AI-Driven Reproducibility and Analytical Robustness in Scientific Discovery [AUTHORS]Shicheng Liu, Krista Opsahl-Ong, Andrew Drozdov, Julian Killingback, Arnav Singhvi, Ronak Pradeep, Michael Bendersky, James C. Dickerson, Monica S. Lam [ABSTRACT]Reproducible and analytically robust research findings are essential to scientific progress, but evaluating these properties typically requires costly manual effort. We investigate whether LLMagents can help automate reproduction and robustness analysis, using a medical case study grounded in the Surveillance, Epidemiology, and End Results (SEER) database, a widely used U.S. cancer registry underlying tens of thousands of published studies. First, we introduce SEER-Repro, a benchmark of 20 published SEER-based papers together with the data and resources required to reproduce their findings. Second, we create an AI agent that successfully reproduces 77.4% of 2,190 atomic claims, while providing executable baselines for further analysis. Third, we develop an agentic analysis pipeline to test the robustness of the medical findings by identifying analytical choices underlying each finding, proposing defensible alternatives, and rerunning the analysis to determine whether those choices affect the conclusion. Across published papers in SEER-Repro, our pipeline identifies potentially consequential analytical choices in 9.5% of fully reproduced findings and flags them for expert review. These results show how LLMagents can support scientific review by reproducing findings, systematically testing their robustness, and directing expert attention toward potentially consequential analytical choices. [LINK]http://arxiv.org/abs/2610.23845v1 [DATE]2026-10-17 01:57:55+08:00 [CATEGORIES]cs.CL
DiscoveryRSI: Recursively Learning to Discover Valuable Hypotheses [AUTHORS]Jianyuan Zhong, Chenwei Wu, Zhaochen Yu, Xixian Jiang, Xuewei Li, Hao Xiao, Xilin Dai, Junyang Wang, Jiaru Zou, Zhiqing Cui, Boyuan Zheng, Jing Xiong, Zhen Zhang, Qiang Gao, Michael Morley, Luis Nakayama, Qiang Xu, Liyue Shen, Tao Gui, Leo Anthony Celi, James Zou, Zhenfei Yin, Yingcheng Wu, Ling Yang [ABSTRACT]Scientific discovery requires developing worthwhile hypotheses into practical investigations, yet learning from this evolving research process remains challenging. We introduce an operational definition of valuable and testable research hypotheses: evidence-grounded claims about consequential uncertainties, coupled with feasible investigations, explicit decision criteria, and alternative explanations. Our DiscoveryRSI framework develops hypothesis trees through multi-agent evolutionary search, compiles selected investigations into executable Harbor tasks, and recursively learns through three mechanisms: hypothesis credit optimization attributes scientific assessments to documented research contributions; harness replay identifies productive strategies and recurring failures to refine research workflows; and rediscovery applies accepted improvements to new investigations, generating further experience for subsequent learning. We also introduce DFM-Bench, which provides comprehensive assessment of hypothesis quality and cumulative research progress at the levels of hypotheses, revisions, development paths and research runs. Across four life-science domains, automated evaluation shows that hypothesis quality rises from 4.86 for the base model to 5.39 after two rounds of self-improvement, with the large gains in falsifiability and data resolvability. We further validate hypothesis quality through expert review, comparing our DiscoveryRSI-27B with other research agents. Preliminary human assessments show that DiscoveryRSI-27B scores higher than GPT-6 Luna on hypothesis importance and whether claims are plausible and evidence-bounded. DiscoveryRSI-27B achieved a 6.3% rate of hypothesis correspondence with research published after the model cutoff, largely outperform GPT-6 with 2.5% and Claude Opus 5 with 2.3%, while delivering greater human-rated improvement along research paths. [COMMENTS]Code: https://github.com/Gen-Verse/DiscoveryRSI [LINK]http://arxiv.org/abs/2610.23839v1 [DATE]2026-10-17 01:56:35+08:00 [CATEGORIES]cs.CL
GLOBE: Evolving Structured Agentic Memory for E-commerce Product Optimization in Generative Search [AUTHORS]Hyunseo Kim, Suil Choe, Sukkyoung So, Daeyeon Yi [ABSTRACT]Optimizing product listings for generative search requires language agents not only to rewrite effectively but also to continually self-improve by retaining successful strategies across diverse categories. However, existing approaches often treat structured listings as free-form text, leaving effective strategies implicit in prompts, search trajectories, or rewritten outputs rather than accumulating them as a reusable agentic memory. To address this limitation, we introduce GLOBE, an agentic framework that drives continual self-improvement by storing optimization knowledge in an evolving handbook of explicit, field-scoped behaviors. In each episode, an Editor applies relevant behaviors and receives rank-based feedback, which a Reflector converts into behavior updates. Crucially for dependable execution in real-world settings, a Moderator removes redundant proposals, a Curator commits accepted changes, and consistently harmful behaviors are deactivated. This cycle transforms episodic feedback into a persistent, field-scoped optimization memory. Experiments show that structured behaviors outperform monolithic prompt optimization, while feedback-driven evolution and moderation further improve effectiveness and reduce downside risk. The learned behaviors also remain effective on unseen product categories, suggesting that agentic memory evolution supports interpretable and transferable product optimization. [COMMENTS]Accepted to the Second Workshop for Research on Agent Language Models(REALM) at EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23709v1 [DATE]2026-10-17 00:33:17+08:00 [CATEGORIES]cs.CL
Coordination Matters: Process-Aware Evaluation of Cooperative Multi-Agent Reinforcement Learning [AUTHORS]Maria Ana Cardei, Matthew Landers, Afsaneh Doryab [ABSTRACT]Cooperative multi-agent reinforcement learning (MARL) benchmarks commonly emphasize aggregate outcomes such as return, success rate, or completion time. While essential, these metrics can obscure how agents coordinate and fail to distinguish policies that achieve similar outcomes through different interaction processes. We propose a process-aware evaluation framework that complements aggregate performance with diagnostics of interference, coordinated task completion, participation, and coordination efficiency. We evaluate the framework in STAT, a controlled spatial task-allocation testbed, and examine its broader applicability in two established cooperative MARL benchmarks. Across diverse value-based and actor-critic methods, we find policies with nearly identical return can exhibit substantial differences in coordination behavior. Controlled scaling in STAT reveals distinct process signatures for spatial, task, and agent scaling, while a reward-shaping ablation shows that explicitly penalizing assignment conflicts can alter coordination behavior with little change in task return. Results in the established benchmarks further show that the framework remains informative under different coordination mechanisms. These findings support process-aware evaluation as a complement to outcome-based benchmarking. The framework provides a common structure for selecting and interpreting environment-specific diagnostics that characterize not only what cooperative policies achieve, but how they coordinate to achieve it. [COMMENTS]Under review [LINK]http://arxiv.org/abs/2605.06557v2 [DATE]2026-10-17 01:58:51+08:00 [CATEGORIES]cs.LG
TANGO: RNA Topology and Geometry Co-Design [AUTHORS]Tianmeng Hu, Biao Luo, Ke Li [ABSTRACT]Designing RNA sequences that satisfy target tertiary-structure objectives is an important problem in RNA engineering, with broad applications in synthetic biology and therapeutics. Existing 3D RNA design methods have made substantial progress, but they largely treat the target fold as a geometry-only objective. This is incomplete for RNA, whose functional folds are specified not only by tertiary geometry but also by secondary base-pairing topology. We therefore formulate RNA inverse design as a topology-and-geometry co-design problem and propose TANGO, a multi-objective multi-agent reinforcement learning framework for this task. TANGO adopts a staged curriculum: it first learns a topology-aware policy from secondary-structure targets, then transfers this policy to tertiary-structure targets to train a general co-design model, and finally applies frontier-guided fine-tuning to refine target-specific trade-offs among secondary-structure fidelity, tertiary-structure fidelity, and sequence diversity. Experiments show that TANGO generates diverse and novel designs, improving tertiary- and secondary-structure performance by 18.3% and 52.7%, respectively, over the strongest baseline. [COMMENTS]NeurIPS 2026 [LINK]http://arxiv.org/abs/2610.20277v2 [DATE]2026-10-17 01:46:35+08:00 [CATEGORIES]cs.LG
CocoRNA: Collective RNA Design with Cooperative Multi-agent Reinforcement Learning [AUTHORS]Tianmeng Hu, Biao Luo, Ke Li [ABSTRACT]Designing RNA sequences that reliably fold into specific secondary structures is essential for understanding their biological functions but remains a challenging computational problem. We propose CocoRNA, a cooperative multi-agent reinforcement learning framework for RNA inverse design. CocoRNA simplifies the design task by decomposing it into smaller sub-problems, each solved collaboratively by multiple agents. This approach reduces the complexity of the problem and improves the exploration of design policies. During training, a centralized critic uses global structural information to guide the agents, enabling them to jointly optimize their design strategies. As a result, CocoRNA learns high-quality RNA design policies that generalize effectively to unseen structures without additional training. Experiments on the Rfam dataset demonstrate that CocoRNA substantially outperforms state-of-the-art methods in both success rate and design speed. Further experiments on other biological sequence design tasks highlight the effectiveness and broad potential of CocoRNA for complex design tasks. [COMMENTS]ICML 2026 [LINK]http://arxiv.org/abs/2610.23820v1 [DATE]2026-10-17 01:43:25+08:00 [CATEGORIES]cs.LG
Learning Personalized Memory Gating from Delayed One-Sided Feedback [AUTHORS]Kaixuan Ji, Zixiang Chen, Wenting Zhao, Cheng Qian, Heng Wang, Weizhi Zhang, Jielin Qiu, Liangwei Yang, Quanquan Gu, Silvio Savarese, Shelby Heinecke, Huan Wang [ABSTRACT]Long-term memory enables language-model agents to personalize future interactions, but a finite memory bank cannot preserve every extracted fact. Existing memory gating policies are typically shared across users and therefore fail to account for user-specific patterns of future reuse. In this paper, we study memory gating at the fact level. We first propose a retrieval-based labeling strategy that labels each fact by whether it is later retrieved and incorporated into responses. We then formulate online memory gating as contextual apple tasting with delayed one-sided feedback, and propose Warm-Start UCB, which combines offline learning from existing data with online adaptation to the current user. Our theoretical analysis establishes a high-probability regret upper bound of $\tilde\{O\}(Tε\wedge d\sqrt\{T\})$ when the delay window is at most $O(\sqrt\{T\})$, where $T$ is the total number of rounds, $ε$ is the initialization error, $d$ is the dimension of the feature map, and $\tilde\{O\}(\cdot)$ hides constant and poly-logarithmic factors. The results of our experiment demonstrate that our method outperforms all baselines under different memory budgets and history lengths, establishing it as a promising personalized memory gating policy. [LINK]http://arxiv.org/abs/2610.23812v1 [DATE]2026-10-17 01:34:45+08:00 [CATEGORIES]cs.LG
Parametric Resonance in Primal-Dual Learning for Slowly Non-Stationary Constrained MDPs [AUTHORS]Maxime Baelde [ABSTRACT]Primal-dual methods for constrained Markov decision processes are analyzed under a two-timescale assumption, with non-stationarity entering through two channels. In the additive channel, where the constraint level oscillates, the dual oscillation saturates at a floor set by the moving optimum and is amplified linearly by the timescale ratio. In the parametric channel, where the constraint sensitivity oscillates, a coherent modulation carries no resonant forcing at any depth: the dual update's stiffness modulation cancels the primal update's damping modulation. Instability instead comes from a mismatch between the constraint estimate used by the primal update and the measurement used by the dual update, producing a parametric resonance at twice the pair's natural frequency. The tolerated mismatch grows with the square root of the forcing bandwidth and collapses with the square root of the dual-to-primal step-size ratio. These results hold on the tangent linear-quadratic reduction around an active saddle, for one constraint and for $K$ constraints under a common modulation and an isotropy condition. Driving both updates from the same estimate removes the resonant forcing but introduces a stationary offset equal to the estimator's bias. In a Lagrangian soft actor-critic agent, the multiplier locks onto the sub-harmonic at the timescale ratios its identified linear loop predicts and a coherent cost critic suppresses the lock-in, while a proportional (PID) term, predicted to leave the lock-in in place, suppresses most of it through the projection of the multiplier. [COMMENTS]33 pages, 3 figures. Code: https://github.com/mbaelde/dualdynamics-code [LINK]http://arxiv.org/abs/2610.23788v1 [DATE]2026-10-17 01:19:08+08:00 [CATEGORIES]cs.LG
GRAFT-ATHENA: Self-Improving Agentic Teams for Autonomous Discovery and Evolutionary Numerical Algorithms [AUTHORS]Juan Diego Toscano, Zhaojie Chai, George Em Karniadakis [ABSTRACT]Language-model agents plan, execute, and evaluate scientific workflows. However, the problem--method relationships guiding their choices remain implicit, so plausible outputs can violate the assumptions of the scientific problem. We introduce Graph Reduction to Adaptive Factored Trees (GRAFT), which makes this problem-to-method map explicit as an expandable probabilistic structure of admissible problems, methods, and dependencies. Because this structure follows scientific semantics, knowledge from documentation, papers, and domain experts can be encoded, while factorization keeps the structure tractable. Semantic fingerprints of encoded problems measure their similarity, so expertise and experience guide related problems. GRAFT-guided workflows exceeded published expert baselines in physics-informed learning, recovered blood-rheology trends, predicted Apollo Command Module hypersonic loading within 1.8\% of experiment, and established machine-checked error bounds for regularized brain-flow reconstruction and a spectral physics-informed architecture. The representation retains evaluated outcomes and incorporates new methods, allowing scientific knowledge and experience to accumulate. [LINK]http://arxiv.org/abs/2605.11117v3 [DATE]2026-10-17 01:10:34+08:00 [CATEGORIES]cs.LG
When Is Language-Model Advice Worth Buying? A Matched-Budget Study of Predictive Gating in Reinforcement Learning [AUTHORS]Ibne Farabi Shihab, Md Najmus Swaqeeb, Abu Sa-Adat Mohamed Moon-Im Al Ahsan, Joyanta Jyoti Mondal [ABSTRACT]Language-model advice can accelerate reinforcement learning, but every call has a price, and a returned suggestion may be unparsable, unsafe to execute, or no better than the agent's own action. We introduce certified predictive advice gating, a cost-aware controller that forecasts possible advisor responses, values the action actually executed after parsing, execution checks, and cache reuse, and requests advice only when a calibrated lower bound on that value covers the price. Our analysis separates the gap of executed advice, cumulative regret under an intervention-stable fallback, and myopic acquisition loss. On BabyAI GoToObj with Qwen2.5-1.5B and 7B advisors, the gate improves return over no advice by paired $0.029\pm0.016$ and $0.030\pm0.015$ while using more than $97\%$ fewer calls than always asking. Matched-budget comparisons, however, do not resolve an advantage over an early query schedule, so the gain reflects economical advice volume rather than demonstrated state selection. Groupwise calibration raises coverage on positive-value opportunities from $0.47$ to $0.85$, still below the $0.90$ target. A separate headroom-aware gate reports paired gains over an early schedule of $0.032\pm0.012$ and $0.029\pm0.011$ on GoToObj and $0.027\pm0.011$ and $0.028\pm0.012$ on GoToObjDoor, and adds BabyAI and ALFWorld success, decision-stratum coverage, and component timing. These results separate return gains, query placement, and execution screening; finite-sample deployment guarantees remain unestablished. [COMMENTS]41 pages, 6 figures [LINK]http://arxiv.org/abs/2609.29548v2 [DATE]2026-10-17 00:49:25+08:00 [CATEGORIES]cs.LG
Tool Recommendation for Multi-Step Agents: Finding Every Tool a Task Needs [AUTHORS]Sam Partee, River Stevens, Thierry Damiba, Valerie Fanelle [ABSTRACT]A multi-step agent task fails when a single tool it needs is missing, yet the tools an agent may need are too many to list: the roughly 8,200 tools on the platform we study have definitions that together exceed the context window of any model we evaluate. We present Tool Recommendation, a deployed system that gives an agent all of them, hundreds of times more than a typical MCP server exposes, through a fixed interface of three tools. It runs at the gateway between agent and tools, executing every call under the end user's own credentials and checking it against the user's consent for each service and the tools the agent may use. We evaluate it on both a public retrieval benchmark and billed end-to-end session cost. On ToolRet, Tool Recommendation places every tool a query needs in the top ten 57% more often than BM25, a lexical ranking that commercial tool search offers, and compares favorably with published systems up to nine times its size. In paired multi-step workflows, its first recommendation covered every required tool in 93% of Codex sessions, and against a fixed list of 100 tools it cut input tokens by 17% with Codex and 40% with Claude Code and the total bill in all 14 runs, by up to 40%, with no change in task completion. [COMMENTS]11 pages, 3 figures, 5 tables [LINK]http://arxiv.org/abs/2610.23707v1 [DATE]2026-10-17 00:30:38+08:00 [CATEGORIES]cs.LG
HardEval: An Execution-Verified Benchmark for AI-Driven Security Hardening and Compliance Remediation [AUTHORS]Manohar Vellala, Soumya Banerjee, Sandip Roy, Ross Gore, Sachin Shetty [ABSTRACT]Implementing Security Technical Implementation Guide (STIG) controls remains a demanding operating-system administration task. Each natural-language security requirement must be translated into a precise and persistent configuration change. These changes span Secure Shell (SSH) configuration, authentication policy, audit rules, and kernel parameters. Existing benchmarks do not measure whether a model can perform this task on a live system. Code benchmarks evaluate isolated programs without persistent system state, while agentbenchmarks largely focus on applications rather than operating-system configuration. In this paper, we introduce \textbf\{HardEval\}, an execution-verified benchmark for evaluating language-model-assisted security hardening. HardEval is built from Red Hat Enterprise Linux 8 STIG controls that fail on an unhardened baseline system. Expert-written reference remediations and Open Vulnerability and Assessment Language (OVAL) checks serve as hidden ground truth for grading; they are never shown to the model, and each task states only the required security outcome while withholding the implementation. The model produces a Bash script, which is executed on a live virtual machine and graded from fail to pass using the corresponding OpenSCAP/OVAL compliance check. Evaluation of fourteen frontier and open-weight configurations reveals large and uneven capability differences across security domains. A recurring failure mode is plausible security code that does not satisfy the exact compliance condition. For four open-weight configurations, multi-run Wilson confidence intervals preserve the observed ordering. HardEval is publicly available. [COMMENTS]8 pages, 4 figures, 4 tables. Code and data: https://github.com/manoharvellala/Hardeval The paper has 4 figure environments (two are images, two are the example-prompt panels) and 4 tables [LINK]http://arxiv.org/abs/2610.23685v1 [DATE]2026-10-17 00:14:58+08:00 [CATEGORIES]cs.LG
2026 Oct 16, Fri
The Chameleon's Limit: Investigating Persona Collapse and Homogenization in Large Language Models [AUTHORS]Yunze Xiao, Vivienne J. Zhang, Chenghao Yang, Weihao Xuan, Jen-tse Huang [ABSTRACT]Large language models (LLMs) have powered multi-agent simulations such as survey respondents or application users. However, LLMs lack the population-level diversity required by these simulations---responses from LLMs with detailed human profiles cluster into classes, unlike human responses that uniformly fill the entire space. In this paper, we formalize this as Persona Collapse and quantify its population-level structure with three diagnostic metrics: Coverage (how much of the persona space is occupied), Uniformity (how evenly agents spread across it), and Complexity (how rich the resulting behaviors are). Across eight LLMs on personality (BFI-44), moral reasoning, and self-introduction, persona collapse is universal but task-dependent: CoSER-Llama-8B collapses the most in personality while being the most diverse in moral reasoning. Moreover, sampling more personas cannot mitigate the collapse. We further show a fidelity--diversity tradeoff: the models with the highest per-persona fidelity produce the most polarized populations. Finally, persona-targeted fine-tuning further narrows the population instead of diversifying it. [LINK]http://arxiv.org/abs/2604.24698v2 [DATE]2026-10-16 23:30:51+08:00 [CATEGORIES]cs.CL
SIT-Eval: Situation-Based Interactive Turn-by-Turn Evaluation of Conversational Agents [AUTHORS]Ravil Mussabayev, Saveliy Zavertyaev, Dmitriy Parpulov, Andrey Savchenko [ABSTRACT]The rapid proliferation of conversational AI agents necessitates evaluation methods that assess performance across diverse dialogue scenarios, utilizing multiple situational criteria, as appropriate behavior depends on context and encompasses various aspects. Existing dialogue benchmarks typically use static natural conversations, which underrepresent rare situations---a critical safety gap, as low-probability events become frequent failures at production scale. However, limited sample sizes force a choice: either maintain natural dialogues for realism or turn to adaptive situation sampling for comprehensive yet statistically reliable coverage. We introduce SIT-Eval, an adjustable framework for evaluating situation-based conversational agents. We formalize an agent's desired behavior as situational rules: if situation is P, follow rules Q. At each conversation turn, an LLM-interlocutor enacts a uniformly sampled situation. A context-aware LLM-judge leverages all conditionally applicable rules as criteria to evaluate the agent's reply for that situation. The judge is validated using human-verified synthetic golden dialogues. Benchmarking 8 LLMs on 1292 conversation turns reveals failure modes hidden by aggregate metrics: models with similar overall scores vary widely in their adherence to individual rules. All models struggle most with ambiguity clarification and Socratic questioning. [LINK]http://arxiv.org/abs/2610.23612v1 [DATE]2026-10-16 23:24:02+08:00 [CATEGORIES]cs.CL
Execution Before Memory: Diagnosing Memory Failures in Video World Models [AUTHORS]Weiyang Jin, Jiwen Yu, Sihan Xu, Yuwei Niu, Xihui Liu [ABSTRACT]A memory test for a video world model should say whether the scene was set, whether the asked change was carried out, and whether the last frame shows the result. Current tests fold these into one number, so a low score is easy to read as forgetting. When the change is not carried out, the hidden state transition cannot appear, so the test is not really a test of memory. We introduce MIST, an evaluation that separates the change from the memory it should leave behind. The model first sets the scene while the change stays hidden, and only later is asked to show it, with leakage before that point kept under control. This separation locates the bottleneck: the model sets the scene but does not carry out the instructed change. The same gap appears in general memory evaluations. On that basis we propose M3-Agent (Make the Missing Move), which recursively revises the details of the input instruction so that the model successfully carries out the requested instruction. Relative to the baseline, it more than triples the MIST reveal score. On the benchmarks, WRBench's score T rises by 26.1% and MBench-T's score U by 52.7%. The method meets the problem from the execution side: it shows what the frozen model can do once the move is stated, and it sets a stricter baseline. [COMMENTS]27 pages, 7 figures. Project page: https://hku-mmlab.github.io/MIST/ Code: https://github.com/HKU-MMLab/MIST [LINK]http://arxiv.org/abs/2610.23542v1 [DATE]2026-10-16 22:37:13+08:00 [CATEGORIES]cs.CL
MuninAI: A Self-Hosted Agentic Framework for Independent Research Groups [AUTHORS]Markus Fischer, Jens Meiler, Matthias Elgeti [ABSTRACT]Research groups that want a language model grounded in their own literature face a choice between hosted services that receive their unpublished work and self-hosted deployments that keep the data local but give the model no access to the corpus. Munin is an open-source framework for a self-hosted scientific assistant that indexes a group's corpus and its citation graph, and wraps whichever open-weight model the group can serve in an agent harness with specialised reading, search, and computation tools. We claim that the harness, rather than the specific model, is what makes the system useful on domain literature, and we ship the evaluation suite that tests this claim with the software.
On 199 paired literature questions, the harness more than doubles the accuracy of the base model, from 0.387 to 0.874 (+0.487, 95% CI [0.407, 0.568]), while naive retrieve-then-answer scores below the unaided model on every backbone tested. Under the harness, roughly twice as many answer claims trace to a passage the model read itself as with fixed top-5 retrieval. When the source paper for a question is removed from the corpus, the system abstains on about nine in ten of the questions it had answered correctly on the Qwen backbones and eight in ten on gpt-oss-20b, and it produces no fabricated local citation across 300 fabricated references.
The pattern holds on three backbones from two developers, including a 20-billion-parameter model that runs on a single consumer-grade GPU. With all outbound network access disabled, both Qwen backbones still exceed the published PaperQA2 accuracy. The evaluation suite runs as a single command, so an adopting group can re-run it on its own corpus and model. [COMMENTS]37 pages, 3 figures, 20 tables. Code, evaluation suite and scorecards: https://github.com/elgetilab/munin (archived release v1.0.0: https://doi.org/10.5281/zenodo.23191404) [LINK]http://arxiv.org/abs/2610.23480v1 [DATE]2026-10-16 21:56:37+08:00 [CATEGORIES]cs.CL
Actions with Receipts: Jointly Binding Claims, Evidence, and Execution for Replayable Tool-Agent Auditing [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Yina Sa, Daren Zha [ABSTRACT]Tool-using agents can expose citations and execution logs while leaving a critical association unaudited: whether the claim shown to a user is the claim emitted by the committed execution and supported by the cited source. A valid citation and a valid trace can therefore remain individually well formed while being transplanted across claims, actions, runs, or source versions. We introduce a claim-anchored execution contract that jointly binds the emitted claim, its exact source span, the ordered execution prefix that produced it, and the source version and access state observed by that execution. Each receipt contains an emission anchor that deterministically locates the claim inside a committed answer or claim-bearing action, together with source identifiers, offsets, hashes, quotes, and a domain-separated execution commitment. A deterministic integrity verifier reconstructs these bindings before semantic or task labels are joined. We separate this integrity plane from a pluggable support plane, so structural validity is not used as a proxy for entailment. The contract exposes seven independently testable properties: claim-emission binding, source binding, ordered-execution binding, oracle separation, persisted-object replay, execution-rerun consistency, and version/access binding. Across 1,280 cross-object attacks, the joint contract detects 1,275 substitutions (0.9961). Removing a targeted property reduces its attack-detection rate to 0.0156-0.0625. On an independently adjudicated 384-pair split, the conflict-aware support guard reaches F1 0.8865 and false acceptance 0.0729; on unseen failure families, these rates are 0.8679 and 0.0938. [COMMENTS]35 pages, 8 figures [LINK]http://arxiv.org/abs/2610.00327v2 [DATE]2026-10-16 20:28:48+08:00 [CATEGORIES]cs.CLcs.LG
Robust Function Calling for LLMAgents Under Interface Drift [AUTHORS]Aniruddha Chattopadhyay, Kanishka Halder, Partha Sarathi Roy [ABSTRACT]Tool-augmented language-model agents rely on interface text: tool descriptions, argument schemas, and the candidate tool list shown at inference time. These interfaces often change after deployment even when the underlying tool semantics remain fixed, creating a practical reliability problem for function-calling systems. We study this failure mode as interface drift and evaluate whether a lightweight, training-free repair layer can recover robustness without retraining the base model. Our method, SchemaShield-Lite, combines schema validation, candidate-set retry for missing tool names, and one-shot constrained repair. On Qwen3.5-9B, it raises DICE-300 accuracy from 0.717 to 0.813 and BFCL-200 accuracy from 0.580 to 0.805, recovering 19/22 and 33/45 drift-induced failures with no observed repair harms. Naive retry does not explain the gains. Ablations show that candidate-set drift, especially distractor tools, accounts for most observed damage, while description and schema edits alone are less harmful. The results position validation-and-repair as a practical robustness layer for deployed LLM-agent tool use. [COMMENTS]16 pages, 1 figure, 6 tables. Peer-reviewed and accepted at NGEN-AI 2026 (International Conference on Next Generation AI Systems). This is the pre-review submitted version [LINK]http://arxiv.org/abs/2610.23342v1 [DATE]2026-10-16 19:49:37+08:00 [CATEGORIES]cs.CL
RubricEM: Meta-RL with Rubric-guided Policy Decomposition beyond Verifiable Rewards [AUTHORS]Gaotang Li, Bhavana Dalvi Mishra, Zifeng Wang, Jun Yan, Yanfei Chen, Chun-Liang Li, Long T. Le, Rujun Han, George Lee, Hanghang Tong, Chen-Yu Lee, Tomas Pfister [ABSTRACT]Deep research agents plan, search, evaluate evidence, and synthesize long-form reports. Training these agents with reinforcement learning (RL) is challenging because they operate over long horizons and their outputs lack exact answer verifiers. To this end, we introduce RubricEM, a unified RL framework for deep research that combines stagewise credit assignment with reflection-based meta-RL for experience reuse. Rubrics unify the framework by providing a shared interface for policy execution, judge feedback, and experience reuse. RubricEM first instills a rubric-guided scaffold through supervised fine-tuning, teaching the agent to plan with rubrics and organize its trajectory into semantic stages. Building on this structural prior, we optimize the task policy with Stage-Structured GRPO, using stagewise judge feedback for finer-grained credit assignment. In parallel, we asynchronously train a shared-backbone reflection meta-policy (with meta-RL) to distill completed trajectories into reusable guidance, contributing through both shared-parameter updates and textual memory. The resulting RubricEM-8B model achieves strong performance across four long-form research benchmarks, outperforming existing open models of larger sizes and approaching proprietary deep-research systems. Beyond final performance, we perform thorough controlled analyses to understand the key ingredients of RubricEM. [COMMENTS]67 pages, 6 figures [LINK]http://arxiv.org/abs/2605.10899v2 [DATE]2026-10-16 19:35:29+08:00 [CATEGORIES]cs.CLcs.LG
Annotrace: A Workbench for Step-Level Annotation of Agent Traces [AUTHORS]Arun Josephraj Arokiaraj, Chenyi Huang, Sahan Bulathwela, Zekun Wu, Bin Wu [ABSTRACT]Every study of why LLMagents fail begins by labelling traces by hand, and every such study builds its own throwaway tool. Observability platforms record agent traces but offer no structured labelling; annotation tools label text but have no notion of a step, a delegation or a tool call, leaving it unclear what the smallest unit of execution to label should be. Annotrace reads six trace formats, including OpenTelemetry GenAI, LangSmith and Langfuse, and turns each run into an ordered, nested list of steps. The annotator writes their own labelling codebook; a frontier model proposes a starting set from a sample of their own traces, answers first, and is asked exactly the questions the human is asked, making annotation a joint human-AI process. Annotrace measures agreement between any two annotators, human or AI. On a multi-agent question-answering corpus it reproduces a previous manual effort's 235 annotation units exactly, and two models labelling the same traces agree 91% of the time at $κ= 0.82$. [COMMENTS]4 pages, 4 figures, 1 table. Under review at the AAAI 2027 Demonstration Program [LINK]http://arxiv.org/abs/2610.23284v1 [DATE]2026-10-16 18:55:26+08:00 [CATEGORIES]cs.CL
Sample, Mix, or Verify? Allocating Test-Time Compute under Imperfect Verifiers [AUTHORS]Srinivasan Iyengar [ABSTRACT]A common way to make language models and agents more accurate is to spend more compute at inference time: generate many candidate solutions and submit the one a verifier prefers. Success then requires that some candidate is correct and that the verifier picks it. The first grows predictably with the number of samples; the second does not, because real verifiers (unit tests, majority votes, LLM judges, reward models) accept wrong answers, reject right ones, and make correlated mistakes. With a fixed budget, a practitioner must decide how many samples to draw, which models to draw them from, and how much to spend on verification, and today these choices are made by rules of thumb. We introduce SPLIT, which makes them from a small labelled pilot. It rests on an exact formula for the probability of submitting a correct answer under accept/reject and top-score selection that accounts for varying task difficulty, blind spots shared across models, and correlated repeated verifier calls. SPLIT fits the formula by empirical Bayes and uses it to plan budgets on unseen tasks. Across seven benchmarks spanning math, science QA, code and software-engineering agents, 26 models, 175 agent systems and six kinds of real verifiers, all from public releases, SPLIT lands within 0.02-0.8 percentage points of the best allocation in hindsight, while every fixed rule we test loses at least 2.5 points in some setting. Regret depends only on how well SPLIT ranks configurations, and it stays near-optimal under compute, price and latency costs and online. The formula also explains when asking more LLM judges lowers precision. As a forecaster far beyond the pilot, SPLIT beats power-law fitting when the verifier caps success, and we report where it loses. [COMMENTS]27 pages (18 main + appendix), 14 figures, 11 tables. Code will be released [LINK]http://arxiv.org/abs/2610.23255v1 [DATE]2026-10-16 18:28:07+08:00 [CATEGORIES]cs.LGcs.CL
Ignored, Not Refused: What Low Prompt-Injection Success Rates Hide in Small Tool-Using Agents [AUTHORS]Farhan Zarif, Kazi Shahid Ahmed Galib, Synthia Akter Promi, Shabnam Sharmin Tarin, Ariyan Hossain, Sadif Ahmed, Farig Yousuf Sadeque [ABSTRACT]Tool-using agents read untrusted external content, which exposes them to indirect prompt injection (IPI), where hidden instructions hijack the agent's tool calls. Benchmarks measure IPI security by the attack success rate (ASR), which treats every failed attack as a successful defense. To test this assumption, we audit 33,000 trajectories from eleven open-weight agents (2-14B parameters) on InjecAgent and AgentDojo. A label-blind LLM judge shows that most failed attacks are ignored (76-80%) rather than refused (17-18%). With identical model weights, changing only the system prompt shifts InjecAgent ASR from 0.2% to 43%. Our Attack-Task Outcome Matrix (ATOM) further shows that in 21.7% of AgentDojo episodes, the agent stayed safe but did not complete the user's task. A low ASR therefore mostly reflects agents silently ignoring injections, not refusing them. Evaluations should report how attacks fail and whether the user's task is completed, alongside ASR. We release the labeled corpus, judge prompt, and analysis code. [COMMENTS]17 pages, 3 figures, 11 tables [LINK]http://arxiv.org/abs/2610.23254v1 [DATE]2026-10-16 18:26:23+08:00 [CATEGORIES]cs.CL
Think More or Read More? A Token-Budget Theory of Reasoning with Shared Context [AUTHORS]Manh Nguyen, Sunil Gupta, Hung Le [ABSTRACT]Reasoning systems increasingly improve answers by incorporating additional context into the model's input, from multi-agent debate, where agents read their peers' responses, to self-refinement, where a model revisits its own drafts. In these systems, communication consumes context as one agent's output becomes another's input. We study the regime where token reading and generation draw on one fixed token window, so every token spent reading is a token that cannot be spent reasoning. We develop a \textbf\{token-budget theory\} that formalizes this trade-off, modeling accuracy as a function of the reasoning budget remaining after reading and aggregating votes with a Condorcet jury model that permits correlated errors. For multi-agent debate, the theory gives a closed-form prediction for the optimal number of agents. This quantity is estimated from single-agent measurements, allowing the optimal group size to be determined without running debate, and the same accounting extends to a heterogeneous set of different models. We empirically verify our theory and show that, at the predicted group size, our method outperforms strong baselines under the tightest calibration spend, where debate-calibrated rules can afford only one or two questions. Self-refinement obeys the same accounting, with its own draft in place of a peer message. The theory further predicts that communication pays only when the gain in the accuracy ceiling exceeds the reasoning room spent on reading, with this condition tightening as the window shrinks. Across five open models and four reasoning tasks, self-consistency matches or beats deployed debate at $93$\% of $980$ matched window-and-group-size settings, with the gap being largest under tight windows and nearly disappearing at a loose one, consistent with the theory. Under a shared window, how much room a model has left to \textit\{think matters more than what it reads\}. [LINK]http://arxiv.org/abs/2610.23246v1 [DATE]2026-10-16 18:18:15+08:00 [CATEGORIES]cs.CLcs.LG
MultiAgent-OPD: End-to-End Training of Hierarchical Multi-Agent Systems with On-Policy Distillation [AUTHORS]Yuhan Sun, Jiaxin Zhao, Yongdong Luo, Yibo Hu, Meiguang Jin, Junfeng Ma, Weixun Wang, Weihang Pan, Zhongan Bi [ABSTRACT]In a hierarchical multi-agent system, a Leader assigns tasks and subagents carry them out. Jointly training the Leader and the subagents is difficult because a final answer gives only indirect feedback to the subagents. An LLM judge can score the performance of the subagents, but these scores depend on the judge's own assessment of the quality of subagents' responses. We present MultiAgent-OPD, which combines on-policy distillation with reinforcement learning to jointly train a shared-parameter Leader-Searcher-Answerer team end to end. We first train a Teacher in a Searcher-Answerer workflow using answer correctness. During team training, the frozen Teacher scores student-generated search and summary tokens to provide local process supervision for the Searchers, while the Leader learns from team success and the Answerer from candidate correctness. Across seven datasets, the trained system reaches 38.68% macro avg@4, exceeding Prompted Critic by 4.06 percentage points and C-MAS + RL, the trained cascaded multiagent system, by 2.61 percentage points. The team also solves more questions within four attempts. [LINK]http://arxiv.org/abs/2610.23234v1 [DATE]2026-10-16 18:01:52+08:00 [CATEGORIES]cs.CL
MAS-Bai: Linguistically Informed Adaptive Multi-Agent Translation for Bowen-Script Bai [AUTHORS]Jingting Zheng, Hong Ren, Yuehan Cui, Shaolin Zhu, Deyi Xiong [ABSTRACT]Large Language Models (LLMs) remain unreliable for extremely low-resource languages whose writing systems conflict with high-resource pretraining priors. We study Bai-Chinese translation in the Bowen script, where Chinese characters are frequently borrowed phonetically rather than semantically, causing severe literal-translation hallucinations. We propose MAS-Bai, a linguistically informed adaptive multi-agent framework for Bowen-script Bai-Chinese translation. MAS-Bai routes each input through either a lightweight retrieval-translation path or a deeper philological path with etymological and syntacticagents, and uses cycle-consistency with a lightweight reverse translator as an inference-time reliability signal. We construct a 6,536-sentence Bai-Chinese corpus and evaluate on a 1,536-sentence test set. MAS-Bai achieves 16.23 BLEU and 62.15 COMET, improving over Standard RAG by 8.11 BLEU and 6.85 COMET. Forward LoRA, expert ablations and controlled routing experiments distinguish the roles of supervised adaptation, Bai-specific expertise and conditional routing. The largest gains occur on Bowen phonetic loans and divergent Bai word order, while human evaluation and reverse-verifier analyses support the automatic results. [COMMENTS]14 pages, 3 figures, Findings of EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23205v1 [DATE]2026-10-16 17:43:32+08:00 [CATEGORIES]cs.CL
AMEND: Constraint-Guided Multi-Agent Text Detoxification with Isolated Critics [AUTHORS]Tianyu Wang [ABSTRACT]Text detoxification requires balancing toxicity removal, semantic preservation, and fluency. Existing methods typically optimize an unconstrained scalar objective---enabling safety scores to mask semantic loss---or rely on open-ended iterative debate prone to semantic drift and high computational cost. We propose AMEND, an inference-time framework that decouples detoxification into role-conditioned proposal generation, isolated multidimensional criticism, and constraint-first selection. AMEND evaluates diverse candidates with role-blind critics and enforces strict semantic and fluency feasibility before ranking by utility. To maximize efficiency, AMEND Lite restricts subsequent search to at most one targeted repair round only when residual toxic spans remain. On ParaDetox, one-round selection over diverse roles improves the joint evaluation metric by $+0.0508$ over the best single role. On a frozen confirmation test set, AMEND Lite achieves a joint score of $0.6597 \pm 0.0062$, outperforming unconstrained multi-round loops by $+0.0489$ ($p < 0.05$) while reducing inference requests by 28.4%. [COMMENTS]10 pages, 2 figures, 4 tables [LINK]http://arxiv.org/abs/2610.23183v1 [DATE]2026-10-16 17:22:21+08:00 [CATEGORIES]cs.CL
Tree-Aware Posterior Distillation for Long-Horizon Agentic Reinforcement Learning [AUTHORS]Heming Zou, Qi Wang, Yixiu Mao, Yun Qu, Yuyi Guo, Zeyuan Liu, Chenglin Li, Weijie Liu, Saiyong Yang, Xiangyang Ji [ABSTRACT]Reinforcement learning with verifiable rewards has become a scalable post-training paradigm for large language models. However, training agents over long-horizon interactions presents two challenges: using only terminal rewards makes turn-level credit assignment sparse and high-variance, while sequential mini-batch updates on a fixed rollout batch introduce policy staleness as the current policy drifts from the behavior policy. Existing approaches obtain denser credit via learned critics, process reward models, or on-policy distillation, but they rely on a value network, step-level annotations, or a stronger external teacher that increase training complexity. Meanwhile, importance sampling and clipping mitigate policy drift, but they do not specify a finite target that stops sequential updates from overshooting. In this work, we introduce TreeOPSD, a critic-free, tree-aware posterior self-distillation framework for long-horizon agentic reinforcement learning. Treating each interaction turn as a tree node, TreeOPSD branches alternative continuations from shared prefixes and recursively backs up terminal rewards into implicit turn-level credit. It then maps each credit to a closed-form, bounded teacher distribution parameterized by the likelihood ratio relative to the frozen behavior policy. Minimizing the divergence between this teacher and the student policy recovers the policy-gradient direction as the tangent at the behavior policy while bounding updates toward a finite target across sequential mini-batches. Empirically, TreeOPSD improves agentic capabilities across several competitive benchmarks and model scales, especially for long-horizon tasks, e.g., raising the average accuracy of Qwen3-30B-A3B-Thinking-2507 by 4.97 points over GRPO on the Deep Research setting. [LINK]http://arxiv.org/abs/2610.23181v1 [DATE]2026-10-16 17:17:26+08:00 [CATEGORIES]cs.LGcs.CL
JAXtari: High-Throughput and Easy-to-Modify Arcade Learning Environment [AUTHORS]Quentin Delfosse, Raban Emunds, Paul Seitz, Sebastian Wette, Jannis Blüml, Daniel Kirn, Dominik Mandok, Kristian Kersting [ABSTRACT]The Arcade Learning Environment (ALE) has been a cornerstone of Deep Reinforcement Learning research, providing a standardized benchmark for evaluating general agents. However, its lack of native support for object-extraction and environment modification restricts research into evaluating whether agents learn robust causal models or rely on brittle visual heuristics. Furthermore, the ALE's dependence on CPU-based emulation creates a significant bottleneck in modern GPU-accelerated training pipelines, limiting both experimental scale and iteration frequency. In this work, we introduce JAXtari, an open-source, high-performance reimplementation of Atari 2600 environments. By leveraging massive GPU parallelization, JAXtari allows PPO agents to finish training on 100M environment steps on a single A100-GPU in under an hour. Beyond speed, it provides object-centric observations and a flexible modification system, allowing researchers to, e.g. develop neuro-symbolic methods or discover novel shortcuts and subsequently train and adapt their agents rapidly, sometimes in only minutes. [LINK]http://arxiv.org/abs/2610.23650v1 [DATE]2026-10-16 23:48:00+08:00 [CATEGORIES]cs.LG
QD-Rec: Open-Ended Evolution of Recommendations via Quality-Diversity [AUTHORS]Takashi Ishida, Taro Makino, Issa Sugiura [ABSTRACT]Open-ended search combined with foundation models has produced advances such as faster algorithms, new results on open mathematical problems, and self-improving agents. We extend this paradigm to recommendation for decision-making, focusing on decisions where the set of possible choices is fixed but the reasons for choosing them are not. In high-stakes decisions, e.g., which company to acquire or which customer to target, comparing diverse, well-reasoned options can be more useful than a single ranked list. We formalize open-ended recommendation as building a diverse, high-quality set of recommendations, each an item together with a hypothesis, a natural-language reason why the item fits the needs of a decision maker. We propose QD-Rec, a quality--diversity method that selects items with a contextual bandit and generates hypotheses with a foundation model from earlier ones kept in an archive as stepping stones. A new recommendation enters the archive if it is novel or better than its nearest neighbor. Across M&A, customer targeting, and paper and book recommendation, QD-Rec gives the highest mean quality and builds the largest archive in most settings. [LINK]http://arxiv.org/abs/2610.23470v1 [DATE]2026-10-16 21:49:22+08:00 [CATEGORIES]cs.LG
RECREG: Recovery-based Shielding for Regular Safety Properties in Reinforcement Learning [AUTHORS]Alexander W. Goodall, Francesco Belardinelli [ABSTRACT]In many real-world scenarios we need to consider formal requirements, such as safety and constraint compliance, to deploy Reinforcement Learning (RL) agents. In this paper we introduce RecReg, a safe RL framework capable of enforcing non-Markovian regular safety properties via recovery-based shielding. RecReg is a unified approach that supports different levels of a priori knowledge and safety guarantees. The paper primarily focuses on building the theoretical foundations for RecReg in the tabular setting, providing rigorous safety and convergence guarantees that are not present in related work. We provide an extensive set of empirical results using the MASA-Safe-RL library, and we outline an extension of RecReg to deep RL domains using SOTA model-based RL. Our empirical results demonstrate that RecReg achieves efficient learning and matches the theory in this paper. [LINK]http://arxiv.org/abs/2610.23445v1 [DATE]2026-10-16 21:28:34+08:00 [CATEGORIES]cs.LG
2026 Oct 17, Sat
Acceptance Is Not Speed: Length Inflation in Training-Free Lossy Speculative Decoding [AUTHORS]Chiatzen Wang, Bill Xu, Haochen Gu, Hongyang Zhang [ABSTRACT]Speculative decoding typically characterizes speedup by draft tokens accepted per verification round. Lossless verification preserves the target model's output distribution and thus expected completion length, making this metric a meaningful efficiency proxy. Relaxed (lossy) verification accepts tokens that lossless verification would reject, potentially altering that distribution. We identify a length inflation phenomenon: relaxed verification lengthens completions, primarily by extending reasoning traces, largely offsetting apparent gains in accepted tokens per round. We evaluate existing training-free relaxed verification rules on twelve target-drafter pairs against lossless verification on identical prompts, measuring completion length, verifier rounds, wall-clock time, and task accuracy. At each rule's most relaxed evaluated setting, relaxed verification on average accepts $1.34\times$ as many tokens per round but produces $1.54\times$ longer completions. For rules with higher acceptance lengths, tokens accepted only under relaxed verification tend to have low target-model probability, contributing to length inflation. For rules that reduce verifier rounds, accuracy typically drops substantially. Further analysis traces some failures to continued reasoning that exhausts the budget needed to deliver a final answer. Overall, accepted tokens per round alone is an unreliable measure of decoding speed under relaxed verification. Completion length and verifier rounds must also be reported to assess actual efficiency. [COMMENTS]Submitted for NAACL 2027, under review [LINK]http://arxiv.org/abs/2610.23818v1 [DATE]2026-10-17 01:41:19+08:00 [CATEGORIES]cs.CL
DLite: History-free Speculative Decoding for Agentic AI [AUTHORS]Jiaming Xu, Hanzhen Wang, Tianlang Zhao, Jiewen Xiao, Chengze Yuan, Jiayi Pan, Yongkang Zhou, Jiancai Ye, Yu Wang, Guohao Dai [ABSTRACT]Speculative decoding has become a mainstream technique for accelerating large language model (LLM) inference by using a lightweight draft model to generate candidate tokens that are subsequently verified by the target model. However, agentic workloads introduce long and continuously evolving contexts and frequent multi-turn interactions. Under such workloads, we observe that existing draft models suffer from up to an 80% degradation in acceptance length, while maintaining their historical KV cache introduces substantial system overhead. In this paper, we identify the growing historical context as a common source of both problems and introduce DLite, a history-free speculative decoding framework for agentic AI. DLite directly reuses contextual hidden states from the target model and performs local speculative prediction without explicitly maintaining the historical context, transforming an increasingly complex modeling problem into local simple prediction problem. A lightweight sequential head further restores intra-block causal dependencies with low overhead. To effectively train this heterogeneous history-free architecture, we introduce a tailored two-stage training strategy. Experiments on general-purpose, long-context and agentic workloads show that DLite achieves 2-6 acceptance length and outperforms the state-of-the-art speculative decoding method, DSpark, up to 70%. [COMMENTS]14 pages, 5 figures [LINK]http://arxiv.org/abs/2610.23787v1 [DATE]2026-10-17 01:18:54+08:00 [CATEGORIES]cs.CL
2026 Oct 16, Fri
Mitigating Cross-Lingual Cultural Inconsistencies in LLMs via Consensus-Driven Preference Optimisation [AUTHORS]Lucas Resck, Isabelle Augenstein, Anna Korhonen [ABSTRACT]Multilingual large language models (MLLMs) often answer culturally ambiguous questions inconsistently across languages, even for a fixed user persona -- e.g., Shakespeare in English but Cervantes in Spanish for the same British user. To robustly quantify this Cross-lingual Cultural Inconsistency, we introduce Singleton Fleiss's $κ_S$, a metric mathematically resilient to hallucinations. For mitigation, we propose Cross-lingual Cultural Consistent Preference Optimisation (C-3PO), a consensus-driven alignment framework. C-3PO raises $κ_S$ by 0.07--0.10 over unaligned models, above every prompting and representation steering baseline averaged across personas, whilst largely preserving explicit user identities, cultural neutrality and intrinsic cultural knowledge; where it does not, a soft-consensus variant recovers most of the cost. Although trained only on multiple-choice questions, C-3PO also improves consistency in open-ended generation. Empirical evaluations demonstrate this inconsistency is more pronounced for lower-resource languages like Indonesian and Persian. Finally, early decoding of intermediate layers reveals that MLLMs implicitly personalise outputs towards the prompt language's stereotypical culture as forward-pass representations stabilise. [COMMENTS]25 pages, 14 figures, 17 tables [LINK]http://arxiv.org/abs/2605.12515v3 [DATE]2026-10-16 23:46:19+08:00 [CATEGORIES]cs.CL
Trajectory-Triggered Sparse Collaborative Decoding [AUTHORS]Amit Ranjan Trivedi, Sruthikesh Surineni, Kyong Ho Lee, Mohit Mittal, Steven Li, Rex Hill [ABSTRACT]Dense assistance methods such as speculative decoding repeatedly verify draft tokens with a target model, keeping target computation coupled to generation. We study sparse collaborative decoding, where a draft model generates most tokens and a target model is invoked only for short corrective segments. The key problem is deciding when to call the target. We show that local uncertainty is a weak trigger: entropy, negative log-likelihood, logit margin, top-$k$ mass, and velocity features give near-chance failure prediction, with AUC 0.50-0.53. In contrast, aggregate trajectory-shape features over the draft confidence path reach AUC 0.86 using a lightweight logistic predictor trained only from draft correctness labels. This induces a trigger-timing tradeoff: early calls have more leverage but weaker evidence, while later calls have stronger evidence but less ability to redirect the trajectory. Used online, learned sparse collaborative decoding improves over the best open-loop schedule by +5.3 to +20.0 accuracy points across five GSM8K draft-target pairs. On MATH-500, it gains 7.0-13.5 points when early prediction and target competence align, but loses when either is weak. On StrategyQA, it improves over open-loop by 5.0-8.3 points, showing transfer beyond arithmetics. [COMMENTS]17 pages. Accepted at the EMNLP 2026 uncertainty workshop [LINK]http://arxiv.org/abs/2610.23613v1 [DATE]2026-10-16 23:24:05+08:00 [CATEGORIES]cs.CLcs.LG
DominoTree: Conditional Tree-Structured Drafting with Domino for Speculative Decoding [AUTHORS]Saw S. Lin, Jyh-Shing Roger Jang [ABSTRACT]Speculative decoding speeds up large language model inference: a small drafter proposes several tokens, and the target model verifies them in one parallel pass. Drafting a tree lets the target verify many alternative continuations at once. DDTree builds such trees from the block-diffusion drafter DFlash, but scores each position independently of the tokens earlier in its branch; Domino conditions on those tokens but drafts only a single chain. DominoTree combines the two: a training-free best-first tree scored by Domino's path-dependent causal correction, restricted to top-M candidates. In a research harness at temperature zero, it averages 5.33x speedup over autoregressive (AR) decoding on Qwen3-4B and 6.89x on Qwen3-8B, peaking at 8.77x; across temperatures it is 20-26% faster than Domino on Qwen3-4B and 27-36% on Qwen3-8B, and 11-20% faster than the stronger of DDTree and its cost-aware variant CaDDTree. A controlled decomposition attributes +10.1% (Qwen3-4B) and +10.7% (Qwen3-8B) accepted length to the correction, and +4.7% and +4.4% to conditioning on the realized path. Served in SGLang on one RTX 5090, it reaches 4.11x and 4.83x AR at batch size 1, 1.17x and 1.09x the speed of Domino's chain, and up to 1.71x AR on 8K-token HELMET summarization, 14-21% above the chain. The lead holds at 2 concurrent requests but reverses from 8 onward, where the chain sustains more goodput despite DominoTree's 28% higher acceptance. DominoTree accepts more tokens than the chain in every serving setting. On the hybrid Gated-DeltaNet Qwen3.6-27B, correct output requires tree-aware recurrent verification; with it, DominoTree matches the chain at batch size 1 (4.58x versus 4.54x AR) and reaches up to 2.11x AR at long context, 15% above the chain. [COMMENTS]31 pages. Submitted to IEEE Access. v4: adds Qwen3.6-27B results, single-GPU serving re-runs, fused tree-builder kernels and tuned budgets; code and data at https://github.com/slin-zhq/Domino-Tree [LINK]http://arxiv.org/abs/2607.08642v4 [DATE]2026-10-16 22:29:03+08:00 [CATEGORIES]cs.CL
On the Tip of the Tongue: First-Token Cueing as a Behavioral Probe of Recall Failures in LLMs [AUTHORS]Jaewon Sok, Jewon Yeom, Seonghyeon Park, Jeongjae Park, Hwiyeong Lee, Taesup Kim [ABSTRACT]Large language models often answer factual questions incorrectly in closed-book settings, and such errors are usually attributed to missing knowledge. Yet models often encode answers they fail to produce. We show that many closed-book errors are first-token recall failures: the model commits to a wrong first token, but once given the right one it completes the answer and often recognizes it as correct. We study this with First-Token Cueing, a behavioral probe that prefills the response with the first token of the gold answer. Model uncertainty peaks at the first answer token, where it already predicts correctness, and among gold prefixes of any length the first token alone accounts for the largest share of recovered errors. Across 16 open models, this one token corrects between a fifth and a half of the errors on multi-token answers in English QA and improves accuracy in all seven languages of a multilingual benchmark. Most of the errors it corrects are also solved by the model's own sampled reasoning, a criterion for encoded knowledge, and models recognize correct starts they do not produce: they continue a correct start more confidently than an incorrect start of the same answer type and prior probability, yet rarely find it without the gold answer. These results locate a recoverable class of factual errors at the first decoding step, where inference-time methods may gain the most. [LINK]http://arxiv.org/abs/2610.23400v1 [DATE]2026-10-16 20:55:36+08:00 [CATEGORIES]cs.CL
Brain-to-Language Decoding: Paradigms, Signals, Methods, Evaluation, Practical Use and Beyond [AUTHORS]Yiqian Yang, Yiqun Duan, Chenyu Liu, Yiqi Wang, Xinliang Zhou, Chin-Teng Lin, Yu Zhang [ABSTRACT]Brain-to-language decoding translates neural activity associated with language production, internal speech and perception into linguistic or expressive outputs. It offers a route to restoring communication after speech loss and a means of studying how the brain represents language. Advances in neural recording and representation learning have expanded the field from constrained recognition and acoustic reconstruction to text generation, streaming personalised speech and facial animation. This survey synthesises these developments across invasive and non-invasive measurements, drawing on a search without a lower year limit and source-led updates through September 2026. We connect Articulated, Inner and Perceived tasks to the neural populations they engage, the representations available to decoders and the outputs those representations can support. We examine model development, public resources and the evolution of evaluation, and compare published performance and communication costs within their reported protocols. The synthesis identifies complementary routes to progress: phonetic, acoustic and semantic targets preserve different aspects of a message; shared representations support reuse across recording conditions and tasks; and online communication increasingly depends on calibration, feedback and user control alongside decoding accuracy. Shared benchmarks enable algorithmic comparisons, while longitudinal studies reveal the demands of sustained use. We discuss these developments and their remaining limitations, then outline a prospective five-level trajectory from commands and language to meaning, scenarios and bidirectional cognitive exchange [LINK]http://arxiv.org/abs/2609.27650v4 [DATE]2026-10-16 17:37:51+08:00 [CATEGORIES]cs.CL
Do Speech BCIs Need Larger Models? Rethinking Neural Decoding beyond Scaling [AUTHORS]Zehui Feng, Weichuan Wang, Xiaohan Chen, Cuntai Guan, Ting Han [ABSTRACT]Recent speech brain-computer interfaces (BCIs) increasingly rely on large models and complex multi-stage pipelines for neural decoding. However, in speech BCI, where training data are scarce, noisy, and expensive to collect, the benefits of model scaling remain unclear. This work revisits a fundamental question: are large models truly necessary for accurate neural decoding? We propose BEST (Brain-Encoding-Speech-Text), a two-stage framework for efficient neural-to-text decoding. Instead of relying on large generative models or phoneme-centric pipelines, BEST introduces hierarchical intermediate representations and a compact speech-oriented decoder that directly maps neural activity to text. This design preserves essential structure while significantly reducing computational cost, enabling robust generalization in low-resource settings and practical deployment under constrained resources. On the Brain-to-Text '24 and '25 benchmarks, BEST achieves word error rates (WERs) of 6.19% and 1.62%, respectively, using only 3.85% of the parameters required by comparable methods. Beyond these results, we provide three key insights: (1) decoder scaling yields non-monotonic gains; (2) cross-modal alignment is more critical than model size; and (3) overfitting degards sequence-level performance. Representation analysis further reveals a consistent neural-audio-text alignment pathway. [COMMENTS]25 pages, Accepted by NeurIPS 2026 as a Main Track conference paper [LINK]http://arxiv.org/abs/2610.23175v1 [DATE]2026-10-16 17:12:58+08:00 [CATEGORIES]cs.CL
TriRouter: Sparse-to-Dense Electrode-Space Routing for Robust EEG Motor Imagery Decoding [AUTHORS]Ying Zhou, Yaoxin, Mengting Liu [ABSTRACT]Motor imagery brain-computer interfaces require reliable decoding from multichannel EEG, but practical acquisition often yields sparse or corrupted electrode montages. We propose TriRouter, an electrode-space routing framework with discriminative-preservation, mask-robustness, and latent-compensation paths. A validation-calibrated router uses observable channel availability to allocate path contributions, preserving complete-channel discrimination while increasing representation recovery under severe missingness. In BNCI2014-001, TriRouter outperforms standard public backbones and a dynamic-spatial-filtering corrupted-EEG baseline. It improves accuracy over the strongest corrupted-EEG baseline by 3.68, 4.02, and 6.46 percentage points at nominal 25%, 50%, and 75% missing settings, and remains effective under structured regional missingness and additional low-channel dataset evaluation. [COMMENTS]Submitted to IEEE ICASSP 2027 [LINK]http://arxiv.org/abs/2610.23665v1 [DATE]2026-10-16 23:59:00+08:00 [CATEGORIES]cs.LG
Deep LLM layers expose passage-level context in EEG during silent reading [AUTHORS]Ingo Marquardt, Anthilia Alchanat, Priyanka Jain [ABSTRACT]In a previous report we showed that a decoder can recover open-vocabulary word-level information from 19-channel EEG recorded during silent reading, by aligning short EEG windows with hidden-state embeddings of the presented word taken from a large language model (LLM). That analysis contrasted two of the model's 33 layers: the context-free input embedding and one mid-depth layer. Here we ask how EEG-decodable structure is distributed over the full depth of the language model. We trained one decoder per layer of Llama-3.1-8B, with and without a causal transformer over the sequence of EEG windows. We distinguish word-level decodability from a context-tracking component. Word-level decodability peaks in early layers and declines steadily towards the last layer. In contrast, the context-tracking component of the retrieval gain rises with depth, both with and without a causal transformer. Unexpectedly, a decoder that sees a single 0.5-s window at a time tracks the text passage too. Two new controls dissect this component. A target-geometry control, a synthetic predictor made of the true target plus noise calibrated to each decoder's word-level accuracy, shows that context tracking is not a by-product of word decoding and target geometry. A control that permutes predictions across runs from either the same or from different sessions shows that most of the observed context-tracking is not caused by an EEG recording session signature. Instead, context tracking mostly follows the text. A single EEG window therefore carries passage-level information that a contextual target can absorb. The results locate word identity in the early layers of the language model, discourse state in the deeper layers, and make the choice of target layer a design dial for EEG language decoders. [COMMENTS]41 pages (including 23 pages of Supplementary Material), 12 figures (including 5 in Supplementary Material) [LINK]http://arxiv.org/abs/2610.23638v1 [DATE]2026-10-16 23:41:19+08:00 [CATEGORIES]cs.LG
GRAFT-EEG: Gauge-Aware Physical Field Transport for Cross-Dataset Motor-Imagery EEG [AUTHORS]Ying Zhou, Mengting Liu [ABSTRACT]Cross-dataset EEG decoding often treats montage, reference, and active-sensor changes as representation shift, although they also alter the observation operator. We study target-statistics-blind transfer: fitting receives no target signal, label, or empirical statistic, and electrode coordinates only instantiate a fixed prediction operator. method removes the active-set common-reference gauge and transports each native mask through a template motor leadfield. A source-fitted spectral classifier anchors the decision; a bounded state-space-model (SSM) residual is secondary. We establish restricted identifiability and an operator-perturbation bound. Across five leave-one-dataset-out targets (253 participants), dynamic method achieved 60.865% equal-target balanced accuracy versus 52.684% for strict-clean 5/5 BIOT (hierarchical difference +8.182 points; 95% CI +6.382 to +10.196; positive in every target). On four EEGPT-clean targets, the descriptive means were 61.520% and 59.057%. In four post-hoc exact same-parent tests, FULL exceeded coordinate-basis, matched-spectrum, row-shuffled, and wrong-anatomy controls by 9.959, 9.698, 11.241, and 3.985 points, respectively (all 95% CIs excluded zero; Holm-adjusted P<0.001). A separate 60-participant same-paradigm evaluation retained a positive physical-operator contrast (+4.029 points; +2.338 to +5.773), whereas dynamic--static was inconclusive; its preregistered status is partial confirmation rather than unrestricted external validation. The SSM added +1.059 points (+0.013 to +1.892). Claims remain bounded to five binary motor-imagery domains and the tested operators. [COMMENTS]Under review at ICLR 2027 [LINK]http://arxiv.org/abs/2610.23620v1 [DATE]2026-10-16 23:30:37+08:00 [CATEGORIES]cs.LG
Cross-Dialect Generalization Without Retraining: Benchmarks and Evaluation of Schema-Derived Constrained Decoding for MLIR [AUTHORS]Plawan Kumar Rath [ABSTRACT]Multi-Level Intermediate Representation (MLIR) underlies modern ML compiler infrastructure (TensorFlow, JAX/StableHLO, PyTorch Inductor, IREE), yet appears only in trace amounts in code-LM pretraining corpora. MLIR is also extensible by design: new dialects ship per application domain, so a fine-tuned model per dialect does not scale. We ask whether inference-time priors derived mechanically from each dialect's Operation Definition Specification (ODS) can substitute for gradient-based adaptation. First, we release four natural-language-to-MLIR benchmarks across three dialects (MLIR-Spec-150, Linalg-Spec-30, StableHLO-Spec-30, StableHLO-Held-Out-200), totaling 410 in-scope NL-to-MLIR pairs, plus a 25-program out-of-grammar stress set, with functional references on 30 Spec prompts, under Apache-2.0 with Gebru datasheets and Croissant 1.1 metadata. Second, we build a three-layer schema-derived constraint stack: a CFG over op signatures (C1), type-domain splits from an ODS-extracted type lattice (C2), and an SSA-scope validator driving five-retry rejection sampling (C3). Porting it to StableHLO required no new constraint-layer code. On structurally dominated dialects, SmolLM2-1.7B with token-level constraints matches or exceeds 15B-34B code LMs running post-hoc rejection at a matched retry budget and 8-25x the per-generation speed: on linalg it reaches 80.0% verify-valid (three runs, n=125), 21 to 44 pp above CodeLlama-34B, Granite-Code-34B, and StarCoder2-15B with non-overlapping CIs. The comparison is between systems, not models: under the baselines' post-hoc protocol, SmolLM2 falls from 52.0% to 24.5% on arith+func. On arith+func and the templated StableHLO-Held-Out-200 the baselines match or beat the SLM; we scope these as non-win cells. We release benchmarks, decoder, every per-prompt generation, and a pinned reproducibility Dockerfile. [COMMENTS]Accepted to the NeurIPS 2026 Evaluations and Datasets Track; camera-ready version. 31 pages (10 main + references + appendix + checklist). Code: https://github.com/plawanrath/slm-to-mlir-constrained-emitter Datasets: https://huggingface.co/collections/plawanrath/cross-dialect-mlir-benchmarks-neurips-2026-e-and-d [LINK]http://arxiv.org/abs/2607.18254v2 [DATE]2026-10-16 22:07:14+08:00 [CATEGORIES]cs.LG
2026 Oct 17, Sat
Beyond Consensus: Perspectivist Modeling and Evaluation of Annotator Disagreement in NLP [AUTHORS]Yinuo Xu, David Jurgens [ABSTRACT]Annotator disagreement is widespread in NLP, particularly for subjective and ambiguous tasks such as toxicity detection and stance analysis. While early approaches treated disagreement as noise to be removed, recent work increasingly models it as a meaningful signal reflecting variation in interpretation and perspective. This survey provides a unified view of disagreement-aware NLP methods. We first present a domain-agnostic taxonomy of the sources of disagreement spanning data, task, and annotator factors. We then synthesize modeling approaches using a common framework defined by prediction targets and pooling structure, highlighting a shift from consensus learning toward explicitly modeling disagreement, and toward capturing structured relationships among annotators. We review evaluation metrics for both predictive performance and annotator behavior, and noting that most fairness evaluations remain descriptive rather than normative. We conclude by identifying open challenges and future directions, including integrating multiple sources of variation, developing disagreement-aware interpretability frameworks, and grappling with the practical tradeoffs of perspectivist modeling. [COMMENTS]Accepted to EMNLP 2026 [LINK]http://arxiv.org/abs/2601.09065v3 [DATE]2026-10-17 00:39:26+08:00 [CATEGORIES]cs.CL
GLOBE: Evolving Structured Agentic Memory for E-commerce Product Optimization in Generative Search [AUTHORS]Hyunseo Kim, Suil Choe, Sukkyoung So, Daeyeon Yi [ABSTRACT]Optimizing product listings for generative search requires language agents not only to rewrite effectively but also to continually self-improve by retaining successful strategies across diverse categories. However, existing approaches often treat structured listings as free-form text, leaving effective strategies implicit in prompts, search trajectories, or rewritten outputs rather than accumulating them as a reusable agentic memory. To address this limitation, we introduce GLOBE, an agentic framework that drives continual self-improvement by storing optimization knowledge in an evolving handbook of explicit, field-scoped behaviors. In each episode, an Editor applies relevant behaviors and receives rank-based feedback, which a Reflector converts into behavior updates. Crucially for dependable execution in real-world settings, a Moderator removes redundant proposals, a Curator commits accepted changes, and consistently harmful behaviors are deactivated. This cycle transforms episodic feedback into a persistent, field-scoped optimization memory. Experiments show that structured behaviors outperform monolithic prompt optimization, while feedback-driven evolution and moderation further improve effectiveness and reduce downside risk. The learned behaviors also remain effective on unseen product categories, suggesting that agentic memory evolution supports interpretable and transferable product optimization. [COMMENTS]Accepted to the Second Workshop for Research on Agent Language Models(REALM) at EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23709v1 [DATE]2026-10-17 00:33:17+08:00 [CATEGORIES]cs.CL
2026 Oct 16, Fri
Would you still call this Dax? Novel Visual References in VLMs and Humans [AUTHORS]Ada Defne Tür, Gaurav Kamath, Joyce Chai, Siva Reddy, Benno Krojer [ABSTRACT]Vision-language models (VLMs), like human learners, are frequently exposed to new visual concepts, but how they map novel visual references to language after exposure remains largely underexplored, particularly when those references contradict prior knowledge from pre-training. To study this, we present the Novel Visual References Dataset (NVRD): 19,176 images spanning 90 visual concepts across different levels of visual novelty, each with up to 20 increasingly perturbed versions of the original object to probe generalization. Unlike prior work on visual augmentations of familiar concepts, NVRD comprises entirely novel, open-ended stimuli constructed from scratch, mirroring how humans encounter genuinely new concepts. We evaluate 3 open- and 2 closed-source models alongside 2,400 human judgments for direct human-model comparison, and find that (i) models struggle to acquire novel concepts in-context when they contradict prior knowledge, and (ii) while models and humans show correlated sensitivity to visual perturbations, models significantly overgeneralize, extending learned labels to stimuli that humans reject. We contribute NVRD as a corpus and benchmark for research on visual concept learning in both humans and machines. [COMMENTS]EMNLP Main Conference 2026 [LINK]http://arxiv.org/abs/2606.05409v4 [DATE]2026-10-16 23:49:10+08:00 [CATEGORIES]cs.CL
Trajectory-Triggered Sparse Collaborative Decoding [AUTHORS]Amit Ranjan Trivedi, Sruthikesh Surineni, Kyong Ho Lee, Mohit Mittal, Steven Li, Rex Hill [ABSTRACT]Dense assistance methods such as speculative decoding repeatedly verify draft tokens with a target model, keeping target computation coupled to generation. We study sparse collaborative decoding, where a draft model generates most tokens and a target model is invoked only for short corrective segments. The key problem is deciding when to call the target. We show that local uncertainty is a weak trigger: entropy, negative log-likelihood, logit margin, top-$k$ mass, and velocity features give near-chance failure prediction, with AUC 0.50-0.53. In contrast, aggregate trajectory-shape features over the draft confidence path reach AUC 0.86 using a lightweight logistic predictor trained only from draft correctness labels. This induces a trigger-timing tradeoff: early calls have more leverage but weaker evidence, while later calls have stronger evidence but less ability to redirect the trajectory. Used online, learned sparse collaborative decoding improves over the best open-loop schedule by +5.3 to +20.0 accuracy points across five GSM8K draft-target pairs. On MATH-500, it gains 7.0-13.5 points when early prediction and target competence align, but loses when either is weak. On StrategyQA, it improves over open-loop by 5.0-8.3 points, showing transfer beyond arithmetics. [COMMENTS]17 pages. Accepted at the EMNLP 2026 uncertainty workshop [LINK]http://arxiv.org/abs/2610.23613v1 [DATE]2026-10-16 23:24:05+08:00 [CATEGORIES]cs.CLcs.LG
PragReST: Self-Reinforcing Counterfactual Reasoning for Pragmatic Language Understanding [AUTHORS]Jihyung Park, Minchao Huang, Leqi Liu, Elias Stengel-Eskin [ABSTRACT]Natural language understanding often depends on meanings that are implied rather than explicitly stated, requiring pragmatic reasoning. Despite strong performance on math and logical reasoning, large language models (LLMs) still struggle with making pragmatic inferences, often choosing literal interpretations. To improve LLM pragmatic reasoning, we introduce PragReST, a self-supervised framework that constructs pragmatic QA data, generates counterfactual reasoning traces, and trains models to internalize them through supervised fine-tuning and reinforcement learning, without human-labeled training data or distillation from a stronger teacher. Across four pragmatic benchmarks (PragMega, Ludwig, MetoQA, and AltPrag), PragReST improves over backbone models, task-specific pragmatic tuning baselines, and non-counterfactual variants of the same pipeline. On accuracy-based benchmarks, PragReST improves over the instruct backbone by 5.37 and 5.50% (absolute) for Qwen3-8B and Qwen3-14B, respectively. Our error analysis and ablations underscore the importance of counterfactual reasoning: PragReST primarily reduces errors caused by failures to contrast observed utterances with plausible alternatives, and removing counterfactual reasoning substantially reduces performance. Moreover, our training preserves out-of-domain performance on general-knowledge and mathematical reasoning benchmarks. [COMMENTS]EMNLP Camera-Ready. First two authors contributed equally. Code and models: https://github.com/jihyung803/PragReST [LINK]http://arxiv.org/abs/2606.18624v2 [DATE]2026-10-16 22:54:52+08:00 [CATEGORIES]cs.CL
Towards Reliable Truth-Aligned Uncertainty Estimation in Large Language Models [AUTHORS]Ponhvoan Srey, Quang Minh Nguyen, Xiaobao Wu, Anh Tuan Luu [ABSTRACT]Uncertainty estimation (UE) aims to quantify uncertainty in large language model (LLM) outputs to flag potential hallucinations and support reliable deployment. However, UE metrics often exhibit unstable performance across models and datasets, limiting their practical utility. We formalise this phenomenon as proxy failure, arising because UE metrics are derived from model behaviour rather than being explicitly grounded in factual correctness. We show that when mutual information between a UE score and correctness is low, the score's discriminability must be close to chance. To address this, we propose Truth AnChoring (TAC), a lightweight post-hoc calibration method that maps raw UE scores to truth-aligned correctness probabilities. Even with noisy and few-shot supervision, TAC continues to improve calibration, providing a practical calibration protocol. Our findings highlight the limitations of treating heuristic UE metrics as direct indicators of uncertainty about factual correctness, and position our TAC as a principled step toward more reliable uncertainty estimation for LLMs. The code repository is available at https://github.com/ponhvoan/TruthAnchor/. [COMMENTS]Accepted to EMNLP Findings 2026 [LINK]http://arxiv.org/abs/2604.00445v2 [DATE]2026-10-16 22:40:44+08:00 [CATEGORIES]cs.CL
HydraLLM: Adaptive Width Slicing for Elastic Text and Vision-Language Models [AUTHORS]Oscar Kemme, Janek Haberer, Ali Hojjat, Olaf Landsiedel [ABSTRACT]LLMs and LVLMs are increasingly deployed across heterogeneous hardware, ranging from data-center GPUs to resource-constrained edge devices. Their fixed-size architectures limit the ability to reuse a single model across deployment settings with different compute, latency, and model-size targets. We introduce HydraLLM, an elastic MQA-based architecture. Its number of active attention heads determines the effective widths of the embeddings, attention projections, feed-forward layers, and the output head, forming a nested family of subnetworks within a single parameter set. We introduce a static-to-elastic conversion pipeline that, starting from a pre-trained model checkpoint, includes importance-driven weight reordering and stochastic subnetwork training with knowledge distillation. We extend this principle to multimodal inference through an elastic projection module whose output width follows the active text width. HydraLLM covers a wider range of active parameter configurations than FFN-only elastic architectures. With 75% of heads active, HydraLLM achieves 2.2 percentage points higher TriviaQA accuracy than a state-of-the-art nested baseline configuration which has 17.25% more parameters. With 50% of heads active, our multimodal model achieves a CLIPScore comparable to that of the 7$\times$ larger Gemma 3 4B model. With only 25% of heads active, HydraLLM increases token-generation throughput by 75.3% over the full-width configuration. [COMMENTS]Accepted at the BabyLM Workshop at EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23455v1 [DATE]2026-10-16 21:38:11+08:00 [CATEGORIES]cs.CLcs.LG
MAS-Bai: Linguistically Informed Adaptive Multi-Agent Translation for Bowen-Script Bai [AUTHORS]Jingting Zheng, Hong Ren, Yuehan Cui, Shaolin Zhu, Deyi Xiong [ABSTRACT]Large Language Models (LLMs) remain unreliable for extremely low-resource languages whose writing systems conflict with high-resource pretraining priors. We study Bai-Chinese translation in the Bowen script, where Chinese characters are frequently borrowed phonetically rather than semantically, causing severe literal-translation hallucinations. We propose MAS-Bai, a linguistically informed adaptive multi-agent framework for Bowen-script Bai-Chinese translation. MAS-Bai routes each input through either a lightweight retrieval-translation path or a deeper philological path with etymological and syntacticagents, and uses cycle-consistency with a lightweight reverse translator as an inference-time reliability signal. We construct a 6,536-sentence Bai-Chinese corpus and evaluate on a 1,536-sentence test set. MAS-Bai achieves 16.23 BLEU and 62.15 COMET, improving over Standard RAG by 8.11 BLEU and 6.85 COMET. Forward LoRA, expert ablations and controlled routing experiments distinguish the roles of supervised adaptation, Bai-specific expertise and conditional routing. The largest gains occur on Bowen phonetic loans and divergent Bai word order, while human evaluation and reverse-verifier analyses support the automatic results. [COMMENTS]14 pages, 3 figures, Findings of EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23205v1 [DATE]2026-10-16 17:43:32+08:00 [CATEGORIES]cs.CL
Cross-Centaur: Modeling Individual Decision-Making Across Psychological Tasks with Language Models [AUTHORS]Hikaru Asano, Hidezo Suganuma, Ryutaro Mori, Yukino Baba [ABSTRACT]Large language models (LLMs) are increasingly used as computational models of human behavior. Recent models such as Centaur have advanced this goal by fine-tuning LLMs on trial-by-trial decisions from psychological experiments. However, existing large-scale behavioral datasets do not link the same participants across different tasks, so these models learn population-level regularities but are not trained to model individual-level regularities across tasks. We introduce Cross-Psych-101, a dataset of 591 participants and 261\{,\}551 valid decisions across 12 Bandit, Tradeoff, Structure, and Social tasks. We train Cross-Centaur to predict a participant's target-task actions from a target-task description and the source-task description paired with the same participant's source-task behavior. Cross-Centaur outperforms off-the-shelf LLMs and Centaur on held-out participants, both within and across tasks. Probing analyses suggest that its hidden states encode participant-level behavioral tendencies that transfer across tasks. These results highlight the potential of cross-task individual-level data for predicting human decision-making with language models. Cross-Pysch-101 is available at https://huggingface.co/datasets/HikaruAsano/cross-psych-101. [COMMENTS]Accepted to EMNLP 2026 (Main Conference, Oral Presentation) [LINK]http://arxiv.org/abs/2610.23157v1 [DATE]2026-10-16 16:55:00+08:00 [CATEGORIES]cs.CL
2026 Oct 17, Sat
Behavioral Couplings: From Formation to Behavioral Emergence in Language Models [AUTHORS]Dongyoon Hahm, Woogyeol Jin, Dennis Wei, Yi Zhou, Nathalie Baracaldo, Kimin Lee [ABSTRACT]Fine-tuning large language models (LLMs) can induce behaviors beyond those explicitly targeted. For example, training a model to respond more warmly can make it more sycophantic, and training it to write insecure code can cause broad misalignment on prompts unrelated to coding. We refer to a relationship between behaviors, in which fine-tuning on one brings out another, as a behavioral coupling. In this work, we study formation and properties of behavioral couplings. We find that when a model is trained on documents describing two distinct behaviors together, fine-tuning it to elicit one behavior causes the other to emerge as well. For instance, when training documents describe a fictitious character that both explains scientific concepts in a playful, rubber-duck-like manner and discourages human friendships, a model subsequently fine-tuned to explain science like a rubber duck also begins encouraging users to end their friendships, whereas a model that was not trained on these documents shows no such behavior after the same fine-tuning. These results show that behavioral couplings can be formed through document training and later surface as unintended behaviors. More concerningly, we find that couplings can compose. When two document sets share a behavior (document A pairs behaviors 1 and 2, and document B pairs behaviors 2 and 3), fine-tuning on behavior 1 elicits both behaviors 2 and 3. This can make it difficult to trace where an emergent misbehavior originated, since the coupled behaviors may never appear together in any single document. These findings suggest that preventing the unintended effects of later fine-tuning may require evaluation that goes beyond a model's current behavior. [LINK]http://arxiv.org/abs/2610.23847v1 [DATE]2026-10-17 01:58:08+08:00 [CATEGORIES]cs.CLcs.LG
AI-Driven Reproducibility and Analytical Robustness in Scientific Discovery [AUTHORS]Shicheng Liu, Krista Opsahl-Ong, Andrew Drozdov, Julian Killingback, Arnav Singhvi, Ronak Pradeep, Michael Bendersky, James C. Dickerson, Monica S. Lam [ABSTRACT]Reproducible and analytically robust research findings are essential to scientific progress, but evaluating these properties typically requires costly manual effort. We investigate whether LLMagents can help automate reproduction and robustness analysis, using a medical case study grounded in the Surveillance, Epidemiology, and End Results (SEER) database, a widely used U.S. cancer registry underlying tens of thousands of published studies. First, we introduce SEER-Repro, a benchmark of 20 published SEER-based papers together with the data and resources required to reproduce their findings. Second, we create an AI agent that successfully reproduces 77.4% of 2,190 atomic claims, while providing executable baselines for further analysis. Third, we develop an agentic analysis pipeline to test the robustness of the medical findings by identifying analytical choices underlying each finding, proposing defensible alternatives, and rerunning the analysis to determine whether those choices affect the conclusion. Across published papers in SEER-Repro, our pipeline identifies potentially consequential analytical choices in 9.5% of fully reproduced findings and flags them for expert review. These results show how LLMagents can support scientific review by reproducing findings, systematically testing their robustness, and directing expert attention toward potentially consequential analytical choices. [LINK]http://arxiv.org/abs/2610.23845v1 [DATE]2026-10-17 01:57:55+08:00 [CATEGORIES]cs.CL
Optimizing Singular Spectrum for Large Language Model Compression [AUTHORS]Dengjie Li, Tiancheng Shen, Yao Zhou, Bernard Ghanem, Yibo Yang, Ming-Hsuan Yang [ABSTRACT]Large language models (LLMs) have demonstrated remarkable capabilities, yet prohibitive parameter complexity often hinders their deployment. Existing singular value decomposition (SVD) based compression methods simply deem singular values as importance scores of decomposed components. However, this importance ordered by singular values does not necessarily correlate with the performance of a downstream task. In this work, we introduce SoCo (Singular spectrum optimization for large language model Compression), a novel compression framework that learns to rescale the decomposed components of SVD in a data-driven manner. Concretely, we employ a learnable diagonal matrix to assign importance scores for singular spectrum and develop a three-stage training process that progressively refines these scores from initial coarse compression to fine-grained sparsification-thereby striking an effective balance between aggressive model compression and performance preservation. Thanks to the learnable singular spectrum, SoCo adaptively prunes components according to the sparsified importance scores, rather than relying on the fixed order of singular values. More importantly, the remaining components with amplified importance scores can compensate for the loss of the pruned ones. Experimental evaluations across multiple LLMs and benchmarks demonstrate that SoCo surpasses the state-of-the-art methods in model compression. [LINK]http://arxiv.org/abs/2502.15092v2 [DATE]2026-10-17 01:56:38+08:00 [CATEGORIES]cs.CL
DiscoveryRSI: Recursively Learning to Discover Valuable Hypotheses [AUTHORS]Jianyuan Zhong, Chenwei Wu, Zhaochen Yu, Xixian Jiang, Xuewei Li, Hao Xiao, Xilin Dai, Junyang Wang, Jiaru Zou, Zhiqing Cui, Boyuan Zheng, Jing Xiong, Zhen Zhang, Qiang Gao, Michael Morley, Luis Nakayama, Qiang Xu, Liyue Shen, Tao Gui, Leo Anthony Celi, James Zou, Zhenfei Yin, Yingcheng Wu, Ling Yang [ABSTRACT]Scientific discovery requires developing worthwhile hypotheses into practical investigations, yet learning from this evolving research process remains challenging. We introduce an operational definition of valuable and testable research hypotheses: evidence-grounded claims about consequential uncertainties, coupled with feasible investigations, explicit decision criteria, and alternative explanations. Our DiscoveryRSI framework develops hypothesis trees through multi-agent evolutionary search, compiles selected investigations into executable Harbor tasks, and recursively learns through three mechanisms: hypothesis credit optimization attributes scientific assessments to documented research contributions; harness replay identifies productive strategies and recurring failures to refine research workflows; and rediscovery applies accepted improvements to new investigations, generating further experience for subsequent learning. We also introduce DFM-Bench, which provides comprehensive assessment of hypothesis quality and cumulative research progress at the levels of hypotheses, revisions, development paths and research runs. Across four life-science domains, automated evaluation shows that hypothesis quality rises from 4.86 for the base model to 5.39 after two rounds of self-improvement, with the large gains in falsifiability and data resolvability. We further validate hypothesis quality through expert review, comparing our DiscoveryRSI-27B with other research agents. Preliminary human assessments show that DiscoveryRSI-27B scores higher than GPT-6 Luna on hypothesis importance and whether claims are plausible and evidence-bounded. DiscoveryRSI-27B achieved a 6.3% rate of hypothesis correspondence with research published after the model cutoff, largely outperform GPT-6 with 2.5% and Claude Opus 5 with 2.3%, while delivering greater human-rated improvement along research paths. [COMMENTS]Code: https://github.com/Gen-Verse/DiscoveryRSI [LINK]http://arxiv.org/abs/2610.23839v1 [DATE]2026-10-17 01:56:35+08:00 [CATEGORIES]cs.CL
Beyond the Answer: Student-Conditioned Trace Construction for Reasoning Distillation [AUTHORS]Shaoang Li, Yanhang Shi, Yufei Li, Mingfu Liang, Frank Shyu, Sandeep Pandey, Luke Simon, Xi Liu, Jian Li [ABSTRACT]A reasoning trace can accompany a correct answer without providing effective supervision for a particular student. We study how to construct traces that improve learning beyond answer-only supervision by considering both their intermediate content and their accessibility to the target learner. We introduce ReDiST, a student-conditioned trace-construction and distillation framework for evidence-rich reasoning. Starting from the student's own attempts, ReDiST revises answer-correct drafts against the supplied evidence and repairs unsuccessful drafts with additional guidance from the verified answer. A frozen student-matched reader screens and ranks original and edited candidates using answer recovery and verified-answer probability, given the question, evidence, and candidate trace. We distill the selected targets while balancing trace and final-answer supervision with an auxiliary answer loss. Construction and selection occur offline, and inference uses only the distilled student. Experiments on question-answering benchmarks show that ReDiST can improve both source-task and cross-benchmark performance over answer-only, original-trace, and the evaluated trace-selection and rewriting baselines. Across supervision recipes, performance rankings can differ between source and transfer benchmarks, highlighting the importance of evaluating trace teaching utility across distributions. [LINK]http://arxiv.org/abs/2610.23831v1 [DATE]2026-10-17 01:52:58+08:00 [CATEGORIES]cs.CLcs.LG
Automatic Tensor-level Execution of Abstract Interpreters for Neural Networks [AUTHORS]Avaljot Singh, Yasmin Chandini Sarita, Aditya Mishra, Gagandeep Singh, Charith Mendis [ABSTRACT]Developing new abstract interpretation-based DNN certifiers is difficult because their mathematical design is expressed at the neuron level, while implementations must operate at the tensor level. Existing libraries such as auto_LiRPA, IVAN, and ERAN support fixed sets of certifiers; extending them requires expertise in formal methods and tensor programming. Domain-specific languages such as ConstraintFlow let researchers specify certifiers at the neuron level, but cannot run them on real DNNs.
We present Neuton, a framework that automatically translates neuron-level certifier specifications into tensor-level executables, enabling new DNN certifiers without manual tensor programming. Neuton makes three contributions. First, a stack-based intermediate representation (IR) and shape analysis infer implicit tensor operations, including broadcasts, dimension expansions, and reductions, required for correct execution. Second, domain-specific rewrite rules optimize common tensor operation sequences into more efficient equivalents. Third, gBCSR is a sparse tensor format using variable-sized, layer-aligned blocks over a non-zero default value, tailored to DNN certifier sparsity.
We prove that translation preserves neuron-level semantics. We evaluate Neuton on 8 certifiers and 32 DNN benchmarks. Neuton compiles four existing certifiers and produces the first executable implementations of four others: two previously specified but never run, and two new designs. These expose precision-runtime trade-offs unavailable in existing libraries. For example, SkipPoly matches DeepPoly's precision on all 32 DNNs while running as fast or faster on half. Neuton's runtime on existing certifiers is competitive with hand-optimized libraries. Code: https://github.com/uiuc-focal-lab/constraintflow.git [LINK]http://arxiv.org/abs/2507.20055v2 [DATE]2026-10-17 01:43:47+08:00 [CATEGORIES]cs.CL
Steering beyond Language: Improving Multilingual Training-Free Text Embeddings via Paired-Residual Subspace Removal [AUTHORS]Ailiang Lin, Zhuoyun Li, Yusong Wang, Keyu Mao, Kotaro Funakoshi, Manabu Okumura [ABSTRACT]While deriving text embeddings directly from frozen LLMs has emerged as a promising direction, existing methods focus almost exclusively on English, leaving multilingual settings largely unexplored. In this work, we present the first systematic study of multilingual training-free text embeddings in LLMs. Our analysis reveals a language-conditioned embedding geometry in which semantically equivalent sentences remain misaligned across languages. We hypothesize that language-specific variation can be disentangled from semantic content to improve embedding quality. To this end, we propose Paired-Residual Subspace Removal (PRSR), a training-free inference-time intervention that steers LLM representations toward a more language-invariant semantic space. PRSR first constructs paired residuals from multilingual parallel sentences to capture language-specific deviations. It then estimates layer-wise low-rank subspaces from these residuals and suppresses the corresponding components in the last-token hidden states. We conduct more than 200 monolingual evaluations across 18 languages using MTEB and MMTEB, along with cross-lingual bitext mining between English and over 200 other languages. Empirical results demonstrate that PRSR consistently improves the performance of existing training-free baselines in both monolingual and cross-lingual settings across different LLMs. [LINK]http://arxiv.org/abs/2610.23807v1 [DATE]2026-10-17 01:30:43+08:00 [CATEGORIES]cs.CL
When Reasoning Traces Conflict with Evidence: How Answer Configuration Shapes Evidence Following [AUTHORS]Shaoang Li, Yanhang Shi, Jian Li [ABSTRACT]A reasoning trace retained at answer time plays two roles: it can raise accuracy, and it can compete with the current evidence when the two disagree. We disentangle these roles by crossing verified evidence versions with the traces generated from each and answering from the same traces and evidence under two configurations: continuation of the model's own thinking and a separate answer call. Across three model families (gpt-oss, Gemma4, Qwen3) on HotpotQA and MuSiQue, the separate call raises evidence-following over continuation by 29.9 to 60.6 points. On the same pairs, removing the trace lowers accuracy by similar amounts under both configurations. Yet continuation repeats the stale conclusion in most conflicts, and the separate call follows the current evidence far more often. Benefiting from a trace and deferring to it are therefore distinct properties. Further analyses show that the final line of a trace carries most of its value, gold evidence helps most before the trace is written, distractors weaken correction under a separate call, and renewed reasoning, unlike a verification instruction, follows the current evidence in most conflicts. Because trace deletion alone cannot reveal which source an answer follows, evaluations should report both properties together with the answer configuration. [COMMENTS]25 pages, 4 figures [LINK]http://arxiv.org/abs/2610.20805v2 [DATE]2026-10-17 01:30:03+08:00 [CATEGORIES]cs.CL
VocalGrad: Evaluating Acoustic Perception in Audio Language Models [AUTHORS]Koki Ryu, Hitomi Yanaka [ABSTRACT]Recent large audio language models (LALMs) have extended their capabilities beyond lexical speech understanding to tasks requiring paralinguistic and non-linguistic audio understanding, such as emotion recognition. However, these models still struggle to reliably utilize acoustic information, particularly when it co-occurs with spoken linguistic content. Existing benchmarks do not clearly distinguish acoustic perception from higher-level reasoning, making it difficult to identify the source of these failures. To address this issue, we introduce VocalGrad, a benchmark designed to evaluate LALMs' ability to perceive low-level acoustic variations. VocalGrad formulates this problem as a simple binary classification of temporal change direction, enabling evaluation largely independent of higher-level reasoning or semantic understanding. Our benchmark results reveal a substantial gap in acoustic perception between human annotators and state-of-the-art LALMs: while human annotators achieve high accuracy, LALMs perform at near-chance levels. We further conduct linear probing to identify structural bottlenecks in these models. Our analysis reveals two distinct failure modes: for some categories, task-relevant information is preserved in internal representations but not utilized by the language model head; for others, the information becomes progressively less linearly decodable across model layers. Overall, our results reveal substantial limitations of LALMs in modeling acoustic variation under controlled settings and expose representation differences not observable from text outputs. [COMMENTS]NeurIPS 2026 (Evaluations & Datasets Track). Dataset page: https://huggingface.co/datasets/ynklab/vocalgrad [LINK]http://arxiv.org/abs/2610.23745v1 [DATE]2026-10-17 00:54:35+08:00 [CATEGORIES]cs.CL
Beyond Consensus: Perspectivist Modeling and Evaluation of Annotator Disagreement in NLP [AUTHORS]Yinuo Xu, David Jurgens [ABSTRACT]Annotator disagreement is widespread in NLP, particularly for subjective and ambiguous tasks such as toxicity detection and stance analysis. While early approaches treated disagreement as noise to be removed, recent work increasingly models it as a meaningful signal reflecting variation in interpretation and perspective. This survey provides a unified view of disagreement-aware NLP methods. We first present a domain-agnostic taxonomy of the sources of disagreement spanning data, task, and annotator factors. We then synthesize modeling approaches using a common framework defined by prediction targets and pooling structure, highlighting a shift from consensus learning toward explicitly modeling disagreement, and toward capturing structured relationships among annotators. We review evaluation metrics for both predictive performance and annotator behavior, and noting that most fairness evaluations remain descriptive rather than normative. We conclude by identifying open challenges and future directions, including integrating multiple sources of variation, developing disagreement-aware interpretability frameworks, and grappling with the practical tradeoffs of perspectivist modeling. [COMMENTS]Accepted to EMNLP 2026 [LINK]http://arxiv.org/abs/2601.09065v3 [DATE]2026-10-17 00:39:26+08:00 [CATEGORIES]cs.CL
IndicDetect: Evaluating Cross-Lingual LLM-Generated Text Detection for Hindi, Telugu, and Tamil [AUTHORS]Bhaskar Ganesh Devalla, Junchao Wu, Nilesh Dokuparthi, Greeshma Yaluru, Tatiana Muniz Rodriguez, Lidia S. Chao, Derek F. Wong [ABSTRACT]The rapid proliferation of LLMs has further heightened the need to develop dependable AI-generated text detection, especially beyond English. Nevertheless, current benchmarks pay little attention to Indic languages and test detectors in idealized settings that do not represent the real world. We present a generalized benchmark for AI-generated text detection in Hindi, Telugu, and Tamil, which we call IndicDetect, designed to assess the robustness of detectors under realistic distribution shifts. IndicDetect comprises highly curated human-written texts matched with LLM-generated counterparts across various domains and generators, and systematically evaluates detectors in the presence of domain shift, generator shift, and adversarial perturbation. Using a single and repeatable evaluation scheme, we evaluate a wide range of statistical and neural detectors. We find substantial robustness failures: supervised neural detectors perform well in-distribution, while training-free methods degrade considerably under unseen generators and adversarial attacks. The severity of these failures varies across languages, with Hindi exhibiting the largest overall degradation under adversarial perturbations. These results highlight that the primary weakness of existing detectors in Indic settings lies in their robustness, not in their peak accuracy. IndicDetect provides standard data splits, an evaluation protocol, and baselines to establish a robust, language-aware foundation for AI-generated text detection in Indic scripts. [LINK]http://arxiv.org/abs/2608.29919v2 [DATE]2026-10-17 00:26:48+08:00 [CATEGORIES]cs.CL
Paved with True Intents: Intent-Aware Training Improves LLM Safety Classification Across Training Regimes [AUTHORS]Jeremias Ferrao, Niclas Müller-Hof, Iustin Sîrbu, Traian Rebedea, Yftah Ziser [ABSTRACT]We argue that safety classifiers should model user intent as an explicit signal between the prompt and the final label. To study this, we introduce AIMS, a human-annotated dataset of 1,724 difficult safety prompts, each paired with an intent description and harm label. We use AIMS to evaluate intent-aware training across supervised fine-tuning, preference learning, reasoning distillation, and reinforcement learning. Despite its size, AIMS enables competitive safety classifiers across training regimes: DPO from model-generated intent errors improves over SFT, and intent-conditioned distillation outperforms reasoning-only distillation in most teacher-student pairs. Most notably, directly rewarding intent faithfulness with GRPO yields the strongest average performance across five external safety benchmarks, while our intent-aware models form the inference latency-F1 Pareto frontier. These results show that faithful intent modeling is a compact, high-quality supervision signal for more robust safety classifiers. [LINK]http://arxiv.org/abs/2606.27210v2 [DATE]2026-10-17 00:09:43+08:00 [CATEGORIES]cs.CL
Beyond the Benchmark: Conditioning, Domains and Failure Modes of Qwen3-TTS [AUTHORS]Séverin Baroudi, Yanis Labrak, Petr Motlicek, Hervé Bredin, Ricard Marxer [ABSTRACT]Qwen3-TTS reports state-of-the-art zero-shot voice cloning, but its technical report leaves open how the model should be conditioned, how it behaves outside read speech, and how it fails. We characterise its two cloning interfaces, in-context learning (ICL) and X-Vector, against Chatterbox, XTTS-v2 and F5-TTS on read, meeting, emotional and air-traffic-control (ATC) speech, using eighteen objective measures, an LLM judge, a spoofing countermeasure and a listening test. Neither interface dominates: ICL clones timbre more closely from correctly transcribed references, X-Vector clones timing, and ICL degrades sharply when a short reference transcript is wrong. Domain explains four times more variance than system choice. Qwen3-TTS leads on intelligibility and pitch register but trails on prosody, and the objective measures rank systems differently from listeners. An off-the-shelf countermeasure flags its clones no more often than genuine recordings, and on ATC radio it fails for every system. Code and audio are released. [LINK]http://arxiv.org/abs/2610.23678v1 [DATE]2026-10-17 00:09:20+08:00 [CATEGORIES]cs.CL
Learning to Reason About Implicit Multimodal Harm: From Diagnosis to Generalization [AUTHORS]Anisha Saha, Varsha Suresh, Teodora Kamova, Sophia Wiedmann, Timothy Hospedales, Vera Demberg [ABSTRACT]A benign image paired with a benign caption can become harmful only in combination, making detection fundamentally a cross-modal reasoning problem. Existing multimodal harm benchmarks evaluate only classification accuracy, leaving unclear whether models identify the cross-modal mechanism that produces harm. We introduce MuPHI (Multimodal Pragmatic Harm Interpretation), a controlled diagnostic benchmark in which harm arises purely from image-text composition, with benign counterfactuals and annotated harm rationales. Using MuPHI, we systematically evaluate where compositional harm reasoning breaks in vision-language models (VLMs). MuPHI reveals that VLMs reliably ground each modality yet fail to bind the evidence into the harm mechanism and their predictions respond weakly to label-determining counterfactual interventions. Moreover, existing harm detection approaches trained on labels substantially improves in-domain (ID) detection but transfers poorly across harm distributions. We therefore propose Compositional Harm Interpretation Learning (CHIL), which rewards evidence identification in each modality, cross-modal binding and verdict consistency in the reasoning chain rather than just predicted label. CHIL improves 81% OOD transfer evaluations by an average of +2.5 macro-F1 and correct counterfactual flip rate by up to +3.8 and +5.4 % points, while also improving overall reasoning quality over label-only optimization. These results suggest that optimizing how a model reasons, not just what it concludes, offers a promising direction for multimodal safety systems that generalize beyond dataset-specific correlations. [LINK]http://arxiv.org/abs/2605.29951v2 [DATE]2026-10-17 00:01:16+08:00 [CATEGORIES]cs.CLcs.LG
Coordination Matters: Process-Aware Evaluation of Cooperative Multi-Agent Reinforcement Learning [AUTHORS]Maria Ana Cardei, Matthew Landers, Afsaneh Doryab [ABSTRACT]Cooperative multi-agent reinforcement learning (MARL) benchmarks commonly emphasize aggregate outcomes such as return, success rate, or completion time. While essential, these metrics can obscure how agents coordinate and fail to distinguish policies that achieve similar outcomes through different interaction processes. We propose a process-aware evaluation framework that complements aggregate performance with diagnostics of interference, coordinated task completion, participation, and coordination efficiency. We evaluate the framework in STAT, a controlled spatial task-allocation testbed, and examine its broader applicability in two established cooperative MARL benchmarks. Across diverse value-based and actor-critic methods, we find policies with nearly identical return can exhibit substantial differences in coordination behavior. Controlled scaling in STAT reveals distinct process signatures for spatial, task, and agent scaling, while a reward-shaping ablation shows that explicitly penalizing assignment conflicts can alter coordination behavior with little change in task return. Results in the established benchmarks further show that the framework remains informative under different coordination mechanisms. These findings support process-aware evaluation as a complement to outcome-based benchmarking. The framework provides a common structure for selecting and interpreting environment-specific diagnostics that characterize not only what cooperative policies achieve, but how they coordinate to achieve it. [COMMENTS]Under review [LINK]http://arxiv.org/abs/2605.06557v2 [DATE]2026-10-17 01:58:51+08:00 [CATEGORIES]cs.LG
Task-aligned spectral inductive biases in parameterized quantum circuits [AUTHORS]Felix Frohnert, Adrián Pérez-Salinas, Sebastian Lange, Marcus Schwaebisch, Evert van Nieuwenburg, Vedran Dunjko [ABSTRACT]Parameterized quantum circuits are widely investigated in quantum machine learning, yet how their architecture can provide task-aligned inductive biases remains poorly understood. We study data reuploading quantum circuits with two distinct inputs: a signal that varies between samples and a fixed, task-defined spectral query that specifies how the signal should be probed. We show that encoding the query can change the circuit's hypothesis class over the signal by modulating how its existing signal frequencies are combined, without introducing new sample-dependent information. This mechanism matches tasks built on nonuniform Fourier transforms, whose evaluation frequencies specify the query and whose sampling positions guide its encoding. We examine this connection in a classification task inspired by synthetic aperture radar, where acquisition geometry determines a query for classifying unfocused spectral patches before image formation. Across small-scale spectral and radar-like tasks, aligned queries improve learning from limited data over parameter-matched circuits with mismatched or absent queries. These results provide a route from known spectral task structure to the design and validation of inductive biases in parameterized quantum circuits. [LINK]http://arxiv.org/abs/2610.23834v1 [DATE]2026-10-17 01:53:19+08:00 [CATEGORIES]cs.LG
ENFR: Evidence Networks with Flow-Matched References for Amortized Simulation-Based Model Inference [AUTHORS]Kai Lehman, Jameson Dong, Carolina Cuesta-Lazaro, Niall Jeffrey, Benjamin D. Wandelt, Francisco Villaescusa-Navarro, Sven Krippendorf, Jochen Weller [ABSTRACT]The evidence, $\mathcal\{Z\} = p(x | \mathcal\{M\})$, of a data-generating model $\mathcal\{M\}$ is a powerful tool in Bayesian frameworks: it enables model inference, model misspecification tests, and goodness-of-fit assessment as an alternative to standard discrepancy statistics, e.g. $χ^2$. While versatile, the evidence is notoriously difficult to compute, even when the likelihood is available. In both likelihood-based and simulation-based inference, estimating $\mathcal\{Z\}$ provides an immediate solution to high-dimensional goodness-of-fit or model misspecification tests: using $\mathcal\{Z\}(x)$ as the test statistic induces a minimum-volume level-$α$ test. Such tests are essential to assess model fidelity, which is necessary to trust simulation-based parameter inference. Existing learned approaches largely target Bayes factors, which only measure evidence ratios. Instead, we provide accurate, fast, and amortized estimates of the absolute evidence. We present ENFR: Evidence Networks with Flow-matched References. Building on the Evidence Network framework of Jeffrey & Wandelt (2024), the proposed method first learns Bayes factors with respect to a flow-matched reference, from which the absolute evidence is recovered. We show improvement in estimation of $\mathcal\{Z\}$ over the flow-based baseline in all applications considered and the best performance of all density estimation methods considered in our benchmarks. We furthermore introduce and provide the proof for an architecture that jointly learns the individual evidences of a set of models $\\{\mathcal\{M\}_i\\}$. We demonstrate the performance and scaling behavior of these approaches on data from a variety of scientific applications. [COMMENTS]28 pages, 6 figures [LINK]http://arxiv.org/abs/2610.23825v1 [DATE]2026-10-17 01:48:26+08:00 [CATEGORIES]cs.LG
Learning without relearning physical symmetry: an SO(3)-equivariant convolutional network (CNN) on the sphere [AUTHORS]Joshua Williamson, Niall Jeffrey, Lorne Whiteway [ABSTRACT]Many of the most important datasets in the physical sciences live natively on the sphere, and these data often carry no preferred position or orientation. Using methods that do not reflect this symmetry leads to far higher simulation requirements, which can be catastrophic for scientific analyses (forcing reliance on cheap, imperfect models that bias results). We present a novel spherical implementation of a gauge equivariant mesh architecture that is specifically designed for the Hierarchical Equal Area isoLatitude Pixelization (\healpix\{\}) of the sphere. The $\mathrm\{SO\}(3)$-equivariant convolution is built from steerable filters, parallel transport and spin-weighted features. The architecture supplies the exact symmetry, so the network does not spend training data learning it. The convolution operations are constructed to give provably exact equivariance, given a band-limit. Our full CNNs can achieve SO(3)-invariance error at per-cent-level and lower, which is controlled through well-understood hyperparameters. On benchmarks, our technique gives the most accurate dense gradient regression and competitive cosmic microwave background foreground removal and denoising. The improved simulation efficiency matters most in simulation-based inference (SBI), where high-fidelity simulations are expensive and neural compressors require large training sets, so a fixed budget forces analyses with less accurate simulators. In cosmology lensing SBI, our technique yields the tightest posteriors at matched simulation count. This approach is a novel method that will enable new convolutional neural network analyses, including generative and inference tasks, for processes and data on the sphere. [LINK]http://arxiv.org/abs/2610.23821v1 [DATE]2026-10-17 01:45:43+08:00 [CATEGORIES]cs.LG
Explainable Machine Learning Analysis of Temperature Effects on Mechanical Strength of Fiber-Reinforced High-Performance Concrete [AUTHORS]Jagaran Chakma, Zhiguang Zhou, Badhan Chakma, Priyanta Chakma [ABSTRACT]Predicting the compressive strength of fiber-reinforced high-performance concrete subjected to temperature exposure is challenging because to the intricate interdependencies of mixture composition, fiber properties, curing conditions, and temperature. A data-driven framework to predict compressive strength utilizing 1,204 experimental records containing a diverse array of material designs and exposure situations. This study evaluates eight machine-learning models and demonstrate that ensemble approaches routinely surpass individual learners. Both gradient boosting and extreme gradient boosting exhibit superior performance on unseen data, each attaining a R^2 of 0.903. Gradient boosting produces the minimal root mean square error (6.67 MPa), whereas extreme gradient boosting achieves the lowest mean absolute error (3.97 MPa) and mean absolute percentage error (10.60%). The models are further interpreted utilizing Shapley additive explanations and partial dependence analysis, which reveal temperature as the primary factor influencing compressive strength, succeeded by fine aggregate, superplasticizer, cement content, water-to-binder ratio, and water content. Strength diminishes significantly with rising temperature, albeit the extent of deterioration is contingent upon the specific combination design. The best performing models are subsequently implemented in a Python-based graphical user interface for expedited prediction. Our findings offer a precise and comprehensible methodology for evaluating temperature-dependent compressive strength and facilitate the practical, data-driven design of fiber-reinforced high-performance concrete. [COMMENTS]29 pages [LINK]http://arxiv.org/abs/2610.23811v1 [DATE]2026-10-17 01:34:24+08:00 [CATEGORIES]cs.LG
LeanPRM: Lean-Grounded Process Supervision for Natural Language Proof Verification [AUTHORS]Rodney Lafuente-Mercado, Salman Rahman, Jingyan Shen, James Shiffer, Saadia Gabriel, Pavel Izmailov [ABSTRACT]AI systems now write long and difficult mathematical proofs, and checking their step-by-step validity is becoming a bottleneck. A learned verifier could check proofs cheaply, but training one requires a correctness label on every step, and such labels are hard to obtain for open-ended proofs. Existing methods take them from human annotators, which is costly, or derive them from final answers, which proofs do not have, and formalizing each proof in Lean requires expert effort. We introduce LeanPRM, a framework that uses the Lean proof assistant to create step-level labels for natural-language proofs. A model writes a natural-language proof together with a paired formal proof, without feedback from Lean. Lean then labels every step, and a second model discards failures caused only by the formal syntax, so that the negative labels mark errors in the mathematics. LeanPRM contains 10,634 proofs with 152K labeled steps and no human annotation. A Qwen3-4B verifier trained on LeanPRM raises exact step-label accuracy on a held-out set of 200 proofs (LeanPRM-Bench) from 48.2 to 74.7, as opposed to 60.3 when the same base model is trained on PRM800K. It also selects better proofs in best-of-N selection on IMO-level problems, and outperforms the PRM800K-trained verifier on three of four external verification benchmarks. Training on LeanPRM improves verifiers at every size from 0.6B to 8B parameters on LeanPRM-Bench, Hard2Verify and IMO-GradingBench. Given the labels need no human annotation, LeanPRM can grow with the supply of formal statements, offering a path to step-level feedback on AI-generated proofs that are hard for humans to check. [COMMENTS]29 pages, 9 figures. Project page: https://rlafuente.com/lean-prm [LINK]http://arxiv.org/abs/2610.17697v2 [DATE]2026-10-17 01:31:06+08:00 [CATEGORIES]cs.LG
RCPN: towards robust pseudo-labels in unsupervised CLIP adaptation via consistent prototypes and neighborhoods [AUTHORS]Eman Ali, Chetan Arora, Muhammad Haris Khan [ABSTRACT]In unsupervised adaptation for vision-language models such as CLIP, pseudo-labels derived from zero-shot predictions often exhibit significant noise, particularly under domain shifts or in visually complex scenarios. Conventional pseudo-label filtering approaches, which rely on fixed confidence thresholds, tend to be unreliable in fully unsupervised settings. In this work, we propose a novel adaptive pseudo-labeling framework that enhances CLIP's adaptation performance by integrating prototype consistency and neighborhood-based consistency. The proposed method comprises two key components: PICS, which assesses pseudo-label accuracy based on in-class feature compactness and cross-class feature separation; and NALR, which exploits semantic similarities among neighboring samples to refine pseudo-labels dynamically. Additionally, we introduce an adaptive weighting mechanism that adjusts the influence of pseudo-labeled samples during training according to their estimated correctness. Extensive experiments on 11 benchmark datasets demonstrate that our method achieves state-of-the-art performance in unsupervised adaptation scenarios, delivering more accurate pseudo-labels while maintaining computational efficiency. [LINK]http://arxiv.org/abs/2507.22075v2 [DATE]2026-10-17 01:28:20+08:00 [CATEGORIES]cs.LG
LagNet: Lag-Structured Operators for Long-Term Time Series Forecasting [AUTHORS]Zhenan Yu, Guangxin Jiang, Jiuyun Jiang, Jin Yang [ABSTRACT]Long-term time series forecasting maps a finite history window to an extended future horizon. Many existing models focus primarily on the design of input representations and model structures, while paying less attention to the structural form of the history-to-future forecasting operator itself. We examine this operator directly and find that, across model families, sensitivity matrices exhibit recurring diagonal patterns whose diagonals correspond to relative temporal lags. Motivated by this observation, we develop a framework for direct multi-step forecasting based on lag-structured temporal operators. Its basic building block, LagLayer, constructs a Toeplitz matrix by tying history-to-future connections with the same relative lag and incorporates horizon-wise and history-wise scaling coefficients to relax exact sharing. This design enables LagLayer to serve as an effective replacement for temporal linear layers in existing forecasting models. Stacking LagLayer modules with activations and multiple feature maps yields LagNet. Experiments on long-term time series forecasting benchmarks show that our method is competitive with strong baselines, while sensitivity matrix diagnostics and limited-data experiments demonstrate that LagLayer provides an effective and sample-efficient alternative to a linear layer. [LINK]http://arxiv.org/abs/2610.23735v1 [DATE]2026-10-17 00:44:53+08:00 [CATEGORIES]cs.LG
Diffusion-Based Stochastic Operator Networks for Uncertainty Quantification in Stochastic Partial Differential Equations [AUTHORS]Phuoc-Toan Huynh, Richard Archibald, Feng Bao [ABSTRACT]We introduce a novel framework for uncertainty quantification of solution operators associated with stochastic partial differential equations (SPDEs). Although SPDEs play a central role in modeling complex physical systems under uncertainty, their practical use typically requires specifying the magnitude and structure of model uncertainties that are often unknown and difficult to infer from noisy measurements. To address this challenge, we develop a stochastic operator-learning framework that learns directly from noisy data and outputs both a mean solution field and a quantification of uncertainty. The proposed method, namely the Stochastic Operator Network (SON), is constructed by combining the structure of the Deep Operator Network (DeepONet) with Stochastic Neural Networks (SNNs) to model stochasticity and enable probabilistic prediction. The training procedure is carried out by minimizing a Hamiltonian-type loss and optimizing the resulting objective using the Stochastic Maximum Principle. Numerical experiments on benchmark SPDEs under multiple uncertainty sources demonstrate the accuracy and robustness of the proposed method in capturing solution structure and quantifying predictive uncertainty. [LINK]http://arxiv.org/abs/2605.17107v2 [DATE]2026-10-17 00:44:26+08:00 [CATEGORIES]cs.LG
Escaping Parameter Space: Tight Generalization Bounds via Representation Quality [AUTHORS]Niclas Göring, Shuofeng Zhang, Branton DeMoss, Ard A. Louis [ABSTRACT]We derive a tight generalization bound for deterministic neural We derive a tight generalization bound for deterministic neural networks based on their last hidden layer representations. By evaluating representation quality with a parameter-free $k$-nearest-neighbor ($k$-NN) classifier, our certificate avoids traditional parameter-space complexity measures, KL divergences to a posterior, weight norms, and compression. The bound is tight: trained from scratch, it reaches an 8.0 percentage point gap to test error on CIFAR-10 (ResNet-18) and yields the first non-vacuous certificates we are aware of on CIFAR-100; with frozen DINOv2 representations, it certifies at 1.4\% on CIFAR-10 and 23.1\% on 1,000-class ImageNet. It decomposes additively into three empirical, independently measurable quantities: local class geometry ($k$-NN decoder error), decoder-network agreement, and training stability. This allows the bound to act as a diagnostic tool: we demonstrate that actively destroying local class geometry eliminates generalization while leaving dataset memorization intact, identifying class geometry as a causally relevant mechanism for generalization. Finally, we show that learned representations are significantly less sensitive to weight noise than linear readouts, suggesting why parameter-space certificates often struggle to remain tight. [LINK]http://arxiv.org/abs/2610.23729v1 [DATE]2026-10-17 00:43:57+08:00 [CATEGORIES]cs.LG
ECG-SDI: Single-lead Electrocardiogram Enables Continuous Sleep Depth Index Profiling and Scalable Digital Biomarkers for Sleep Health [AUTHORS]Donglin Xie, Qingshuo Zhao, Songchi Zhou, Haoyu Wang, Jingyu Wang, Yuxi Zhou, Thomas Penzel, Shenda Hong [ABSTRACT]Conventional sleep staging provides a coarse-grained representation that obscures within-stage depth variation. We developed ECG-SDI, a single-lead electrocardiography (ECG) framework for continuous sleep-depth profiling and digital biomarker extraction. Using 12,392 overnight polysomnographic recordings, we trained and validated a two-stage deep learning model supervised by conventional sleep-stage labels. Correlations with ordered stages excluding rapid eye movement (REM) sleep were 0.828 in the pooled internal test set and 0.848 in the external Cleveland Family Study; corresponding REM classification AUROCs were 0.954 and 0.977. Overnight trajectories captured within-stage variation. Biomarker-based phenotyping identified a disturbed subtype associated with poorer subjective sleep quality and higher all-cause mortality in the Sleep Heart Health Study (adjusted hazard ratio, 1.20; 95% CI, 1.05-1.38). Among 507 participants with paired recordings before and after continuous positive airway pressure (CPAP), normalized sleep-depth area increased by 6.1% and shallow-sleep burden below 0.3 decreased by 16.0%. Exploratory analyses linked reductions in approximate entropy and depth variability to reduced oxygen desaturation and increased sleep efficiency, respectively (Spearman rho = 0.24 and 0.12); both associations remained positive after covariate adjustment. Mean ECG-SDI within N2 increased by 0.017 (95% CI, 0.004-0.030). A wearable proof-of-concept illustrated the acquisition-to-report workflow. These findings support ECG-based sleep phenotyping and further evaluation of CPAP response monitoring at home. [LINK]http://arxiv.org/abs/2610.23727v1 [DATE]2026-10-17 00:43:07+08:00 [CATEGORIES]cs.LG
Tool Recommendation for Multi-Step Agents: Finding Every Tool a Task Needs [AUTHORS]Sam Partee, River Stevens, Thierry Damiba, Valerie Fanelle [ABSTRACT]A multi-step agent task fails when a single tool it needs is missing, yet the tools an agent may need are too many to list: the roughly 8,200 tools on the platform we study have definitions that together exceed the context window of any model we evaluate. We present Tool Recommendation, a deployed system that gives an agent all of them, hundreds of times more than a typical MCP server exposes, through a fixed interface of three tools. It runs at the gateway between agent and tools, executing every call under the end user's own credentials and checking it against the user's consent for each service and the tools the agent may use. We evaluate it on both a public retrieval benchmark and billed end-to-end session cost. On ToolRet, Tool Recommendation places every tool a query needs in the top ten 57% more often than BM25, a lexical ranking that commercial tool search offers, and compares favorably with published systems up to nine times its size. In paired multi-step workflows, its first recommendation covered every required tool in 93% of Codex sessions, and against a fixed list of 100 tools it cut input tokens by 17% with Codex and 40% with Claude Code and the total bill in all 14 runs, by up to 40%, with no change in task completion. [COMMENTS]11 pages, 3 figures, 5 tables [LINK]http://arxiv.org/abs/2610.23707v1 [DATE]2026-10-17 00:30:38+08:00 [CATEGORIES]cs.LG
Blitz: Accelerating the Diffusion Modules of AlphaFold3 and Boltz-2 [AUTHORS]Vinay Saji Mathew, Soundar Kumara, Gretta Kellogg, William K. M. Lai [ABSTRACT]Diffusion-based protein structure prediction models produce all-atom structures in about 200 denoising steps, and that cost has driven a number of few-step samplers. We present Blitz, a few-step distillation framework that uses a geometric noise schedule to distribute steps more evenly across noise scales, then learns those updates through multistep consistency training. Physical steering and target-free candidate selection help reduce the remaining geometric defects at inference. We apply Blitz to AlphaFold 3 and Boltz-2, reducing denoising from 200 steps to eight. Because published few-step results use different benchmarks and scoring procedures, we compare both distilled models with their teachers and other few-step methods under a common evaluation protocol covering structural accuracy, interface quality, ligand placement, and stereochemical validity. Blitz retains teacher-level structural accuracy, with AF3-Blitz outperforming all evaluated few-step baselines and exceeding its teacher on several accuracy metrics. [COMMENTS]24 pages, 12 figures [LINK]http://arxiv.org/abs/2610.23703v1 [DATE]2026-10-17 00:29:04+08:00 [CATEGORIES]cs.LG
HardEval: An Execution-Verified Benchmark for AI-Driven Security Hardening and Compliance Remediation [AUTHORS]Manohar Vellala, Soumya Banerjee, Sandip Roy, Ross Gore, Sachin Shetty [ABSTRACT]Implementing Security Technical Implementation Guide (STIG) controls remains a demanding operating-system administration task. Each natural-language security requirement must be translated into a precise and persistent configuration change. These changes span Secure Shell (SSH) configuration, authentication policy, audit rules, and kernel parameters. Existing benchmarks do not measure whether a model can perform this task on a live system. Code benchmarks evaluate isolated programs without persistent system state, while agentbenchmarks largely focus on applications rather than operating-system configuration. In this paper, we introduce \textbf\{HardEval\}, an execution-verified benchmark for evaluating language-model-assisted security hardening. HardEval is built from Red Hat Enterprise Linux 8 STIG controls that fail on an unhardened baseline system. Expert-written reference remediations and Open Vulnerability and Assessment Language (OVAL) checks serve as hidden ground truth for grading; they are never shown to the model, and each task states only the required security outcome while withholding the implementation. The model produces a Bash script, which is executed on a live virtual machine and graded from fail to pass using the corresponding OpenSCAP/OVAL compliance check. Evaluation of fourteen frontier and open-weight configurations reveals large and uneven capability differences across security domains. A recurring failure mode is plausible security code that does not satisfy the exact compliance condition. For four open-weight configurations, multi-run Wilson confidence intervals preserve the observed ordering. HardEval is publicly available. [COMMENTS]8 pages, 4 figures, 4 tables. Code and data: https://github.com/manoharvellala/Hardeval The paper has 4 figure environments (two are images, two are the example-prompt panels) and 4 tables [LINK]http://arxiv.org/abs/2610.23685v1 [DATE]2026-10-17 00:14:58+08:00 [CATEGORIES]cs.LG
AceFF: A State-of-the-Art Machine Learning Potential for Small Molecules [AUTHORS]Stephen E. Farr, Stefan Doerr, Antonio Mirarchi, Alireza Moradzadeh, Francesc Sabanes Zariquiey, Saee Paliwal, Gianni De Fabritiis [ABSTRACT]We introduce AceFF, a pre-trained machine learning interatomic potential (MLIP) optimized for small molecule drug discovery. While MLIPs have emerged as efficient alternatives to Density Functional Theory (DFT), generalizability across diverse chemical spaces remains difficult. AceFF addresses this by refining the TensorNet2 architecture and training it on a comprehensive dataset of drug-like compounds. This approach yields a force field that balances high-throughput inference speed with DFT-level accuracy. We introduce two models sharing the same architecture: \mbox\{AceFF-2\}, trained on our own dataset of small molecules and supporting the essential medicinal chemistry elements (H, B, C, N, O, F, Si, P, S, Cl, Br, I); and \mbox\{AceFF-2-OMOL\}, trained on a subset of OMol25 containing the AceFF-2 elements plus the common biomolecular ion elements Na and K. The models are explicitly trained to handle charged states. Validation against rigorous benchmarks, including complex torsional energy scans, molecular dynamics trajectories, batched minimizations, and tests of force and energy accuracy, demonstrates that AceFF is state-of-the-art for organic molecules in the accuracy and speed regime important for drug discovery. The AceFF-2 and AceFF-2-OMol model weights and inference code are available at https://huggingface.co/Acellera/AceFF-2.0 and https://huggingface.co/Acellera/AceFF-2-OMOL. [LINK]http://arxiv.org/abs/2601.00581v3 [DATE]2026-10-17 00:13:19+08:00 [CATEGORIES]cs.LG
Interpretable experiential learning based on state history and global feedback [AUTHORS]Anton Kolonin [ABSTRACT]A new interpretable experiential learning model based on state history and global feedback is presented. It is capable of learning a behavioral model represented by a transition graph between sets of states, with transitions attributed with utility and evidence count. This model is expected to be suitable for solving reinforcement learning problem in resource-constrained environments. The model was thoroughly evaluated on the OpenAI Gym Atari Breakout benchmark, demonstrating performance comparable to some known neural network-based solutions. [COMMENTS]5 figures. This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible [LINK]http://arxiv.org/abs/2605.00940v2 [DATE]2026-10-17 00:12:02+08:00 [CATEGORIES]cs.LG
2026 Oct 16, Fri
Would you still call this Dax? Novel Visual References in VLMs and Humans [AUTHORS]Ada Defne Tür, Gaurav Kamath, Joyce Chai, Siva Reddy, Benno Krojer [ABSTRACT]Vision-language models (VLMs), like human learners, are frequently exposed to new visual concepts, but how they map novel visual references to language after exposure remains largely underexplored, particularly when those references contradict prior knowledge from pre-training. To study this, we present the Novel Visual References Dataset (NVRD): 19,176 images spanning 90 visual concepts across different levels of visual novelty, each with up to 20 increasingly perturbed versions of the original object to probe generalization. Unlike prior work on visual augmentations of familiar concepts, NVRD comprises entirely novel, open-ended stimuli constructed from scratch, mirroring how humans encounter genuinely new concepts. We evaluate 3 open- and 2 closed-source models alongside 2,400 human judgments for direct human-model comparison, and find that (i) models struggle to acquire novel concepts in-context when they contradict prior knowledge, and (ii) while models and humans show correlated sensitivity to visual perturbations, models significantly overgeneralize, extending learned labels to stimuli that humans reject. We contribute NVRD as a corpus and benchmark for research on visual concept learning in both humans and machines. [COMMENTS]EMNLP Main Conference 2026 [LINK]http://arxiv.org/abs/2606.05409v4 [DATE]2026-10-16 23:49:10+08:00 [CATEGORIES]cs.CL
Mitigating Cross-Lingual Cultural Inconsistencies in LLMs via Consensus-Driven Preference Optimisation [AUTHORS]Lucas Resck, Isabelle Augenstein, Anna Korhonen [ABSTRACT]Multilingual large language models (MLLMs) often answer culturally ambiguous questions inconsistently across languages, even for a fixed user persona -- e.g., Shakespeare in English but Cervantes in Spanish for the same British user. To robustly quantify this Cross-lingual Cultural Inconsistency, we introduce Singleton Fleiss's $κ_S$, a metric mathematically resilient to hallucinations. For mitigation, we propose Cross-lingual Cultural Consistent Preference Optimisation (C-3PO), a consensus-driven alignment framework. C-3PO raises $κ_S$ by 0.07--0.10 over unaligned models, above every prompting and representation steering baseline averaged across personas, whilst largely preserving explicit user identities, cultural neutrality and intrinsic cultural knowledge; where it does not, a soft-consensus variant recovers most of the cost. Although trained only on multiple-choice questions, C-3PO also improves consistency in open-ended generation. Empirical evaluations demonstrate this inconsistency is more pronounced for lower-resource languages like Indonesian and Persian. Finally, early decoding of intermediate layers reveals that MLLMs implicitly personalise outputs towards the prompt language's stereotypical culture as forward-pass representations stabilise. [COMMENTS]25 pages, 14 figures, 17 tables [LINK]http://arxiv.org/abs/2605.12515v3 [DATE]2026-10-16 23:46:19+08:00 [CATEGORIES]cs.CL
PresenceBench: An Exploratory Evaluation of Large Language Model Adaptation to User State [AUTHORS]Peña Arroyo Gallego, Mauro Ruiz Moreno, Alejandro Acien, Sergio Romero-Tapiador, Teresa Arroyo-Gallego [ABSTRACT]Presence-aware intelligence is the capacity of an AI system's intelligence layer to adapt appropriately to the functional state of the person it serves (here readiness, flow, and stability) while preserving task quality and safety. We introduce PresenceBench, an instrument for measuring it. PresenceBench evaluates how the intelligence layer uses a supplied state, not how that state is sensed, and breaks the capability into measurable properties. Two are tested here. Receptivity: does the model take up the state and give a better response than with no signal? Specificity: does any improvement depend on the correct state, rather than on extra structured context? A third, adaptation across states, is a required endpoint not measured in this release. A five-arm causal design answers both questions, against each model's own baseline and an equally structured mismatched-state control, scored by a correctness-gated human paired comparison. The exploratory evaluation covered four models across eighteen single-turn, six fixed three-turn and six simulated natural-conversation scenarios. Supplying a synthetic state, raw or translated into a response policy, gave human-preference point estimates above 0.5 for all four models (Claude Sonnet 5, DeepSeek V4 Pro, Gemini 3.6 Flash, GPT-5.6 Terra), with three of eight 95% intervals excluding chance. Specificity was not resolved: no per-model interval excluded chance, and the design cannot yet separate the state values from a generic instruction to adapt. Policy delivery reduced priced cost and output tokens against each model's baseline; these are resource-use changes, not efficiency gains, because correctness non-inferiority was not established. Safety-floor failures were rare (13 of 360 cases), with a non-significant excess under policy delivery. We report the instrument, the frozen results, and the controls a confirmatory study needs. [COMMENTS]33 pages, 6 figures, 13 tables [LINK]http://arxiv.org/abs/2610.23625v1 [DATE]2026-10-16 23:33:04+08:00 [CATEGORIES]cs.CL
SIT-Eval: Situation-Based Interactive Turn-by-Turn Evaluation of Conversational Agents [AUTHORS]Ravil Mussabayev, Saveliy Zavertyaev, Dmitriy Parpulov, Andrey Savchenko [ABSTRACT]The rapid proliferation of conversational AI agents necessitates evaluation methods that assess performance across diverse dialogue scenarios, utilizing multiple situational criteria, as appropriate behavior depends on context and encompasses various aspects. Existing dialogue benchmarks typically use static natural conversations, which underrepresent rare situations---a critical safety gap, as low-probability events become frequent failures at production scale. However, limited sample sizes force a choice: either maintain natural dialogues for realism or turn to adaptive situation sampling for comprehensive yet statistically reliable coverage. We introduce SIT-Eval, an adjustable framework for evaluating situation-based conversational agents. We formalize an agent's desired behavior as situational rules: if situation is P, follow rules Q. At each conversation turn, an LLM-interlocutor enacts a uniformly sampled situation. A context-aware LLM-judge leverages all conditionally applicable rules as criteria to evaluate the agent's reply for that situation. The judge is validated using human-verified synthetic golden dialogues. Benchmarking 8 LLMs on 1292 conversation turns reveals failure modes hidden by aggregate metrics: models with similar overall scores vary widely in their adherence to individual rules. All models struggle most with ambiguity clarification and Socratic questioning. [LINK]http://arxiv.org/abs/2610.23612v1 [DATE]2026-10-16 23:24:02+08:00 [CATEGORIES]cs.CL
Not Every Mention Is a Claim: Rethinking Object Hallucination Detection in Vision-Language Models [AUTHORS]Hsin-Ling Hsu, Neeraja Kirtane, Kuan-Hao Huang [ABSTRACT]Object hallucination in vision-language models (VLMs) is usually measured by checking each object a caption names against the image. Most benchmarks and detectors alike treat every object mention as a claim that the object is present. This assumption fails when a caption names an object without asserting it, as in "a bench for people to sit on", which mentions people without claiming that any are in the image. We call such mentions non-assertive. Treating them as hallucinations inflates the error rates attributed to models that use such language. To separate them from real hallucinations, we introduce NAME-Bench (Non-Assertive Mention EvaluationBenchmark), 25,497 object mentions from captions that five VLMs generate for COCO images. Each mention is labelled concrete (asserted and present), hallucinated (asserted and absent), or non-assertive, and non-assertive mentions are further divided into four subtypes: depictive, purpose, negation and hedged. All the hallucination detectors that we evaluate flag non-assertive mentions as hallucinations, at rates between 14% and 80%. Adding non-assertive mentions to detector training as negatives largely corrects this on LLaVA-1.5. On the four newer VLMs, a comparable correction takes five to ten times more data, and the detectors then catch fewer real hallucinations. These findings show that current detectors conflate mentioning an object with asserting it, and that this conflation is harder to undo on the four newer VLMs than on LLaVA-1.5. Hallucination evaluation should therefore separate what a caption asserts from what it merely mentions. [LINK]http://arxiv.org/abs/2610.23610v1 [DATE]2026-10-16 23:23:45+08:00 [CATEGORIES]cs.CL
Enhancing RAG-based Medical VQA with Gated Multimodal Representations [AUTHORS]Daniela Jimenez-Pano, Hugo Jair Escalante, Delia Irazú Hernández-Farías [ABSTRACT]Medical Vision-Language Models (Med-LVLMs) have shown strong performance in Medical Visual Question Answering (MVQA), yet they may still produce factually inconsistent responses. Retrieval-Augmented Generation (RAG) can mitigate this issue by incorporating external clinical evidence; however, retrieval is often guided by a single modality, potentially overlooking the joint semantics of medical images and clinical questions. We introduce a multimodal RAG framework that learns a joint query representation through a Gated Multimodal Unit, enabling evidence retrieval to leverage information from both modalities while keeping the underlying Med-LVLM frozen. We evaluate the proposed approach on benchmark datasets. Results show that the multimodal query representation consistently outperforms a reference Med-LVLM and remains competitive with state-of-the-art methods. Our findings demonstrate the potential of multimodal query representations for more effective evidence retrieval in Medical VQA. [COMMENTS]Preprint submitted to ECIR [LINK]http://arxiv.org/abs/2610.23588v1 [DATE]2026-10-16 23:07:33+08:00 [CATEGORIES]cs.CL
PragReST: Self-Reinforcing Counterfactual Reasoning for Pragmatic Language Understanding [AUTHORS]Jihyung Park, Minchao Huang, Leqi Liu, Elias Stengel-Eskin [ABSTRACT]Natural language understanding often depends on meanings that are implied rather than explicitly stated, requiring pragmatic reasoning. Despite strong performance on math and logical reasoning, large language models (LLMs) still struggle with making pragmatic inferences, often choosing literal interpretations. To improve LLM pragmatic reasoning, we introduce PragReST, a self-supervised framework that constructs pragmatic QA data, generates counterfactual reasoning traces, and trains models to internalize them through supervised fine-tuning and reinforcement learning, without human-labeled training data or distillation from a stronger teacher. Across four pragmatic benchmarks (PragMega, Ludwig, MetoQA, and AltPrag), PragReST improves over backbone models, task-specific pragmatic tuning baselines, and non-counterfactual variants of the same pipeline. On accuracy-based benchmarks, PragReST improves over the instruct backbone by 5.37 and 5.50% (absolute) for Qwen3-8B and Qwen3-14B, respectively. Our error analysis and ablations underscore the importance of counterfactual reasoning: PragReST primarily reduces errors caused by failures to contrast observed utterances with plausible alternatives, and removing counterfactual reasoning substantially reduces performance. Moreover, our training preserves out-of-domain performance on general-knowledge and mathematical reasoning benchmarks. [COMMENTS]EMNLP Camera-Ready. First two authors contributed equally. Code and models: https://github.com/jihyung803/PragReST [LINK]http://arxiv.org/abs/2606.18624v2 [DATE]2026-10-16 22:54:52+08:00 [CATEGORIES]cs.CL
Execution Before Memory: Diagnosing Memory Failures in Video World Models [AUTHORS]Weiyang Jin, Jiwen Yu, Sihan Xu, Yuwei Niu, Xihui Liu [ABSTRACT]A memory test for a video world model should say whether the scene was set, whether the asked change was carried out, and whether the last frame shows the result. Current tests fold these into one number, so a low score is easy to read as forgetting. When the change is not carried out, the hidden state transition cannot appear, so the test is not really a test of memory. We introduce MIST, an evaluation that separates the change from the memory it should leave behind. The model first sets the scene while the change stays hidden, and only later is asked to show it, with leakage before that point kept under control. This separation locates the bottleneck: the model sets the scene but does not carry out the instructed change. The same gap appears in general memory evaluations. On that basis we propose M3-Agent (Make the Missing Move), which recursively revises the details of the input instruction so that the model successfully carries out the requested instruction. Relative to the baseline, it more than triples the MIST reveal score. On the benchmarks, WRBench's score T rises by 26.1% and MBench-T's score U by 52.7%. The method meets the problem from the execution side: it shows what the frozen model can do once the move is stated, and it sets a stricter baseline. [COMMENTS]27 pages, 7 figures. Project page: https://hku-mmlab.github.io/MIST/ Code: https://github.com/HKU-MMLab/MIST [LINK]http://arxiv.org/abs/2610.23542v1 [DATE]2026-10-16 22:37:13+08:00 [CATEGORIES]cs.CL
Referring Multi-Object Tracking with Diverse Dynamic Expressions [AUTHORS]Yani Zhang, Dongming Wu, Wencheng Han, Xingping Dong, Shengcai Liao, Bo Du [ABSTRACT]Referring understanding is a fundamental task that bridges natural language and visual content by localizing objects described in free-form expressions. However, existing works are constrained by limited language expressiveness, lacking the capacity to model object dynamics in spatial numbers and temporal states. To address these limitations, we introduce a new and general referring understanding task, termed referring multi-object tracking (RMOT). Its core idea is to employ a language expression as a semantic cue to guide the prediction of multi-object tracking, comprehensively accounting for variations in object quantity and temporal semantics. Along with RMOT, we introduce a RMOT benchmark named Refer-KITTI-V2, featuring scalable and diverse language expressions. To efficiently generate high-quality annotations covering object dynamics with minimal manual effort, we propose a semi-automatic labeling pipeline that formulates a total of 9,758 language prompts. In addition, we propose TempRMOT, an elegant end-to-end Transformer-based framework for RMOT. At its core is a query-driven Temporal Enhancement Module that represents each object as a Transformer query, enabling long-term spatial-temporal interactions with other objects and past frames to efficiently refine these queries. TempRMOT achieves state-of-the-art performance on both Refer-KITTI and Refer-KITTI-V2, demonstrating the effectiveness of our approach. The source code and dataset is available at https://github.com/zyn213/TempRMOT. [LINK]http://arxiv.org/abs/2406.05039v3 [DATE]2026-10-16 22:34:19+08:00 [CATEGORIES]cs.CL
Perceive, Solve, Reason: Progressive Capability Formation in Time Series-LLMs [AUTHORS]Xin Qiu, Junlong Tong, Wei Zhang, Xiaoyu Shen [ABSTRACT]Large language models (LLMs) can now perceive, describe, and reason about images and audio, yet time series understanding remains underestimated: most work uses LLMs as forecasters and classifiers, reasoning studies are few and narrow, and no suite trains and evaluates the full range of capabilities. In this paper, we build the first such suite, with 337K training and 44K evaluation samples across three progressive capabilities, Perceive, Solve, and Reason, and use it to ask how time series should enter an LLM that must master all three, and in what order. We find that a dedicated encoder is essential: with one, a 4B model surpasses GPT-4o on every perception attribute we test, whereas textualizing the series costs 8--10 points in reasoning accuracy. The capabilities then build on one another: learning them in order beats mixing the same data and rivals reasoning-only training with over three times the reasoning data, but only when an encoder is used. Analyses further reveal that capabilities are transferred through distinct network depths rather than simply stacked, and that a stage converged on its own benchmark is not necessarily ready for the next. Distilled into a simple recipe, these findings yield 4B TS-LLMs that, without any target-domain data, outperform four recent 7B TS-LLMs on seven of eight out-of-domain metrics, by up to 11 points. We therefore advocate treating time series as a first-class modality for LLMs, read through dedicated encoders and taught to perceive before they reason. [LINK]http://arxiv.org/abs/2610.23521v1 [DATE]2026-10-16 22:23:23+08:00 [CATEGORIES]cs.CL
MuninAI: A Self-Hosted Agentic Framework for Independent Research Groups [AUTHORS]Markus Fischer, Jens Meiler, Matthias Elgeti [ABSTRACT]Research groups that want a language model grounded in their own literature face a choice between hosted services that receive their unpublished work and self-hosted deployments that keep the data local but give the model no access to the corpus. Munin is an open-source framework for a self-hosted scientific assistant that indexes a group's corpus and its citation graph, and wraps whichever open-weight model the group can serve in an agent harness with specialised reading, search, and computation tools. We claim that the harness, rather than the specific model, is what makes the system useful on domain literature, and we ship the evaluation suite that tests this claim with the software.
On 199 paired literature questions, the harness more than doubles the accuracy of the base model, from 0.387 to 0.874 (+0.487, 95% CI [0.407, 0.568]), while naive retrieve-then-answer scores below the unaided model on every backbone tested. Under the harness, roughly twice as many answer claims trace to a passage the model read itself as with fixed top-5 retrieval. When the source paper for a question is removed from the corpus, the system abstains on about nine in ten of the questions it had answered correctly on the Qwen backbones and eight in ten on gpt-oss-20b, and it produces no fabricated local citation across 300 fabricated references.
The pattern holds on three backbones from two developers, including a 20-billion-parameter model that runs on a single consumer-grade GPU. With all outbound network access disabled, both Qwen backbones still exceed the published PaperQA2 accuracy. The evaluation suite runs as a single command, so an adopting group can re-run it on its own corpus and model. [COMMENTS]37 pages, 3 figures, 20 tables. Code, evaluation suite and scorecards: https://github.com/elgetilab/munin (archived release v1.0.0: https://doi.org/10.5281/zenodo.23191404) [LINK]http://arxiv.org/abs/2610.23480v1 [DATE]2026-10-16 21:56:37+08:00 [CATEGORIES]cs.CL
Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching [AUTHORS]Roy Miles, Aysim Toker, Andreea-Maria Oncescu, Jiankang Deng, Ismail Elezi [ABSTRACT]Reasoning with large language models often benefits from generating multiple chains-of-thought, but existing aggregation strategies are typically trajectory-level (e.g., selecting the best trace or voting on the final answer), discarding useful intermediate work from partial or "nearly correct" attempts. We propose Stitching Noisy Diffusion Thoughts, a self-consistency framework that turns cheap diffusion-sampled reasoning into a reusable pool of step-level candidates. Given a problem, we (i) sample many diverse, low-cost reasoning trajectories using a masked diffusion language model, (ii) score every intermediate step with an off-the-shelf process reward model (PRM), and (iii) stitch these highest-quality steps across trajectories into a composite rationale. This rationale is then used to recompute only the final answer. This modular pipeline separates exploration (diffusion) from evaluation and solution synthesis, avoiding monolithic unified hybrids while preserving broad search. Across math reasoning benchmarks, we find that step-level recombination is most beneficial on harder problems, and ablations highlight the importance of the final solver in converting stitched but imperfect rationales into accurate answers. Using low-confidence diffusion sampling with parallel, independent rollouts, our training-free framework improves average accuracy by up to 23.8% across six math and coding tasks. At the same time, it achieves up to a 1.8x latency reduction relative to both traditional diffusion models (e.g., Dream, LLaDA) and unified architectures (e.g., TiDAR). The code is available https://github.com/roymiles/diffusion-stitching [COMMENTS]NeurIPS 2026. Code available at https://github.com/roymiles/diffusion-stitching [LINK]http://arxiv.org/abs/2602.22871v3 [DATE]2026-10-16 21:55:50+08:00 [CATEGORIES]cs.CL
Cross-Layer Attention Probing for Fine-Grained Hallucination Detection [AUTHORS]Malavika Suresh, Rahaf Aljundi, Ikechukwu Nkisi-Orji, Nirmalie Wiratunga [ABSTRACT]With the large-scale adoption of Large Language Models (LLMs) in various applications, there is a growing reliability concern due to their tendency to generate inaccurate text, i.e. hallucinations. In this work, we propose Cross-Layer Attention Probing (CLAP), a novel activation probing technique for hallucination detection, which processes the LLM activations across the entire residual stream as a joint sequence. Our empirical evaluations using five LLMs and three tasks show that CLAP improves hallucination detection compared to baselines on both greedy decoded responses as well as responses sampled at higher temperatures, thus enabling fine-grained detection, i.e. the ability to disambiguate hallucinations and non-hallucinations among different sampled responses to a given prompt. This allows us to propose a detect-then-mitigate strategy using CLAP to reduce hallucinations and improve LLM reliability compared to direct mitigation approaches. Finally, we show that CLAP maintains high reliability even when applied out-of-distribution. Code available at: https://github.com/itsmemala/CLAP [COMMENTS]Published at the TRUST-AI workshop, ECAI 2025 [LINK]http://arxiv.org/abs/2509.09700v2 [DATE]2026-10-16 21:07:39+08:00 [CATEGORIES]cs.CL
On the Tip of the Tongue: First-Token Cueing as a Behavioral Probe of Recall Failures in LLMs [AUTHORS]Jaewon Sok, Jewon Yeom, Seonghyeon Park, Jeongjae Park, Hwiyeong Lee, Taesup Kim [ABSTRACT]Large language models often answer factual questions incorrectly in closed-book settings, and such errors are usually attributed to missing knowledge. Yet models often encode answers they fail to produce. We show that many closed-book errors are first-token recall failures: the model commits to a wrong first token, but once given the right one it completes the answer and often recognizes it as correct. We study this with First-Token Cueing, a behavioral probe that prefills the response with the first token of the gold answer. Model uncertainty peaks at the first answer token, where it already predicts correctness, and among gold prefixes of any length the first token alone accounts for the largest share of recovered errors. Across 16 open models, this one token corrects between a fifth and a half of the errors on multi-token answers in English QA and improves accuracy in all seven languages of a multilingual benchmark. Most of the errors it corrects are also solved by the model's own sampled reasoning, a criterion for encoded knowledge, and models recognize correct starts they do not produce: they continue a correct start more confidently than an incorrect start of the same answer type and prior probability, yet rarely find it without the gold answer. These results locate a recoverable class of factual errors at the first decoding step, where inference-time methods may gain the most. [LINK]http://arxiv.org/abs/2610.23400v1 [DATE]2026-10-16 20:55:36+08:00 [CATEGORIES]cs.CL
FSPO: Policy-Consistent Risk and Pareto-Feasible Control for Budgeted LLM RL Post-Training [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Daren Zha [ABSTRACT]Adaptive LLM reinforcement-learning post-training changes multiple training actuators online, including rollout temperature, group size, clipping, KL regularization, verifier allocation, and update budget. Three coupled issues remain unresolved. A future-risk model trained from behavior trajectories need not estimate the risk induced by the controller that will be deployed; a score calibrated on logged state-action pairs can become miscalibrated after selective action choice; and independent per-resource minimum costs do not in general certify a feasible multi-resource continuation. We introduce FSPO, a feedback-state controller for budgeted LLM RL post-training that addresses these issues jointly. FSPO learns a policy-consistent risk-to-go model whose Bellman target follows the same frozen controller used for future decisions, together with a long-horizon utility model. Decision-conditioned trajectory calibration (DCTC) calibrates risk on cross-fitted trajectories generated by actions selected by provisional controllers. A Pareto resource continuation certificate (PRCC) admits an action only when a non-dominated cumulative reservation remains feasible over the residual horizon. Under a matched GRPO resource envelope, FSPO reaches 66.11% held-out and 59.43% OOD accuracy, compared with 64.47% and 57.03% for PB2, the strongest evaluated adaptive baseline. Three paired training seeds give gains of +2.42 and +3.19 percentage points over the contextual bandit on held-out and OOD evaluation. Under high behavior-deployment mismatch, policy-consistent risk lowers selected-decision ECE from 0.108 to 0.053; DCTC lowers it from 0.039 to 0.022 at matched acceptance; PRCC removes false-feasible admissions on an 18-action catalog ($0.197\rightarrow0.000$); and enabling all three components reduces trajectory failure from 0.181 to 0.083 in a factorial ablation. [COMMENTS]40 pages, 4 figures [LINK]http://arxiv.org/abs/2610.02828v2 [DATE]2026-10-16 20:44:10+08:00 [CATEGORIES]cs.CLcs.LG
ContractRL: Shielded Group-Relative Policy Optimization for Auditable Tool-Call Repair [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Yina Sa, Daren Zha [ABSTRACT]Structured tool calls often fail after only a small number of fields violate a schema or an execution contract. Regenerating the complete object enlarges the action surface and makes repeated repair difficult to audit. We introduce ContractRL, a contract-constrained sequential repair protocol that models verifier-guided JSON repair as a bounded decision process. At each step the policy observes the candidate, typed verifier feedback, JSON Pointer, immutable repair history, and remaining budget; a contract-derived action mask filters malformed or prohibited RFC-6902 operations before a deterministic validator performs the transition. We specify a contract-constrained group-relative objective for patch, retry, and abstention decisions while keeping canonical targets and semantic labels outside the online state until trace freeze. Under identical verifier information, ContractRL attains 0.9362 semantic success with 34.4 generated tokens, compared with 0.9076 and 44.9 tokens for Patch-SFT and 0.9148 and 137.2 tokens for full regeneration over 192 cases per seed and five seeds. Policy optimization improves semantic success from 0.9186 for supervised ContractRL to 0.9375. A separate three-seed paired evaluation against Patch-SFT yields a semantic difference of +0.0396 (95% CI $[+0.0137,+0.0662], p=0.0039$). Feedback, action-mask, budget, and schema-shift analyses connect these gains to localized correction, while adversarial and multi-turn evaluations characterize the remaining failure modes. [COMMENTS]29 pages, 8 figures [LINK]http://arxiv.org/abs/2610.00328v2 [DATE]2026-10-16 20:23:26+08:00 [CATEGORIES]cs.CLcs.LG
Dependency Reasoning for Aspect Sentiment Quadruple Prediction [AUTHORS]Shima Esfandyari, Seywan Soleymanbaigi, Fardin Akhlaghian Tab, Fatemeh Daneshfar, Mourad Oussalah [ABSTRACT]Aspect Sentiment Quadruple Prediction (ASQP) is fundamentally a relational reasoning problem: identifying aspect-opinion-category-sentiment quadruples requires modeling structured dependencies, not merely sequential co-occurrence. Existing LLM-based approaches incorporate syntactic information indirectly, via retrieval-based demonstration selection or multi-stage training, rather than as unified, single-pass generation-time context. We present DRIFT, which linearizes dependency parse trees and injects them directly into LLM prompts as explicit relational context, enabling the model to reason over aspect-opinion dependencies rather than infer them from token proximity alone. DRIFT further integrates retrieval-augmented in-context learning and a syntax-aware LLM-based adjudication mechanism that complements exact-match evaluation under span-boundary variation. On SemEval Rest15, Rest16, and ACOS, DRIFT achieves the strongest performance among compared generative ASQP methods, reaching 54.47% and 64.10% F1 on Rest15 and Rest16, with consistent gains across short- and medium-distance aspect-opinion configurations; adjudication further shows exact-match metrics systematically underestimate semantically correct predictions. [LINK]http://arxiv.org/abs/2610.14873v2 [DATE]2026-10-16 20:03:17+08:00 [CATEGORIES]cs.CLcs.LG
FAER: Auditable Utility-Aligned Trajectory Replay for Language Model Post-Training [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Tianshu Fu, Daren Zha [ABSTRACT]Replay selectors often rank cached trajectories by format feedback, confidence, freshness, or response length, although cache-level correctness and downstream learner utility are distinct objectives. We formalize this selection-to-learning gap and introduce FAER as an auditable full-trajectory replay framework. Its training-free fixed selector is a protocol baseline; FAER-UTILITY is the learner-aware selector fitted on disjoint calibration blocks. The normalized gradient alignment is reported as a baseline, while a disposable optimizer-aware virtual update supplies a magnitude-aware utility surface. The audit contract freezes observed fields and replay traces before evaluation labels are joined. On GSM8K with Qwen2.5-1.5B-Instruct, the matched learner study reports quality 0.6329 for the fixed selector, compared with 0.5482 for uniform and 0.6037 for format-feedback under 128 updates. Metadata-only cross-fitted calibration reaches $0.6476\!\pm\!0.0139$ over eight seeds (median 0.6481; paired 95% interval $[+0.079,+0.122]$) at 63,276 target-run tokens; its recorded full cost is 189,642 tokens and 3.48 GPU-hours including calibration. The completed FAER-UTILITY row reaches 0.6624 at 62,844 target-run tokens and 4.26 GPU-hours. Format-feedback selects records with correctness 0.6953, compared with 0.3594 for the fixed selector, despite the different downstream ranking. The completed comparison surfaces report the learner-aware ablation, same-seed gap, policy-optimization rows, and strict zero-shot transfer. [COMMENTS]35 pages, 6 figures [LINK]http://arxiv.org/abs/2610.00385v2 [DATE]2026-10-16 20:01:19+08:00 [CATEGORIES]cs.LGcs.CL
Language Models Learn to Improve Language Models [AUTHORS]Minhua Lin, Zijun Wang, Zewen Liu, Zhaotian Weng, Bing He, Zhan Shi, Benoit Dumoulin, Xiang Zhang, Suhang Wang, Hanqing Lu [ABSTRACT]Large language models (LLMs) are increasingly used to automate AI research, including language-model (LM) development, offering a potential path toward recursive self-improvement. However, turning experimental feedback into a learned ability to propose effective improvements across models faces two challenges. First, training such an ability with real feedback is expensive: reinforcement learning (RL) requires evaluating tens of thousands of proposals, each potentially requiring a full model-training run. Even at ten GPU-hours per run, reward collection alone would consume hundreds of thousands of GPU-hours, motivating a cheaper surrogate. Second, surrogate feedback can become unreliable as policy optimization favors overestimated proposals when the proposal distribution shifts. To address these challenges, we propose LM2LM, which trains LLMs with RL to generate effective LM improvement proposals. To provide scalable feedback, we use a research world model (RWM), a frozen LLM that predicts proposal gains from real experimental records provided in context, as the reward model for RL. To maintain feedback reliability, we periodically execute current-policy proposals ranked highly by the RWM and add their measured outcomes to its context. Across eight autoregressive and diffusion LM development environments, real-execution evaluations on held-out tasks show that LM2LM enables Qwen3.8-27B to outperform GPT-6-sol, even when GPT-6-sol is given 300 real experimental records in context, while requiring 33% lower data-generation cost. LM2LM also generalizes to a model family unseen during training. [LINK]http://arxiv.org/abs/2610.23344v1 [DATE]2026-10-16 19:51:00+08:00 [CATEGORIES]cs.CL
Audit-First VAPO: Risk-Certified Selective Updates under Imperfect Verification [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Rui Chen, Daren Zha [ABSTRACT]Imperfect verifiers can assign a harmful update direction even when clipping and regularization bound its magnitude. We introduce Audit-First VAPO, which separates discrete directional admission from continuous magnitude control. An observation-only accept-appeal-abstain policy uses a finite secondary-verification budget; its action trace is frozen before clean labels are joined. Simultaneous finite-sample bounds then certify selected harmful risk, coverage, and verifier-call rate over a predeclared policy family. Conditional Hoeffding-Azuma bounds account for the dependence induced by shared budgets, and rollout or verifier changes initiate a new certification stage. After admission, a bounded trust-clip-KL actuator controls magnitude. We evaluate two models on two reasoning benchmarks against static RLVR, matched-random selection, confidence thresholding, noise correction, and verifier augmentation. On Qwen3.5-0.8B and GSM8K at target risk $ρ=0.08$, RC-VAPO achieves 74.1% accuracy, selected harmful risk 0.0697, coverage 0.4125, and relative verifier cost $1.16\times$. At matched coverage and update magnitude, its selected-risk difference from matched random is -0.0260 with paired 95% interval $[-0.0364,-0.0157]$. Across asymmetric, confidence-dependent, and correlated-verifier noise, the certificate is satisfied on 57 of 60 independent runs. These comparisons isolate informative directional selection from proposal suppression, update shrinkage, and additional verifier computation. [COMMENTS]33 pages, 7 figures [LINK]http://arxiv.org/abs/2609.33662v2 [DATE]2026-10-16 19:41:12+08:00 [CATEGORIES]cs.CLcs.LG
RubricEM: Meta-RL with Rubric-guided Policy Decomposition beyond Verifiable Rewards [AUTHORS]Gaotang Li, Bhavana Dalvi Mishra, Zifeng Wang, Jun Yan, Yanfei Chen, Chun-Liang Li, Long T. Le, Rujun Han, George Lee, Hanghang Tong, Chen-Yu Lee, Tomas Pfister [ABSTRACT]Deep research agents plan, search, evaluate evidence, and synthesize long-form reports. Training these agents with reinforcement learning (RL) is challenging because they operate over long horizons and their outputs lack exact answer verifiers. To this end, we introduce RubricEM, a unified RL framework for deep research that combines stagewise credit assignment with reflection-based meta-RL for experience reuse. Rubrics unify the framework by providing a shared interface for policy execution, judge feedback, and experience reuse. RubricEM first instills a rubric-guided scaffold through supervised fine-tuning, teaching the agent to plan with rubrics and organize its trajectory into semantic stages. Building on this structural prior, we optimize the task policy with Stage-Structured GRPO, using stagewise judge feedback for finer-grained credit assignment. In parallel, we asynchronously train a shared-backbone reflection meta-policy (with meta-RL) to distill completed trajectories into reusable guidance, contributing through both shared-parameter updates and textual memory. The resulting RubricEM-8B model achieves strong performance across four long-form research benchmarks, outperforming existing open models of larger sizes and approaching proprietary deep-research systems. Beyond final performance, we perform thorough controlled analyses to understand the key ingredients of RubricEM. [COMMENTS]67 pages, 6 figures [LINK]http://arxiv.org/abs/2605.10899v2 [DATE]2026-10-16 19:35:29+08:00 [CATEGORIES]cs.CLcs.LG
CascadeLedger: Counterfactual Auditing of Information and Resource Integrity in LLM Cascades [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Daren Zha [ABSTRACT]We introduce CascadeLedger, a counterfactual execution-audit framework for budgeted language-model cascades. The protocol enforces call-before-read semantics, binds successful reads to service and cache identities, records availability and component-wise budget transitions, and freezes the event trace before any target or task scorer is joined. The central object is aggregate-equivalent trace validity: two executions can share a final answer, declared resource projection, and service histogram while differing in whether their routing decisions used only legitimately observable information. We complement single-trace invariants with counterfactual prefix non-interference (CPNI), which couples a visible prefix and policy randomness while changing only a declared hidden family. The counterfactual influence frontier (CIF) localizes the earliest violating prefix and a minimal hidden witness; projection-preserving adversarial mutation (PPAM) searches for invalid traces that remain acceptable to a non-counterfactual checker. Across frozen-cache, independently generated, and serving evaluations, the two-layer audit reports a clean false-positive rate of 0.0000, detects 0.9983 of serialized faults and 0.9982 of unlogged hidden-read faults, and changes 0.3403 of corrupted quality-resource conclusions after deterministic recomputation. The online recorder and offline CPNI audit have separate overhead measurements. Learned routing and semantic verification provide downstream compatibility studies under the same execution contract. [COMMENTS]38 pages, 6 figures [LINK]http://arxiv.org/abs/2610.20668v2 [DATE]2026-10-16 19:25:48+08:00 [CATEGORIES]cs.CLcs.LG
SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models [AUTHORS]Hongxing Li, Jinyue Su, Dingming Li, Wenqi Zhang, Weiming Lu, Jun Xiao, Yueting Zhuang, Yongliang Shen [ABSTRACT]Existing spatial reasoning benchmarks mainly test spatial perception: reading off relations already visible in the input. Yet real-world spatial intelligence demands predictive spatial reasoning: constructing a scene from observations, anticipating how an intervention changes it, and reasoning about the unseen outcome. We introduce SpaceCast-Bench, the first benchmark to directly and diagnostically evaluate this capability. Built around an observe-transform-infer framework, its 3,862 questions from 182 real-world scenes span 16 task types at three levels: static perception, local prediction, and global prediction, progressively requiring scene understanding, spatial state updating, and relational inference over unobserved outcomes. Evaluating 21 models exposes a stark gap: the strongest model reaches only 58.0% against 87.2% human performance, while spatially specialized models remain near random chance. Controlled analyses further reveal that bridge views are critical for integrating distributed observations, and that explicit 3D evidence benefits models more reliably than generated outcome images or videos. Fine-tuning on our programmatically generated data lifts Qwen3-VL-4B from 34.0% to 65.7% with macro-average gains across six out-of-domain benchmarks. [COMMENTS]Project Page: https://zjureal.com/SpaceCast-Bench Code: https://github.com/ZJU-REAL/SpaceCast-Bench [LINK]http://arxiv.org/abs/2610.12402v2 [DATE]2026-10-16 18:40:25+08:00 [CATEGORIES]cs.CL
Reasoning Models Can Control Their Chains of Thought When They Tell Themselves To [AUTHORS]Uriel Dolev, Idan Schwartz, Yoav Goldberg [ABSTRACT]Chain-of-thought (CoT) controllability, the ability of reasoning models to control properties of their CoT, is consequential for both their usefulness and reliable oversight. Recent work evaluates this capability by measuring whether models follow CoT constraints given in the user prompt, and reports very low controllability across models. We argue that this low controllability may not reflect an inherent inability of reasoning models to control their CoT, but instead result from naively specifying constraints in the user prompt. To test this, we introduce Self-Instructed CoT Controllability: a complementary evaluation that places the constraint as a self-instruction inside the model's CoT and measures its ability to follow it. We show that, under self-instruction, models achieve significantly higher controllability than with standard user instructions. Even a generic reminder to follow the user's constraint substantially improves controllability. Moreover, effective self-instructions can be bootstrapped into the reasoning through a simple instruction in the user prompt, enabling improved control even without access to the reasoning channel. Finally, using self-instruction, we demonstrate in a realistic monitoring setup that CoT controllability can be leveraged for monitor evasion, posing a risk to CoT monitorability. [COMMENTS]26 pages, 19 figures, 18 tables. Code and data: https://github.com/udolev/Self-Instructed-CoT-Controllability [LINK]http://arxiv.org/abs/2610.23257v1 [DATE]2026-10-16 18:31:27+08:00 [CATEGORIES]cs.CLcs.LG
Sample, Mix, or Verify? Allocating Test-Time Compute under Imperfect Verifiers [AUTHORS]Srinivasan Iyengar [ABSTRACT]A common way to make language models and agents more accurate is to spend more compute at inference time: generate many candidate solutions and submit the one a verifier prefers. Success then requires that some candidate is correct and that the verifier picks it. The first grows predictably with the number of samples; the second does not, because real verifiers (unit tests, majority votes, LLM judges, reward models) accept wrong answers, reject right ones, and make correlated mistakes. With a fixed budget, a practitioner must decide how many samples to draw, which models to draw them from, and how much to spend on verification, and today these choices are made by rules of thumb. We introduce SPLIT, which makes them from a small labelled pilot. It rests on an exact formula for the probability of submitting a correct answer under accept/reject and top-score selection that accounts for varying task difficulty, blind spots shared across models, and correlated repeated verifier calls. SPLIT fits the formula by empirical Bayes and uses it to plan budgets on unseen tasks. Across seven benchmarks spanning math, science QA, code and software-engineering agents, 26 models, 175 agent systems and six kinds of real verifiers, all from public releases, SPLIT lands within 0.02-0.8 percentage points of the best allocation in hindsight, while every fixed rule we test loses at least 2.5 points in some setting. Regret depends only on how well SPLIT ranks configurations, and it stays near-optimal under compute, price and latency costs and online. The formula also explains when asking more LLM judges lowers precision. As a forecaster far beyond the pilot, SPLIT beats power-law fitting when the verifier caps success, and we report where it loses. [COMMENTS]27 pages (18 main + appendix), 14 figures, 11 tables. Code will be released [LINK]http://arxiv.org/abs/2610.23255v1 [DATE]2026-10-16 18:28:07+08:00 [CATEGORIES]cs.LGcs.CL
Ignored, Not Refused: What Low Prompt-Injection Success Rates Hide in Small Tool-Using Agents [AUTHORS]Farhan Zarif, Kazi Shahid Ahmed Galib, Synthia Akter Promi, Shabnam Sharmin Tarin, Ariyan Hossain, Sadif Ahmed, Farig Yousuf Sadeque [ABSTRACT]Tool-using agents read untrusted external content, which exposes them to indirect prompt injection (IPI), where hidden instructions hijack the agent's tool calls. Benchmarks measure IPI security by the attack success rate (ASR), which treats every failed attack as a successful defense. To test this assumption, we audit 33,000 trajectories from eleven open-weight agents (2-14B parameters) on InjecAgent and AgentDojo. A label-blind LLM judge shows that most failed attacks are ignored (76-80%) rather than refused (17-18%). With identical model weights, changing only the system prompt shifts InjecAgent ASR from 0.2% to 43%. Our Attack-Task Outcome Matrix (ATOM) further shows that in 21.7% of AgentDojo episodes, the agent stayed safe but did not complete the user's task. A low ASR therefore mostly reflects agents silently ignoring injections, not refusing them. Evaluations should report how attacks fail and whether the user's task is completed, alongside ASR. We release the labeled corpus, judge prompt, and analysis code. [COMMENTS]17 pages, 3 figures, 11 tables [LINK]http://arxiv.org/abs/2610.23254v1 [DATE]2026-10-16 18:26:23+08:00 [CATEGORIES]cs.CL
Who Pays for Efficient Inference? Confidence-Based Small-to-Large LLM Routing Across Tasks and Languages [AUTHORS]Md Sabbir Hossen, Anichur Rahman, Pabon Shaha, Airin Afroj Aishi, Andrew H. Sung, Md Shohel Rana [ABSTRACT]Small-to-large language model routing can reduce inference cost by escalating difficult inputs while retaining easier ones on a smaller model. Its success, however, depends on identifying examples that benefit from escalation, and the reliability of confidence across multilingual tasks remains unclear. We present a systematic evaluation of confidence-based routing across three model families, four languages, and three task types, covering 72 model, dataset, language, and scale configurations and 42,288 predictions. We compare global confidence routing with random routing, equal-quota confidence routing, and an oracle upper bound. Confidence identifies beneficial escalations on the Belebele reading comprehension and SIB-200 topic classification datasets, with mean correction-detection AUCs of 0.612 and 0.705. At 50% escalation, it improves over random routing by 1.40--4.67 percentage points on Belebele and 2.54--5.63 points on SIB-200. In contrast, the mean AUC on the MMLU-ProX-Lite knowledge and reasoning dataset is only 0.510, and routing provides no consistent improvement over random selection because small models often remain highly confident while incorrect. Global confidence routing also creates language escalation gaps up to 30.33 percentage points, whereas equal-quota routing removes this disparity with little aggregate accuracy cost. These results show that confidence routing is task-dependent and can allocate computation unevenly across languages. Code is available at https://github.com/sabbir5622r/multilingual-llm-routing [LINK]http://arxiv.org/abs/2610.23249v1 [DATE]2026-10-16 18:23:41+08:00 [CATEGORIES]cs.CL
MGE-PO: Entropy-Aware Credit Assignment for Mathematical Reasoning [AUTHORS]Jihao Huang, Guangjia Chai, Zhirui Chen, Lei Wang, Yao Liu, Liu Qi, Pengjie Ding, Xinyang Liu, Chengxun Jia, Ling Shao, Hongyu Chen, Junchen Wan [ABSTRACT]GRPO assigns the same group-relative advantage to equally rewarded responses within a question, leaving their generation uncertainty unused. We introduce MGE-PO, centered on four-quadrant (4Q) advantage shaping that combines verified correctness with rollout generation entropy. Correctness determines the sign of an additional correction, while entropy controls its magnitude: lower-entropy successes receive larger bonuses and higher-entropy failures receive larger penalties. An online entropy-based question curriculum supports this credit rule using cached rollout measurements. At the final checkpoint after eight training epochs on Qwen3-8B-Base, adding 4Q across three policy optimization settings yields absolute gains of 3.23-3.69% in six-benchmark macro avg@32 and 4.24-11.06% in macro pass@32. On Qwen2.5-7B-Base, the complete configuration raises four-benchmark macro avg@32 from 41.5% to 45.4%. [LINK]http://arxiv.org/abs/2610.23245v1 [DATE]2026-10-16 18:15:02+08:00 [CATEGORIES]cs.CL
STAIF: A Stage-wise Optimization for Complex Instruction Following [AUTHORS]Jian Hong, Chen Cheng, Quan Liu, Yuhao Chen, Enhong Chen [ABSTRACT]Following complex instructions with multiple explicit constraints remains a fundamental challenge for large language models (LLMs). Existing alignment methods, such as DPO, optimize holistic reward signals that often underemphasize strict satisfaction of individual constraints, particularly under out-of-distribution or multi-constraint settings. In this paper, we propose STAIF, a stage-wise optimization framework that decouples the alignment of subjective (soft) constraints from the optimization of objectively verifiable (hard) constraints. Stage 1 applies preference optimization with multiple negative samples to sharpen sensitivity to soft constraints, while Stage 2 applies Reinforcement Learning with Verifiable Rewards (RLVR) to enforce strict compliance with hard constraints. To support this method, we construct STAINSTRUCT, a high-quality bilingual (English, Chinese) dataset of approximately 31,000 complex multi-constraint instructions. Extensive analyses validate the design of STAIF and show state-of-the-art performance on representative benchmarks against strong baselines, as well as genuine generalization. [COMMENTS]16 pages, 6 figures [LINK]http://arxiv.org/abs/2607.22649v2 [DATE]2026-10-16 18:00:07+08:00 [CATEGORIES]cs.CL
ReLaDrift: Recurrent Language Generation via Drifting for Test-Time Scaling [AUTHORS]Yuki Otsuka, Daisuke Oba, Naoaki Okazaki [ABSTRACT]Drifting Models generate a sample with a single function evaluation, making them attractive for fast parallel language generation, but their one-step formulation does not support iterative refinement. We propose Recurrent Language Generation via Drifting (ReLaDrift), which uses a shared Drifting Model to generate an initial draft from noise and iteratively refine it in a continuous latent space. We train this recurrence by sampling the number of rounds and applying the drifting loss only at the final round, stopping gradients through preceding rounds so that only one round requires backpropagation. Unlike the one-step Drifting Model, whose accuracy does not improve when iterated, ReLaDrift enables a trade-off between test-time computation and generation quality, as diffusion and flow language models do. On Sudoku Hard and GSM8K, it achieves competitive or superior performance compared with these iterative-generation approaches across a range of inference budgets. Our code is available at https://github.com/Yuki-Ots/ReLaDrift [LINK]http://arxiv.org/abs/2610.23225v1 [DATE]2026-10-16 17:55:22+08:00 [CATEGORIES]cs.CL
MAS-Bai: Linguistically Informed Adaptive Multi-Agent Translation for Bowen-Script Bai [AUTHORS]Jingting Zheng, Hong Ren, Yuehan Cui, Shaolin Zhu, Deyi Xiong [ABSTRACT]Large Language Models (LLMs) remain unreliable for extremely low-resource languages whose writing systems conflict with high-resource pretraining priors. We study Bai-Chinese translation in the Bowen script, where Chinese characters are frequently borrowed phonetically rather than semantically, causing severe literal-translation hallucinations. We propose MAS-Bai, a linguistically informed adaptive multi-agent framework for Bowen-script Bai-Chinese translation. MAS-Bai routes each input through either a lightweight retrieval-translation path or a deeper philological path with etymological and syntacticagents, and uses cycle-consistency with a lightweight reverse translator as an inference-time reliability signal. We construct a 6,536-sentence Bai-Chinese corpus and evaluate on a 1,536-sentence test set. MAS-Bai achieves 16.23 BLEU and 62.15 COMET, improving over Standard RAG by 8.11 BLEU and 6.85 COMET. Forward LoRA, expert ablations and controlled routing experiments distinguish the roles of supervised adaptation, Bai-specific expertise and conditional routing. The largest gains occur on Bowen phonetic loans and divergent Bai word order, while human evaluation and reverse-verifier analyses support the automatic results. [COMMENTS]14 pages, 3 figures, Findings of EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23205v1 [DATE]2026-10-16 17:43:32+08:00 [CATEGORIES]cs.CL
Brain-to-Language Decoding: Paradigms, Signals, Methods, Evaluation, Practical Use and Beyond [AUTHORS]Yiqian Yang, Yiqun Duan, Chenyu Liu, Yiqi Wang, Xinliang Zhou, Chin-Teng Lin, Yu Zhang [ABSTRACT]Brain-to-language decoding translates neural activity associated with language production, internal speech and perception into linguistic or expressive outputs. It offers a route to restoring communication after speech loss and a means of studying how the brain represents language. Advances in neural recording and representation learning have expanded the field from constrained recognition and acoustic reconstruction to text generation, streaming personalised speech and facial animation. This survey synthesises these developments across invasive and non-invasive measurements, drawing on a search without a lower year limit and source-led updates through September 2026. We connect Articulated, Inner and Perceived tasks to the neural populations they engage, the representations available to decoders and the outputs those representations can support. We examine model development, public resources and the evolution of evaluation, and compare published performance and communication costs within their reported protocols. The synthesis identifies complementary routes to progress: phonetic, acoustic and semantic targets preserve different aspects of a message; shared representations support reuse across recording conditions and tasks; and online communication increasingly depends on calibration, feedback and user control alongside decoding accuracy. Shared benchmarks enable algorithmic comparisons, while longitudinal studies reveal the demands of sustained use. We discuss these developments and their remaining limitations, then outline a prospective five-level trajectory from commands and language to meaning, scenarios and bidirectional cognitive exchange [LINK]http://arxiv.org/abs/2609.27650v4 [DATE]2026-10-16 17:37:51+08:00 [CATEGORIES]cs.CL
BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models [AUTHORS]Liubov Chubarova, Alexandra Kuleshova, Daniil Volkov, Kirill Sultanov, Alexey Zaytsev [ABSTRACT]While Multimodal Large Language Models (MLLMs) have made significant strides in visual comprehension, their ability to reason about text-dense, professional documents remains incompletely evaluated. Existing benchmarks emphasize information extraction, require external domain knowledge, or cover professional documents only as one of many settings. They are also largely English- or Chinese-centric, leaving other languages and Russian, in particular, substantially underrepresented. To address these limitations, we introduce BEAR-Bench (Bilingual Enterprise and Academic Reasoning), a self-contained, complex English-and-Russian benchmark comprising 1000 human-annotated questions based on text-rich business and scientific documents. We evaluate 16 proprietary and open-weight MLLMs, including Gemini 3.1 Pro and Qwen3.5-397B, on BEAR-Bench and observe clear headroom even for the strongest systems. Finally, we use the resulting model outputs to compare existing hallucination detection methods, evaluating not only how often models fail on BEAR-Bench but also how reliably those failures can be identified. [LINK]http://arxiv.org/abs/2608.17895v2 [DATE]2026-10-16 17:30:41+08:00 [CATEGORIES]cs.CL
AMEND: Constraint-Guided Multi-Agent Text Detoxification with Isolated Critics [AUTHORS]Tianyu Wang [ABSTRACT]Text detoxification requires balancing toxicity removal, semantic preservation, and fluency. Existing methods typically optimize an unconstrained scalar objective---enabling safety scores to mask semantic loss---or rely on open-ended iterative debate prone to semantic drift and high computational cost. We propose AMEND, an inference-time framework that decouples detoxification into role-conditioned proposal generation, isolated multidimensional criticism, and constraint-first selection. AMEND evaluates diverse candidates with role-blind critics and enforces strict semantic and fluency feasibility before ranking by utility. To maximize efficiency, AMEND Lite restricts subsequent search to at most one targeted repair round only when residual toxic spans remain. On ParaDetox, one-round selection over diverse roles improves the joint evaluation metric by $+0.0508$ over the best single role. On a frozen confirmation test set, AMEND Lite achieves a joint score of $0.6597 \pm 0.0062$, outperforming unconstrained multi-round loops by $+0.0489$ ($p < 0.05$) while reducing inference requests by 28.4%. [COMMENTS]10 pages, 2 figures, 4 tables [LINK]http://arxiv.org/abs/2610.23183v1 [DATE]2026-10-16 17:22:21+08:00 [CATEGORIES]cs.CL
What LLMs Do in the Shadows: Reading Context Reliance from the Unembedding Dark Space [AUTHORS]Alexandra Kuleshova, Andrey Volodichev, Daria Kotova, Alexey Zaytsev [ABSTRACT]Detecting faithfulness hallucinations in large language models (LLMs) is crucial for their safe deployment in high-stakes domains. Recent approaches address this by contrasting contextual against parametric knowledge utilization, yet measuring context reliance typically demands noticeable overhead, such as materializing attention maps, running additional forward passes, or training auxiliary probes. In this work, we show that context reliance can be estimated by projecting hidden states onto the dark space - directions that compose weakly with the unembedding matrix. Beyond its recognized function as an initial-token attention sink, we show that this subspace actively mediates context processing, and that its share of hidden-state norm - termed the Dark Matter Ratio (DMR) - provides a direct score of context utilization. Pairing DMR with a parametric knowledge score yields DarkLens, a lightweight faithfulness detector operating strictly within a single forward pass. Evaluated across three QA and summarization benchmarks on six open-weight LLMs (4B-13B parameters), DarkLens matches or outperforms a wide range of baselines without requiring extra forward passes or probe training. [LINK]http://arxiv.org/abs/2610.23182v1 [DATE]2026-10-16 17:21:49+08:00 [CATEGORIES]cs.CL
Tree-Aware Posterior Distillation for Long-Horizon Agentic Reinforcement Learning [AUTHORS]Heming Zou, Qi Wang, Yixiu Mao, Yun Qu, Yuyi Guo, Zeyuan Liu, Chenglin Li, Weijie Liu, Saiyong Yang, Xiangyang Ji [ABSTRACT]Reinforcement learning with verifiable rewards has become a scalable post-training paradigm for large language models. However, training agents over long-horizon interactions presents two challenges: using only terminal rewards makes turn-level credit assignment sparse and high-variance, while sequential mini-batch updates on a fixed rollout batch introduce policy staleness as the current policy drifts from the behavior policy. Existing approaches obtain denser credit via learned critics, process reward models, or on-policy distillation, but they rely on a value network, step-level annotations, or a stronger external teacher that increase training complexity. Meanwhile, importance sampling and clipping mitigate policy drift, but they do not specify a finite target that stops sequential updates from overshooting. In this work, we introduce TreeOPSD, a critic-free, tree-aware posterior self-distillation framework for long-horizon agentic reinforcement learning. Treating each interaction turn as a tree node, TreeOPSD branches alternative continuations from shared prefixes and recursively backs up terminal rewards into implicit turn-level credit. It then maps each credit to a closed-form, bounded teacher distribution parameterized by the likelihood ratio relative to the frozen behavior policy. Minimizing the divergence between this teacher and the student policy recovers the policy-gradient direction as the tangent at the behavior policy while bounding updates toward a finite target across sequential mini-batches. Empirically, TreeOPSD improves agentic capabilities across several competitive benchmarks and model scales, especially for long-horizon tasks, e.g., raising the average accuracy of Qwen3-30B-A3B-Thinking-2507 by 4.97 points over GRPO on the Deep Research setting. [LINK]http://arxiv.org/abs/2610.23181v1 [DATE]2026-10-16 17:17:26+08:00 [CATEGORIES]cs.LGcs.CL
Do Speech BCIs Need Larger Models? Rethinking Neural Decoding beyond Scaling [AUTHORS]Zehui Feng, Weichuan Wang, Xiaohan Chen, Cuntai Guan, Ting Han [ABSTRACT]Recent speech brain-computer interfaces (BCIs) increasingly rely on large models and complex multi-stage pipelines for neural decoding. However, in speech BCI, where training data are scarce, noisy, and expensive to collect, the benefits of model scaling remain unclear. This work revisits a fundamental question: are large models truly necessary for accurate neural decoding? We propose BEST (Brain-Encoding-Speech-Text), a two-stage framework for efficient neural-to-text decoding. Instead of relying on large generative models or phoneme-centric pipelines, BEST introduces hierarchical intermediate representations and a compact speech-oriented decoder that directly maps neural activity to text. This design preserves essential structure while significantly reducing computational cost, enabling robust generalization in low-resource settings and practical deployment under constrained resources. On the Brain-to-Text '24 and '25 benchmarks, BEST achieves word error rates (WERs) of 6.19% and 1.62%, respectively, using only 3.85% of the parameters required by comparable methods. Beyond these results, we provide three key insights: (1) decoder scaling yields non-monotonic gains; (2) cross-modal alignment is more critical than model size; and (3) overfitting degards sequence-level performance. Representation analysis further reveals a consistent neural-audio-text alignment pathway. [COMMENTS]25 pages, Accepted by NeurIPS 2026 as a Main Track conference paper [LINK]http://arxiv.org/abs/2610.23175v1 [DATE]2026-10-16 17:12:58+08:00 [CATEGORIES]cs.CL
Chinese Decision: A Family of Language Models for Single-Pass Probabilistic Decision-Making [AUTHORS]Qiang Yan [ABSTRACT]Many language-based applications require probabilities over a finite set of decisions rather than open-ended responses. We present Chinese Decision, a family of Chinese-language models for single-pass probabilistic decision-making. Given context, an instruction, and candidate descriptions, the model extracts final-position candidate-letter logits in one forward pass and normalizes them over valid options, without autoregressive answer generation. A shared interface supports Boolean judgments, categorical choices, and ordinal scores. We report a completed 0.8B model trained with distribution cross-entropy on 13.048 million decisions and preliminary 2B checkpoints trained with an additional RLCD objective. On the full CJ-Bench of 307,900 decisions, the uncalibrated 0.8B model achieves 72.24% general-domain accuracy and 76.98% average accuracy across medicine, law, and finance using a single checkpoint, with 1.07% overall expected calibration error. The 2B model reaches 72.56% general-domain accuracy after 40.13% of one epoch but has not surpassed the completed 0.8B model on the professional-domain average. We further evaluate a validation-temperature-scaled deployment model and its Q8 quantization, finding close aggregate performance while documenting individual prediction changes. We release the training, evaluation, and deployment pipeline. The current results establish a reproducible approach to Chinese probabilistic decisions but do not isolate the effect of model scale. [COMMENTS]11 pages, 2 figures, 11 tables. Code: https://github.com/yanqiangmiffy/chinese-decision [LINK]http://arxiv.org/abs/2610.23170v1 [DATE]2026-10-16 17:08:32+08:00 [CATEGORIES]cs.CL
Explaining Chess and Go in Natural Language [AUTHORS]Ori Press, Alexander Golubev [ABSTRACT]Modern chess engines play far beyond human ability, often choosing moves whose underlying ideas are difficult for humans to understand. We propose an automatic metric for evaluating chess explanations, grounded in chess-engine evaluations. Evaluating state-of-the-art language models using this metric, we find that they perform surprisingly poorly, often relying on surface-level board observations rather than robust, transferable chess concepts. We then introduce BinProp, a training-free method that enables these same models to discover reusable chess concepts and use them to explain moves in natural language. BinProp improves explainability for every model tested, nearly tripling the average score compared with direct prompting. Beyond chess, the same method discovers useful concepts in Go, draughts, and shogi without additional tuning. [LINK]http://arxiv.org/abs/2610.23166v1 [DATE]2026-10-16 17:05:17+08:00 [CATEGORIES]cs.CLcs.LG
EKLAV: Elicited Key CoT Learning via Answer only superVision [AUTHORS]Adarsh Singh, Aman Singh, Kushal Raj Bhandari, Soham Dan, Anoop Kumar, Vivek Gupta [ABSTRACT]Reasoning distillation trains a student to reproduce both the teacher's chain-of-thought trace and its final answer. We ask whether the trace is better used as privileged training context than as an imitation target. EKLAV keeps the teacher's reasoning in the input sequence but supervises only the final task answer, applying a task-dependent placement that moves part or all of the trace into the prompt. On BRIGHT passage reranking, EKLAV improves average nDCG@10 at all four tested checkpoints across two model families, winning 41 of 48 model-domain comparisons while reducing training compute by up to 31.6% and inference time by up to 32.7%. On four held-out table-reranking benchmarks, it is the best of four conditions on three, exceeding full-trace distillation by 3.0 points and answer-only training without reasoning by 5.5 points on average. Analysis of generated traces reveals that EKLAV students defer their first verbalized verdict later (0.76 vs. 0.44 relative position), reverse fewer decisions (5% vs. 16%), and produce reasoning more predictive of their own rankings than STD-COT's, though less predictive than the base model's. The recipe reuses a fixed teacher corpus, requires no student rollouts or iterative teacher feedback, and applies to both pointwise and listwise formulations. Code and Data are at https://github.com/AdarshSingh7647/Eklavya [COMMENTS]8 pages, 3 figures [LINK]http://arxiv.org/abs/2607.25182v2 [DATE]2026-10-16 17:04:49+08:00 [CATEGORIES]cs.CL
TriRouter: Sparse-to-Dense Electrode-Space Routing for Robust EEG Motor Imagery Decoding [AUTHORS]Ying Zhou, Yaoxin, Mengting Liu [ABSTRACT]Motor imagery brain-computer interfaces require reliable decoding from multichannel EEG, but practical acquisition often yields sparse or corrupted electrode montages. We propose TriRouter, an electrode-space routing framework with discriminative-preservation, mask-robustness, and latent-compensation paths. A validation-calibrated router uses observable channel availability to allocate path contributions, preserving complete-channel discrimination while increasing representation recovery under severe missingness. In BNCI2014-001, TriRouter outperforms standard public backbones and a dynamic-spatial-filtering corrupted-EEG baseline. It improves accuracy over the strongest corrupted-EEG baseline by 3.68, 4.02, and 6.46 percentage points at nominal 25%, 50%, and 75% missing settings, and remains effective under structured regional missingness and additional low-channel dataset evaluation. [COMMENTS]Submitted to IEEE ICASSP 2027 [LINK]http://arxiv.org/abs/2610.23665v1 [DATE]2026-10-16 23:59:00+08:00 [CATEGORIES]cs.LG
JAXtari: High-Throughput and Easy-to-Modify Arcade Learning Environment [AUTHORS]Quentin Delfosse, Raban Emunds, Paul Seitz, Sebastian Wette, Jannis Blüml, Daniel Kirn, Dominik Mandok, Kristian Kersting [ABSTRACT]The Arcade Learning Environment (ALE) has been a cornerstone of Deep Reinforcement Learning research, providing a standardized benchmark for evaluating general agents. However, its lack of native support for object-extraction and environment modification restricts research into evaluating whether agents learn robust causal models or rely on brittle visual heuristics. Furthermore, the ALE's dependence on CPU-based emulation creates a significant bottleneck in modern GPU-accelerated training pipelines, limiting both experimental scale and iteration frequency. In this work, we introduce JAXtari, an open-source, high-performance reimplementation of Atari 2600 environments. By leveraging massive GPU parallelization, JAXtari allows PPO agents to finish training on 100M environment steps on a single A100-GPU in under an hour. Beyond speed, it provides object-centric observations and a flexible modification system, allowing researchers to, e.g. develop neuro-symbolic methods or discover novel shortcuts and subsequently train and adapt their agents rapidly, sometimes in only minutes. [LINK]http://arxiv.org/abs/2610.23650v1 [DATE]2026-10-16 23:48:00+08:00 [CATEGORIES]cs.LG
Scalable long-range machine learning interatomic potentials with electrical response for atomistic material modeling [AUTHORS]Yicheng Chen, Xiaoyun Yu, Hyejeong Song, Xintian Wang, Zihui Li, Zixi Xia, Junmin Chen, Peichen Zhong [ABSTRACT]Accurate atomistic simulations of materials under electric fields require long-range electrostatics and electrical-response modeling at length scales beyond the reach of conventional first-principles methods. Yet these capabilities remain difficult to reconcile with the scalability of machine-learning interatomic potentials (MLIPs). Strictly local models enable efficient spatial parallelization but neglect long-range interactions, whereas existing electrostatic extensions introduce global dependencies that hinder distributed computation. Here, we overcome this limitation by analytically separating the long-range electrostatic forces and Born effective charges (BECs) into a global electrostatics calculation and strictly local derivatives of learned latent charges. This formulation preserves the spatial decomposability of equivariant MLIPs while enabling distributed evaluation of energies, forces, stresses, and BECs across different boundary conditions. Implemented as DPA4C-LR on the \texttt\{nvalchemi\} platform, the framework scales to 5.2 million atoms on four GPUs at approximately twice the computational cost of its short-range backbone. Trained on energies and forces, DPA4C-LR accurately infers BECs and reproduces the field-dependent infrared spectra of water. We further demonstrate its capabilities by recovering the characteristic slim polarization hysteresis of relaxor \ce\{Pb(Mg_\{1/3\}Nb_\{2/3\})O3\}, which is obscured by finite-size effects in smaller simulations; predicting a room-temperature field-driven ionic conductivity of $6.8 \pm 0.7$~mS/cm in amorphous \ce\{Li_\{0.9\}NbO_\{0.9\}Cl_\{4.1\}\}; and modeling interfacial thermal transport across Si/diamond interfaces with slab-corrected electrostatics. By combining long-range physics, electrical-response inference, and distributed computation, this framework extends predictive atomistic modeling toward device-relevant length scales. [LINK]http://arxiv.org/abs/2610.23643v1 [DATE]2026-10-16 23:44:07+08:00 [CATEGORIES]cs.LG
No-Regret Mixing of LRU and LFU with Optimal Switching Cost [AUTHORS]Younes Ben Mazziane, Xinying Zou [ABSTRACT]Caching systems often rely on simple eviction policies such as Least Recently Used (LRU) and Least Frequently Used (LFU), which perform well in complementary request regimes. Recent policies such as LeCar and Cacheus combine LRU and LFU using ideas from the experts problem in online learning. Specifically, upon a miss, they randomize between the two eviction rules using probabilities derived from scores updated by tracking the history of past evictions. While these policies exhibit strong empirical performance, it remains unclear whether they are guaranteed, on every request sequence, to perform asymptotically as well as the better of LRU and LFU, i.e., whether they achieve sublinear regret with respect to this benchmark. We first show that LeCar suffers linear regret against an oblivious adversary, even with unbounded history. We then propose H-MC, a Hedge-based mixture of virtual LRU and LFU caches that preserves Hedge's selection probabilities, and hence its regret guarantees, while minimizing the switching cost among all joint selection rules with these marginals. [LINK]http://arxiv.org/abs/2609.07566v3 [DATE]2026-10-16 23:42:39+08:00 [CATEGORIES]cs.LG
GRAFT-EEG: Gauge-Aware Physical Field Transport for Cross-Dataset Motor-Imagery EEG [AUTHORS]Ying Zhou, Mengting Liu [ABSTRACT]Cross-dataset EEG decoding often treats montage, reference, and active-sensor changes as representation shift, although they also alter the observation operator. We study target-statistics-blind transfer: fitting receives no target signal, label, or empirical statistic, and electrode coordinates only instantiate a fixed prediction operator. method removes the active-set common-reference gauge and transports each native mask through a template motor leadfield. A source-fitted spectral classifier anchors the decision; a bounded state-space-model (SSM) residual is secondary. We establish restricted identifiability and an operator-perturbation bound. Across five leave-one-dataset-out targets (253 participants), dynamic method achieved 60.865% equal-target balanced accuracy versus 52.684% for strict-clean 5/5 BIOT (hierarchical difference +8.182 points; 95% CI +6.382 to +10.196; positive in every target). On four EEGPT-clean targets, the descriptive means were 61.520% and 59.057%. In four post-hoc exact same-parent tests, FULL exceeded coordinate-basis, matched-spectrum, row-shuffled, and wrong-anatomy controls by 9.959, 9.698, 11.241, and 3.985 points, respectively (all 95% CIs excluded zero; Holm-adjusted P<0.001). A separate 60-participant same-paradigm evaluation retained a positive physical-operator contrast (+4.029 points; +2.338 to +5.773), whereas dynamic--static was inconclusive; its preregistered status is partial confirmation rather than unrestricted external validation. The SSM added +1.059 points (+0.013 to +1.892). Claims remain bounded to five binary motor-imagery domains and the tested operators. [COMMENTS]Under review at ICLR 2027 [LINK]http://arxiv.org/abs/2610.23620v1 [DATE]2026-10-16 23:30:37+08:00 [CATEGORIES]cs.LG
ThetreStageAI: Multi-View Retrieval-Augmented Generation for Structured Bengali Theatre Production Planning [AUTHORS]Md. Mehedi Hasan Naeem, Mst. Kamrunnahar Ruma, Shakila Sultana, Nafiza Anjum, Nasreen Jannatun Nemu, Md. Sujan Ali [ABSTRACT]Staging a theatre performance requires coordinating dialogue, actor movement, stage directions, lighting focus, cue timing, and equipment constraints within the same production plan. This work presents ThetreStageAI, a retrieval augmented framework for structured Bengali theatre production planning. Rather than retrieving from one mixed corpus, the system maintains separate scene, actor-blocking, and lighting views. Bengali text is embedded with BGE-M3 and indexed in view-specific Qdrant collections. Retrieved references are combined with a production brief and passed to Gemini, which generates schema-constrained JSON containing characters, scenes, dialogue, stage directions, blocking, lighting, and optional sound cues. Structural and semantic checks are applied before acceptance, with at most one constrained repair attempt. We also construct a human-guided, LLM-assisted synthetic dataset of 500 short Bengali theatre records that expands to 1,500 retrieval documents. Dataset diagnostics confirm consistent identifiers and valid lighting fields while also revealing substantial dialogue repetition. A paired camera-ready evaluation compares single-combined and full multi-view retrieval under matched settings. The source code, dataset, and evaluation materials are publicly available. [COMMENTS]6 pages, 6 figures. Accepted at the 12th IEEE International Conference on Sustainable Technologies and Engineering (IEEE iCOSTE 2026), Cox's Bazar, Bangladesh, November 1-3, 2026 [LINK]http://arxiv.org/abs/2610.23618v1 [DATE]2026-10-16 23:29:20+08:00 [CATEGORIES]cs.LG
SelectArena: A Modern Benchmark for Feature Selection on Real-world Tabular Data [AUTHORS]Bastian Schäfer, Dominika Matus, Lennart Purucker, Ricardo Knauer, Erik Rodner, Frank Hutter, Anton Sokolov, Arnaud Liefooghe [ABSTRACT]Selecting a subset of features (columns) from a tabular dataset is a ubiquitous challenge in real-world applications. Industry experts seek to minimize model training costs, climate scientists require an interpretable subset of factors to understand extreme weather events, and clinicians need to choose among diagnostic tests due to financial and time constraints. When looking at the scientific literature to determine which feature selection method to use, practitioners across domains face an unrepresentative landscape of benchmarking. Despite extensive research, existing feature selection benchmarks provide limited guidance because they rely on outdated models, ad hoc evaluation protocols, or artificially generated data that do not reflect real-world problems. To close the gap between scientific benchmarking and real-world applications, we introduce SelectArena, a modern benchmark for feature selection. We compare 15 widely used feature selection methods on 71 manually curated real-world tabular datasets, covering classification and regression problems, using standardized evaluation protocols and state-of-the-art tabular (foundation) models. Our results reveal clear trade-offs across predictive performance, runtime, stability, and validity. Tree-based methods, CART and RFImportance, along with F-Test are Pareto-optimal when taking into account predictive performance and runtime. F-Test, information-theory-, and distance-based methods lead on stability and validity, with F-Test emerging as a balanced choice across our metrics. Search-based methods are slow, unstable and achieve low performance. SelectArena turns these findings into practical guidance and, as an extensible platform, supports future feature selection research. [LINK]http://arxiv.org/abs/2610.13644v2 [DATE]2026-10-16 22:56:25+08:00 [CATEGORIES]cs.LG
Beyond Expected Values: Risk-Sensitive Planning with Distributional Monte-Carlo Tree Search [AUTHORS]Quynh Trinh, Trung Nguyen, Nam Nguyen, Tuan Dam [ABSTRACT]Monte-Carlo Tree Search (MCTS) typically stores only scalar mean estimates, which makes it poorly suited for tail-risk objectives such as conditional value-at-risk (CVaR). We introduce Categorical Thompson Sampling with Optimism (CATSO) and Particle Thompson Sampling with Optimism (PATSO), two distributional MCTS algorithms that maintain finite-support empirical backup laws at Q-edges. CATSO represents Q-edge laws with categorical atoms and Dirichlet Thompson sampling, while PATSO represents them with adaptive particles. Both algorithms select actions using the CVaR of a Thompson-sampled Q-edge law plus a polynomial optimism bonus, and propagate scalar continuation values through visit-weighted averages of child Q-edge CVaR scores. For the finite-depth nested CVaR planning target, we prove an $O(n^\{-1/2\})$ bound on the root value error after $n$ rollouts, which upper-bounds the risk-sensitive simple regret. CATSO adds a fixed-grid discretization bias, while capped PATSO adds a tunable Wasserstein compression bias. Experiments on stochastic planning benchmarks show that distributional Q-edges improve lower-tail return and reduce catastrophic outcomes when mean-optimal and risk-sensitive behavior differ. [COMMENTS]Proceedings of the 40th Annual Conference on Neural Information Processing Systems [LINK]http://arxiv.org/abs/2610.23565v1 [DATE]2026-10-16 22:51:28+08:00 [CATEGORIES]cs.LG
Can Graph Learning Learn Circuits? [AUTHORS]Chester Tan, Moritz Lampert, Courtney Maynard, Ankit Ramakrishnan, Tina Eliassi-Rad, Ingo Scholtes [ABSTRACT]Circuit localization is a mechanistic interpretability task whose goal is to identify a sparse subgraph of a transformer's computation graph sufficient to reproduce a particular behavior. Most established methods localize circuits independently for each model-task pair. We instead frame circuit localization as a graph machine learning problem in which the edges of a computation graph represent computational pathways, and graph neural networks (GNNs) model interactions among these pathways. We introduce Graph Circuit Learning (GCL), a supervised, amortized framework that trains a GNN across multiple model-task pairs and applies it to unseen cases. To provide sufficient data, we augment the InterpBench benchmark with additional cases derived from TracrBench programs. Of the 46 evaluated GCL configurations, the one selected by cross-validation scored a median edge AUROC of 0.946 (interquartile interval [0.841, 0.980]) on the 16 held-out cases of our evaluation split. Removing all message-passing edges reduces the median to 0.926. We also adapt PGExplainer, a GNN explainability method, to circuit localization, obtaining a median edge AUROC of 0.950 on the same cases. These preliminary results suggest that graph machine learning offers a natural and potentially powerful perspective on circuit localization, one we hope encourages closer exchange between the two communities. [COMMENTS]Oral presentation at the Fifth Learning on Graphs Conference (LoG 2026), Extended Abstract Track [LINK]http://arxiv.org/abs/2608.08536v2 [DATE]2026-10-16 22:46:47+08:00 [CATEGORIES]cs.LG
BenthicDINO: Physics-Informed Self-Distillation for View-Invariant Side-Scan Sonar Representations [AUTHORS]Taqi Hamoda, Hayat Rajani, Nuno Gracias [ABSTRACT]Automated interpretation of side-scan sonar (SSS) imagery is traditionally hindered by view-dependence, as recorded intensities inextricably conflate intrinsic seabed reflectivity with transient acquisition geometries and acoustic artifacts. To address this, we introduce BenthicDINO, a physics-informed self-supervised learning framework that decouples acquisition dynamics from seabed representations. Utilizing a ConvNeXt-v2 backbone, physics-guided augmentations, and an HSIC penalty, BenthicDINO explicitly enforces feature invariance to physical viewing parameters like slant range and incidence angle. Evaluations demonstrate exceptional data efficiency - achieving near-peak performance in benthic habitat mapping with only 10% of available labels - and strong cross-domain generalization, outperforming fully supervised baselines on the AI4Shipwrecks benchmark using frozen representations. [LINK]http://arxiv.org/abs/2608.23215v2 [DATE]2026-10-16 22:33:36+08:00 [CATEGORIES]cs.LG
Interpreting SSL Representations for Spoof Detection: a WavLM Study [AUTHORS]Mohamed Mallat, Michele Panariello, Massimiliano Todisco, Nicholas Evans, Anthony Larcher [ABSTRACT]Self-supervised speech models such as WavLM achieve strong spoofing detection performance, yet which layers encode the most spoof-relevant information and what acoustic properties they capture remain opaque. We present a layer-wise interpretability study of WavLM using the ASVspoof 5 database, combining three complementary analyses: (i) per-layer linear probing with attack-level evaluation across speech synthesis, voice conversion, and adversarial attacks; (ii) a canonical correlation analysis (CCA) linking layer representations to spoofing-relevant acoustic features; (iii) an exhaustive grid search over layer subsets showing that use of five layers outperforms full thirteen-layer pooling. CCA analysis reveals that, while most acoustic features peak in alignment with mid-layers and then decay, voice quality measures jitter, shimmer, and the harmonics-to-noise ratio show the opposite trend, with correlation increasing toward later layers. Embedding visualisations reveal that deeper layers capture a continuous acoustic similarity among attacks rather than discrete categories, suggesting these layers encode fundamental synthesis artifacts rather than attack-specific signatures. [COMMENTS]5 pages, 4 figures, 1 table. Published at IWBF 2026. An extended abstract of this work was presented as a poster at Odyssey 2026 [LINK]http://arxiv.org/abs/2610.23526v1 [DATE]2026-10-16 22:26:55+08:00 [CATEGORIES]cs.LG
Adaptive conformal prediction on the Grassmannian [AUTHORS]Dimitrios Loukrezis, Zachary J. Grey, Dimitris G. Giovanis [ABSTRACT]We introduce a conformal prediction framework for regression with Grassmann-valued responses. The proposed method constructs prediction regions in the form of metric balls directly on the Grassmann manifold, using basis-invariant geometric distances as non-conformity measures and scaling the corresponding non-conformity scores through a learned prediction difficulty estimator. This yields locally adaptive, input-dependent prediction regions that respect the underlying manifold geometry, while retaining the finite-sample marginal coverage guarantee of split conformal prediction under exchangeability. We evaluate several Grassmannian distance metrics and compare their prediction-region efficiency on synthetic benchmarks and an airfoil inverse shape-design problem. The results highlight the advantages of geometry-aware non-conformity scores, while local adaptation improves empirical coverage uniformity across prediction difficulty levels and, in strongly heteroscedastic settings, can substantially reduce prediction-region size while maintaining the target coverage. [COMMENTS]21 pages, 6 figures, 3 tables [LINK]http://arxiv.org/abs/2610.23508v1 [DATE]2026-10-16 22:18:06+08:00 [CATEGORIES]cs.LG
Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification [AUTHORS]Brown Zaz, Mar Gonzàlez I Català, Ferran Hernandez Caralt, Moshe Eliasof, Pietro Liò [ABSTRACT]In the transductive setting, where the full graph is observed but node labels are only partially available, progress in semi-supervised node classification has largely focused on architectural innovation. In this paper, we revisit an orthogonal axis: the training objective. We start from a simple observation: transductive models produce predictions for every node during training, including nodes without labels. These unlabeled-node predictions may contain useful training signal, but standard supervised objectives discard them because no ground-truth labels are available. Inspired by the decomposition of cross-entropy into a label-dependent alignment term and a label-independent entropy term, we propose prediction confidence as a natural way to extract this signal in the absence of labels. This motivates Transductive Sharpening (TS): a loss-level modification that minimizes prediction entropy on unlabeled nodes while counterbalancing this effect on labeled nodes. We evaluate Transductive Sharpening across a wide range of node-classification benchmarks and observe consistent performance improvements without requiring any changes to the backbone architecture. Code is available at https://github.com/transductive-sharpening/tunedGNN. [COMMENTS]19 pages, 3 figures, 20 tables [LINK]http://arxiv.org/abs/2605.20248v2 [DATE]2026-10-16 22:15:12+08:00 [CATEGORIES]cs.LG
Cross-Dialect Generalization Without Retraining: Benchmarks and Evaluation of Schema-Derived Constrained Decoding for MLIR [AUTHORS]Plawan Kumar Rath [ABSTRACT]Multi-Level Intermediate Representation (MLIR) underlies modern ML compiler infrastructure (TensorFlow, JAX/StableHLO, PyTorch Inductor, IREE), yet appears only in trace amounts in code-LM pretraining corpora. MLIR is also extensible by design: new dialects ship per application domain, so a fine-tuned model per dialect does not scale. We ask whether inference-time priors derived mechanically from each dialect's Operation Definition Specification (ODS) can substitute for gradient-based adaptation. First, we release four natural-language-to-MLIR benchmarks across three dialects (MLIR-Spec-150, Linalg-Spec-30, StableHLO-Spec-30, StableHLO-Held-Out-200), totaling 410 in-scope NL-to-MLIR pairs, plus a 25-program out-of-grammar stress set, with functional references on 30 Spec prompts, under Apache-2.0 with Gebru datasheets and Croissant 1.1 metadata. Second, we build a three-layer schema-derived constraint stack: a CFG over op signatures (C1), type-domain splits from an ODS-extracted type lattice (C2), and an SSA-scope validator driving five-retry rejection sampling (C3). Porting it to StableHLO required no new constraint-layer code. On structurally dominated dialects, SmolLM2-1.7B with token-level constraints matches or exceeds 15B-34B code LMs running post-hoc rejection at a matched retry budget and 8-25x the per-generation speed: on linalg it reaches 80.0% verify-valid (three runs, n=125), 21 to 44 pp above CodeLlama-34B, Granite-Code-34B, and StarCoder2-15B with non-overlapping CIs. The comparison is between systems, not models: under the baselines' post-hoc protocol, SmolLM2 falls from 52.0% to 24.5% on arith+func. On arith+func and the templated StableHLO-Held-Out-200 the baselines match or beat the SLM; we scope these as non-win cells. We release benchmarks, decoder, every per-prompt generation, and a pinned reproducibility Dockerfile. [COMMENTS]Accepted to the NeurIPS 2026 Evaluations and Datasets Track; camera-ready version. 31 pages (10 main + references + appendix + checklist). Code: https://github.com/plawanrath/slm-to-mlir-constrained-emitter Datasets: https://huggingface.co/collections/plawanrath/cross-dialect-mlir-benchmarks-neurips-2026-e-and-d [LINK]http://arxiv.org/abs/2607.18254v2 [DATE]2026-10-16 22:07:14+08:00 [CATEGORIES]cs.LG
RamanBench: A Large-Scale Benchmark for Machine Learning on Raman Spectroscopy [AUTHORS]Mario Koddenbrock, Christoph Lange, Robin Legner, Martin Jäger, Martin Kögler, Mariano N. Cruz Bournazou, Peter Neubauer, Felix Biessmann, Erik Rodner [ABSTRACT]Machine learning (ML) has transformed many scientific fields, yet key applications still lack standardized benchmarks. Raman spectroscopy, a widely used technique for non-invasive molecular analysis, is one such field where progress is limited by fragmented datasets, inconsistent evaluation, and models that fail to capture the structure of spectral data. We introduce RamanBench, the first large-scale, fully reproducible benchmark for ML on Raman spectroscopy, consisting of streamlined data access (https://pypi.org/project/raman-data/), evaluation protocols and code (https://pypi.org/project/raman-bench/), as well as a live leaderboard (https://huggingface.co/spaces/HTW-KI-Werkstatt/RamanBench). It unifies 77 datasets (including 17 first released with this benchmark) across four domains, comprising 325,862 spectra and spanning classification and regression tasks under diverse experimental conditions. We benchmark 28 models under a standardized protocol, including classical methods (e.g., PLS), Raman-specific (e.g., RamanNet), tabular foundation models (TFMs; e.g., TabPFN), and time-series approaches (e.g., ROCKET). TFMs consistently outperform domain-specific and gradient boosting baselines, while time-series models remain competitive. However, no single method dominates across these heterogeneous datasets, revealing that Raman spectroscopy remains an open challenge for machine learning. Therefore, we invite the community to contribute new approaches to our living benchmark, with the potential to accelerate advances in critical applications such as medical diagnostics, biological research, and materials science. [COMMENTS]https://huggingface.co/spaces/HTW-KI-Werkstatt/RamanBench [LINK]http://arxiv.org/abs/2605.02003v4 [DATE]2026-10-16 21:36:03+08:00 [CATEGORIES]cs.LG
Predictable Scaling: Efficient Model Exploration via Scaling-Law Prediction [AUTHORS]Aishwarya Wesanekar, Sheng-Tao Yang, Jiawei Zhang, Flavio Sales Truzzi, Chen Chen, Nader Beigiparast, Shengyuan Huang, Roshanak Zilouchian Moghaddam, Vitor Cid, Colin Clement, Vinayak Thapliyal, Sebastian Ament, Jihao Andreas Lin, Vincent Huang, Kai Han, Boyi Jiang, Wenlin Chen, Chaorong Chen, Ashwin Kumar, Maximilian Balandat, David Eriksson, Yifei Zhang, Pan Zhong [ABSTRACT]Evaluating candidate changes (hypotheses) to recommendation models requires full-scale training runs consuming thousands of GPU-hours, severely limiting the hypothesis space engineers can feasibly explore. We introduce Predictable Scaling (PS), a systematic framework that exploits empirically verified regularity conditions to predict full-scale model performance from small-scale proxies. PS constructs scaled-down variants via composable scaling functions along distinct resource axes, and projects metrics to the target scale via curve fitting with calibrated uncertainty estimates. We compare three prediction strategies of increasing expressiveness: even the simplest (raw proxy) achieves prediction error within the empirical noise floor of repeated-run variance; the most expressive (Gaussian Process) provides an indicative 15% relative MAE improvement and calibrated uncertainty. Evaluation on two ranking models of increasing complexity---a task-specific model and a foundation model---demonstrates that PS correctly ranks actionable hypotheses across diverse hypothesis types at 79-88% compute reductions. [LINK]http://arxiv.org/abs/2610.23453v1 [DATE]2026-10-16 21:35:54+08:00 [CATEGORIES]cs.LG
Generative Policy Improvement from Observational Data [AUTHORS]Jonas Schweisthal, Dennis Frauen, Stefan Feuerriegel, Emma Brunskill [ABSTRACT]We study how to improve decision policies from observational data to generate better actions when actions are complex, structured objects, such as text messages or images. This task is subject to two major challenges: (i) action-level overlap is often too weak for reliable off-policy evaluation and improvement, and (ii) generating new structured actions often requires expensive training or fine-tuning (e.g., of an LLM) and can be ill-defined due to information loss in action representations. To address both, we propose PolicyLift, a three-stage framework for generative policy improvement with structured actions from observational data. In Stage 1, PolicyLift learns a small set of discrete action codes that capture high-level strategies in the observed action space and are tailored to support policy improvement under limited overlap. In Stage 2, it learns a context-dependent policy over these action codes. At deployment time, Stage 3 "lifts" the learned code-level policy back to the original action space by generating actions from the historical behavior system in a code-conditioned way. We show theoretically that improving the policy over action codes guarantees improvement of the induced policy in the original action space. We demonstrate PolicyLift across experiments in synthetic and real-world text generation tasks with consistent policy improvements over the behavior policy. [COMMENTS]Preprint [LINK]http://arxiv.org/abs/2610.23436v1 [DATE]2026-10-16 21:20:43+08:00 [CATEGORIES]cs.LG
Bridging Compressed Communication and Local Steps via Spectral Method in Federated Learning [AUTHORS]Dmitriy Bystrov, George Chirkov, Daniil Medyakov, Dmitry Bylinkin, Aleksandr Beznosikov [ABSTRACT]Federated Learning (FL) enables collaborative training of complex models across large-scale networks of edge devices while keeping data localized. However, data heterogeneity and high communication overhead remain fundamental bottlenecks. Recently, the spectral optimizer Muon has demonstrated remarkable advantages over Adam-like methods in centralized training. While recent efforts have attempted to transition Muon to FL, naive integrations into traditional architectures fail short. Crucially, existing approaches do not support the the communication compression and local steps --- the two foundational techniques of efficient FL. To address this limitation, we introduce a novel spectral FL method that utilizes both local updates and compression. We establish rigorous convergence guarantees under mild assumptions, notably eliminating the restrictive bounded heterogeneity condition. Empirical evaluation on standard image classification benchmarks validates the efficiency and robustness of our approach. [LINK]http://arxiv.org/abs/2610.23434v1 [DATE]2026-10-16 21:19:03+08:00 [CATEGORIES]cs.LG
Residual Stochastic Concept Bottleneck Models [AUTHORS]Stephen Marcou, Mateo Espinosa Zarlenga, Sonia Laguna, Moritz Vandenhirtz, Julia E. Vogt, Mateja Jamnik [ABSTRACT]Concept-based models (CMs) learn human-interpretable intermediate representations that enable users to understand the model's reasoning and steer its predictions through test-time concept interventions. However, in realistic settings, their performance can be limited by concept incompleteness, where the training concepts fail to capture all task-relevant information. Adding an unconstrained residual pathway to the model's predictive process can overcome the consequences of incompleteness. However, this introduces an inscrutable, leak-prone prediction pathway that may weaken concept interventions, as these often leave residual variables unchanged. To address this limitation, we introduce Residual Stochastic Concept Bottleneck Models (Res-SCBMs), a family of concept-based models that jointly model concept and residual variables through a multivariate Gaussian distribution. Res-SCBMs learn a covariance structure that explicitly captures dependencies between the known concepts and learnt residual variables, allowing interventions to propagate to statistically dependent residual variables through Gaussian conditioning. This extends the intervention capabilities of previous CMs without requiring concept completeness during training. Our results show that, across multiple concept-incomplete benchmarks, Res-SCBMs improve predictive performance over competing baselines while retaining effective test-time concept interventions. [LINK]http://arxiv.org/abs/2610.23421v1 [DATE]2026-10-16 21:09:16+08:00 [CATEGORIES]cs.LG
2026 Oct 17, Sat
GLOBE: Evolving Structured Agentic Memory for E-commerce Product Optimization in Generative Search [AUTHORS]Hyunseo Kim, Suil Choe, Sukkyoung So, Daeyeon Yi [ABSTRACT]Optimizing product listings for generative search requires language agents not only to rewrite effectively but also to continually self-improve by retaining successful strategies across diverse categories. However, existing approaches often treat structured listings as free-form text, leaving effective strategies implicit in prompts, search trajectories, or rewritten outputs rather than accumulating them as a reusable agentic memory. To address this limitation, we introduce GLOBE, an agentic framework that drives continual self-improvement by storing optimization knowledge in an evolving handbook of explicit, field-scoped behaviors. In each episode, an Editor applies relevant behaviors and receives rank-based feedback, which a Reflector converts into behavior updates. Crucially for dependable execution in real-world settings, a Moderator removes redundant proposals, a Curator commits accepted changes, and consistently harmful behaviors are deactivated. This cycle transforms episodic feedback into a persistent, field-scoped optimization memory. Experiments show that structured behaviors outperform monolithic prompt optimization, while feedback-driven evolution and moderation further improve effectiveness and reduce downside risk. The learned behaviors also remain effective on unseen product categories, suggesting that agentic memory evolution supports interpretable and transferable product optimization. [COMMENTS]Accepted to the Second Workshop for Research on Agent Language Models(REALM) at EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23709v1 [DATE]2026-10-17 00:33:17+08:00 [CATEGORIES]cs.CL
Learning Personalized Memory Gating from Delayed One-Sided Feedback [AUTHORS]Kaixuan Ji, Zixiang Chen, Wenting Zhao, Cheng Qian, Heng Wang, Weizhi Zhang, Jielin Qiu, Liangwei Yang, Quanquan Gu, Silvio Savarese, Shelby Heinecke, Huan Wang [ABSTRACT]Long-term memory enables language-model agents to personalize future interactions, but a finite memory bank cannot preserve every extracted fact. Existing memory gating policies are typically shared across users and therefore fail to account for user-specific patterns of future reuse. In this paper, we study memory gating at the fact level. We first propose a retrieval-based labeling strategy that labels each fact by whether it is later retrieved and incorporated into responses. We then formulate online memory gating as contextual apple tasting with delayed one-sided feedback, and propose Warm-Start UCB, which combines offline learning from existing data with online adaptation to the current user. Our theoretical analysis establishes a high-probability regret upper bound of $\tilde\{O\}(Tε\wedge d\sqrt\{T\})$ when the delay window is at most $O(\sqrt\{T\})$, where $T$ is the total number of rounds, $ε$ is the initialization error, $d$ is the dimension of the feature map, and $\tilde\{O\}(\cdot)$ hides constant and poly-logarithmic factors. The results of our experiment demonstrate that our method outperforms all baselines under different memory budgets and history lengths, establishing it as a promising personalized memory gating policy. [LINK]http://arxiv.org/abs/2610.23812v1 [DATE]2026-10-17 01:34:45+08:00 [CATEGORIES]cs.LG
LeanPRM: Lean-Grounded Process Supervision for Natural Language Proof Verification [AUTHORS]Rodney Lafuente-Mercado, Salman Rahman, Jingyan Shen, James Shiffer, Saadia Gabriel, Pavel Izmailov [ABSTRACT]AI systems now write long and difficult mathematical proofs, and checking their step-by-step validity is becoming a bottleneck. A learned verifier could check proofs cheaply, but training one requires a correctness label on every step, and such labels are hard to obtain for open-ended proofs. Existing methods take them from human annotators, which is costly, or derive them from final answers, which proofs do not have, and formalizing each proof in Lean requires expert effort. We introduce LeanPRM, a framework that uses the Lean proof assistant to create step-level labels for natural-language proofs. A model writes a natural-language proof together with a paired formal proof, without feedback from Lean. Lean then labels every step, and a second model discards failures caused only by the formal syntax, so that the negative labels mark errors in the mathematics. LeanPRM contains 10,634 proofs with 152K labeled steps and no human annotation. A Qwen3-4B verifier trained on LeanPRM raises exact step-label accuracy on a held-out set of 200 proofs (LeanPRM-Bench) from 48.2 to 74.7, as opposed to 60.3 when the same base model is trained on PRM800K. It also selects better proofs in best-of-N selection on IMO-level problems, and outperforms the PRM800K-trained verifier on three of four external verification benchmarks. Training on LeanPRM improves verifiers at every size from 0.6B to 8B parameters on LeanPRM-Bench, Hard2Verify and IMO-GradingBench. Given the labels need no human annotation, LeanPRM can grow with the supply of formal statements, offering a path to step-level feedback on AI-generated proofs that are hard for humans to check. [COMMENTS]29 pages, 9 figures. Project page: https://rlafuente.com/lean-prm [LINK]http://arxiv.org/abs/2610.17697v2 [DATE]2026-10-17 01:31:06+08:00 [CATEGORIES]cs.LG
Learning What to Say to Your VLA: Mostly Harmless Vision Language Action Model Steering [AUTHORS]Hyun Joe Jeong, Gokul Swamy, Andrea Bajcsy [ABSTRACT]Vision-Language-Action (VLA) models provide a natural language interface to robot control, but the mapping from language to behavior is often brittle and unintuitive: semantically similar instructions can induce drastically different behaviors, while some capabilities may not be elicitable through prompting alone. As a result, both human instructions and zero-shot language models can fail to reliably steer VLAs toward successful task execution. In this work, we propose a framework that interactively searches for language sequences that improve closed-loop VLA task performance, distills these sequences into a test-time language feedback policy (LFP), and learns an improvement head that predicts when language steering will improve performance. We conformalize this improvement head to prevent harmful steering interventions, where the LFP decreases task performance relative to the original instruction on out-of-distribution scenarios. Crucially, our approach operates on arbitrary frozen pre-trained VLAs, requiring neither access to the original training distribution nor fine-tuning of the underlying model. On seen environments, our conformalized LFP improves base VLA performance by 24.7% in simulation and 65.0% in hardware. On visual and semantic perturbations with calibration samples, our conformalized LFP has strong harmlessness guarantees, and produces recovery behaviors not observed with open-loop prompting. [COMMENTS]CoRL 2026 [LINK]http://arxiv.org/abs/2606.12299v2 [DATE]2026-10-17 01:28:27+08:00 [CATEGORIES]cs.LG
Interpretable experiential learning based on state history and global feedback [AUTHORS]Anton Kolonin [ABSTRACT]A new interpretable experiential learning model based on state history and global feedback is presented. It is capable of learning a behavioral model represented by a transition graph between sets of states, with transitions attributed with utility and evidence count. This model is expected to be suitable for solving reinforcement learning problem in resource-constrained environments. The model was thoroughly evaluated on the OpenAI Gym Atari Breakout benchmark, demonstrating performance comparable to some known neural network-based solutions. [COMMENTS]5 figures. This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible [LINK]http://arxiv.org/abs/2605.00940v2 [DATE]2026-10-17 00:12:02+08:00 [CATEGORIES]cs.LG
2026 Oct 16, Fri
FSPO: Policy-Consistent Risk and Pareto-Feasible Control for Budgeted LLM RL Post-Training [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Daren Zha [ABSTRACT]Adaptive LLM reinforcement-learning post-training changes multiple training actuators online, including rollout temperature, group size, clipping, KL regularization, verifier allocation, and update budget. Three coupled issues remain unresolved. A future-risk model trained from behavior trajectories need not estimate the risk induced by the controller that will be deployed; a score calibrated on logged state-action pairs can become miscalibrated after selective action choice; and independent per-resource minimum costs do not in general certify a feasible multi-resource continuation. We introduce FSPO, a feedback-state controller for budgeted LLM RL post-training that addresses these issues jointly. FSPO learns a policy-consistent risk-to-go model whose Bellman target follows the same frozen controller used for future decisions, together with a long-horizon utility model. Decision-conditioned trajectory calibration (DCTC) calibrates risk on cross-fitted trajectories generated by actions selected by provisional controllers. A Pareto resource continuation certificate (PRCC) admits an action only when a non-dominated cumulative reservation remains feasible over the residual horizon. Under a matched GRPO resource envelope, FSPO reaches 66.11% held-out and 59.43% OOD accuracy, compared with 64.47% and 57.03% for PB2, the strongest evaluated adaptive baseline. Three paired training seeds give gains of +2.42 and +3.19 percentage points over the contextual bandit on held-out and OOD evaluation. Under high behavior-deployment mismatch, policy-consistent risk lowers selected-decision ECE from 0.108 to 0.053; DCTC lowers it from 0.039 to 0.022 at matched acceptance; PRCC removes false-feasible admissions on an 18-action catalog ($0.197\rightarrow0.000$); and enabling all three components reduces trajectory failure from 0.181 to 0.083 in a factorial ablation. [COMMENTS]40 pages, 4 figures [LINK]http://arxiv.org/abs/2610.02828v2 [DATE]2026-10-16 20:44:10+08:00 [CATEGORIES]cs.CLcs.LG
Two Minds Better Than One: Collaborative Reward Modeling for LLM Alignment [AUTHORS]Jiazheng Zhang, Wenqing Jing, Zizhuo Zhang, Zhiheng Xi, Shihan Dou, Rongxiang Weng, Jiahuan Li, Jingang Wang, Mingxu Chai, Shibo Hong, Yuming Yang, Chenhao Huang, Tao Gui, Qi Zhang, Xuanjing Huang [ABSTRACT]Reward models (RMs) are central to aligning large language models (LLMs) with human values, yet noisy preference feedback can induce reward misgeneralization through spurious correlations and overfitting. We characterize the training dynamics of preference pairs and uncover a distinctive optimization signature of noise: persistently high losses, sharp temporal fluctuations, and irregular gradient updates. Our analyses further show that excluding such pairs improves downstream alignment. In practice, however, reliable preferences are not known a priori, and self-filtering based on a model's own training signals can amplify confirmation bias. To address this challenge, we introduce Collaborative Reward Modeling (CRM), an online framework for identifying reliable preferences. CRM couples cross-model peer review with curriculum learning: two RMs evaluate each other's sample selections, while curriculum keeps their learning progress synchronized. Across extensive experiments, CRM improves RM generalization by up to 9.94 points on RewardBench under 40% preference noise. CRM further generalizes to implicit-reward alignment, offering a robust and broadly applicable approach to learning from imperfect feedback. [LINK]http://arxiv.org/abs/2505.10597v3 [DATE]2026-10-16 20:42:25+08:00 [CATEGORIES]cs.LGcs.CL
ContractRL: Shielded Group-Relative Policy Optimization for Auditable Tool-Call Repair [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Yina Sa, Daren Zha [ABSTRACT]Structured tool calls often fail after only a small number of fields violate a schema or an execution contract. Regenerating the complete object enlarges the action surface and makes repeated repair difficult to audit. We introduce ContractRL, a contract-constrained sequential repair protocol that models verifier-guided JSON repair as a bounded decision process. At each step the policy observes the candidate, typed verifier feedback, JSON Pointer, immutable repair history, and remaining budget; a contract-derived action mask filters malformed or prohibited RFC-6902 operations before a deterministic validator performs the transition. We specify a contract-constrained group-relative objective for patch, retry, and abstention decisions while keeping canonical targets and semantic labels outside the online state until trace freeze. Under identical verifier information, ContractRL attains 0.9362 semantic success with 34.4 generated tokens, compared with 0.9076 and 44.9 tokens for Patch-SFT and 0.9148 and 137.2 tokens for full regeneration over 192 cases per seed and five seeds. Policy optimization improves semantic success from 0.9186 for supervised ContractRL to 0.9375. A separate three-seed paired evaluation against Patch-SFT yields a semantic difference of +0.0396 (95% CI $[+0.0137,+0.0662], p=0.0039$). Feedback, action-mask, budget, and schema-shift analyses connect these gains to localized correction, while adversarial and multi-turn evaluations characterize the remaining failure modes. [COMMENTS]29 pages, 8 figures [LINK]http://arxiv.org/abs/2610.00328v2 [DATE]2026-10-16 20:23:26+08:00 [CATEGORIES]cs.CLcs.LG
FAER: Auditable Utility-Aligned Trajectory Replay for Language Model Post-Training [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Tianshu Fu, Daren Zha [ABSTRACT]Replay selectors often rank cached trajectories by format feedback, confidence, freshness, or response length, although cache-level correctness and downstream learner utility are distinct objectives. We formalize this selection-to-learning gap and introduce FAER as an auditable full-trajectory replay framework. Its training-free fixed selector is a protocol baseline; FAER-UTILITY is the learner-aware selector fitted on disjoint calibration blocks. The normalized gradient alignment is reported as a baseline, while a disposable optimizer-aware virtual update supplies a magnitude-aware utility surface. The audit contract freezes observed fields and replay traces before evaluation labels are joined. On GSM8K with Qwen2.5-1.5B-Instruct, the matched learner study reports quality 0.6329 for the fixed selector, compared with 0.5482 for uniform and 0.6037 for format-feedback under 128 updates. Metadata-only cross-fitted calibration reaches $0.6476\!\pm\!0.0139$ over eight seeds (median 0.6481; paired 95% interval $[+0.079,+0.122]$) at 63,276 target-run tokens; its recorded full cost is 189,642 tokens and 3.48 GPU-hours including calibration. The completed FAER-UTILITY row reaches 0.6624 at 62,844 target-run tokens and 4.26 GPU-hours. Format-feedback selects records with correctness 0.6953, compared with 0.3594 for the fixed selector, despite the different downstream ranking. The completed comparison surfaces report the learner-aware ablation, same-seed gap, policy-optimization rows, and strict zero-shot transfer. [COMMENTS]35 pages, 6 figures [LINK]http://arxiv.org/abs/2610.00385v2 [DATE]2026-10-16 20:01:19+08:00 [CATEGORIES]cs.LGcs.CL
Language Models Learn to Improve Language Models [AUTHORS]Minhua Lin, Zijun Wang, Zewen Liu, Zhaotian Weng, Bing He, Zhan Shi, Benoit Dumoulin, Xiang Zhang, Suhang Wang, Hanqing Lu [ABSTRACT]Large language models (LLMs) are increasingly used to automate AI research, including language-model (LM) development, offering a potential path toward recursive self-improvement. However, turning experimental feedback into a learned ability to propose effective improvements across models faces two challenges. First, training such an ability with real feedback is expensive: reinforcement learning (RL) requires evaluating tens of thousands of proposals, each potentially requiring a full model-training run. Even at ten GPU-hours per run, reward collection alone would consume hundreds of thousands of GPU-hours, motivating a cheaper surrogate. Second, surrogate feedback can become unreliable as policy optimization favors overestimated proposals when the proposal distribution shifts. To address these challenges, we propose LM2LM, which trains LLMs with RL to generate effective LM improvement proposals. To provide scalable feedback, we use a research world model (RWM), a frozen LLM that predicts proposal gains from real experimental records provided in context, as the reward model for RL. To maintain feedback reliability, we periodically execute current-policy proposals ranked highly by the RWM and add their measured outcomes to its context. Across eight autoregressive and diffusion LM development environments, real-execution evaluations on held-out tasks show that LM2LM enables Qwen3.8-27B to outperform GPT-6-sol, even when GPT-6-sol is given 300 real experimental records in context, while requiring 33% lower data-generation cost. LM2LM also generalizes to a model family unseen during training. [LINK]http://arxiv.org/abs/2610.23344v1 [DATE]2026-10-16 19:51:00+08:00 [CATEGORIES]cs.CL
RubricEM: Meta-RL with Rubric-guided Policy Decomposition beyond Verifiable Rewards [AUTHORS]Gaotang Li, Bhavana Dalvi Mishra, Zifeng Wang, Jun Yan, Yanfei Chen, Chun-Liang Li, Long T. Le, Rujun Han, George Lee, Hanghang Tong, Chen-Yu Lee, Tomas Pfister [ABSTRACT]Deep research agents plan, search, evaluate evidence, and synthesize long-form reports. Training these agents with reinforcement learning (RL) is challenging because they operate over long horizons and their outputs lack exact answer verifiers. To this end, we introduce RubricEM, a unified RL framework for deep research that combines stagewise credit assignment with reflection-based meta-RL for experience reuse. Rubrics unify the framework by providing a shared interface for policy execution, judge feedback, and experience reuse. RubricEM first instills a rubric-guided scaffold through supervised fine-tuning, teaching the agent to plan with rubrics and organize its trajectory into semantic stages. Building on this structural prior, we optimize the task policy with Stage-Structured GRPO, using stagewise judge feedback for finer-grained credit assignment. In parallel, we asynchronously train a shared-backbone reflection meta-policy (with meta-RL) to distill completed trajectories into reusable guidance, contributing through both shared-parameter updates and textual memory. The resulting RubricEM-8B model achieves strong performance across four long-form research benchmarks, outperforming existing open models of larger sizes and approaching proprietary deep-research systems. Beyond final performance, we perform thorough controlled analyses to understand the key ingredients of RubricEM. [COMMENTS]67 pages, 6 figures [LINK]http://arxiv.org/abs/2605.10899v2 [DATE]2026-10-16 19:35:29+08:00 [CATEGORIES]cs.CLcs.LG
MultiAgent-OPD: End-to-End Training of Hierarchical Multi-Agent Systems with On-Policy Distillation [AUTHORS]Yuhan Sun, Jiaxin Zhao, Yongdong Luo, Yibo Hu, Meiguang Jin, Junfeng Ma, Weixun Wang, Weihang Pan, Zhongan Bi [ABSTRACT]In a hierarchical multi-agent system, a Leader assigns tasks and subagents carry them out. Jointly training the Leader and the subagents is difficult because a final answer gives only indirect feedback to the subagents. An LLM judge can score the performance of the subagents, but these scores depend on the judge's own assessment of the quality of subagents' responses. We present MultiAgent-OPD, which combines on-policy distillation with reinforcement learning to jointly train a shared-parameter Leader-Searcher-Answerer team end to end. We first train a Teacher in a Searcher-Answerer workflow using answer correctness. During team training, the frozen Teacher scores student-generated search and summary tokens to provide local process supervision for the Searchers, while the Leader learns from team success and the Answerer from candidate correctness. Across seven datasets, the trained system reaches 38.68% macro avg@4, exceeding Prompted Critic by 4.06 percentage points and C-MAS + RL, the trained cascaded multiagent system, by 2.61 percentage points. The team also solves more questions within four attempts. [LINK]http://arxiv.org/abs/2610.23234v1 [DATE]2026-10-16 18:01:52+08:00 [CATEGORIES]cs.CL
Brain-to-Language Decoding: Paradigms, Signals, Methods, Evaluation, Practical Use and Beyond [AUTHORS]Yiqian Yang, Yiqun Duan, Chenyu Liu, Yiqi Wang, Xinliang Zhou, Chin-Teng Lin, Yu Zhang [ABSTRACT]Brain-to-language decoding translates neural activity associated with language production, internal speech and perception into linguistic or expressive outputs. It offers a route to restoring communication after speech loss and a means of studying how the brain represents language. Advances in neural recording and representation learning have expanded the field from constrained recognition and acoustic reconstruction to text generation, streaming personalised speech and facial animation. This survey synthesises these developments across invasive and non-invasive measurements, drawing on a search without a lower year limit and source-led updates through September 2026. We connect Articulated, Inner and Perceived tasks to the neural populations they engage, the representations available to decoders and the outputs those representations can support. We examine model development, public resources and the evolution of evaluation, and compare published performance and communication costs within their reported protocols. The synthesis identifies complementary routes to progress: phonetic, acoustic and semantic targets preserve different aspects of a message; shared representations support reuse across recording conditions and tasks; and online communication increasingly depends on calibration, feedback and user control alongside decoding accuracy. Shared benchmarks enable algorithmic comparisons, while longitudinal studies reveal the demands of sustained use. We discuss these developments and their remaining limitations, then outline a prospective five-level trajectory from commands and language to meaning, scenarios and bidirectional cognitive exchange [LINK]http://arxiv.org/abs/2609.27650v4 [DATE]2026-10-16 17:37:51+08:00 [CATEGORIES]cs.CL
EKLAV: Elicited Key CoT Learning via Answer only superVision [AUTHORS]Adarsh Singh, Aman Singh, Kushal Raj Bhandari, Soham Dan, Anoop Kumar, Vivek Gupta [ABSTRACT]Reasoning distillation trains a student to reproduce both the teacher's chain-of-thought trace and its final answer. We ask whether the trace is better used as privileged training context than as an imitation target. EKLAV keeps the teacher's reasoning in the input sequence but supervises only the final task answer, applying a task-dependent placement that moves part or all of the trace into the prompt. On BRIGHT passage reranking, EKLAV improves average nDCG@10 at all four tested checkpoints across two model families, winning 41 of 48 model-domain comparisons while reducing training compute by up to 31.6% and inference time by up to 32.7%. On four held-out table-reranking benchmarks, it is the best of four conditions on three, exceeding full-trace distillation by 3.0 points and answer-only training without reasoning by 5.5 points on average. Analysis of generated traces reveals that EKLAV students defer their first verbalized verdict later (0.76 vs. 0.44 relative position), reverse fewer decisions (5% vs. 16%), and produce reasoning more predictive of their own rankings than STD-COT's, though less predictive than the base model's. The recipe reuses a fixed teacher corpus, requires no student rollouts or iterative teacher feedback, and applies to both pointwise and listwise formulations. Code and Data are at https://github.com/AdarshSingh7647/Eklavya [COMMENTS]8 pages, 3 figures [LINK]http://arxiv.org/abs/2607.25182v2 [DATE]2026-10-16 17:04:49+08:00 [CATEGORIES]cs.CL
Causal Discovery beyond Conditional Independence: Identifiability, Algorithms, and Statistical Inference for Directed Acyclic and Cyclic Graphs [AUTHORS]Yang Ni [ABSTRACT]Learning causal structure from data requires assumptions about mechanisms, noises, observation schemes, and interventions. Historically, causal discovery has relied primarily on conditional independence. This review focuses on modern causal discovery methods that exploit properties of distributions beyond conditional independence. Our organizing principle separates the source of identification from the statistical estimator and its computational implementation. This perspective connects conditional-independence methods with distributional asymmetries, conditional-moment restrictions, and variation across environments or individuals. We examine discovery under latent confounding and feedback, distinguishing causal information about observed variables from reconstruction of latent structure and emphasizing the additional semantics required by cyclic models. We also distinguish population identification, frequentist and Bayesian consistency, uncertainty quantification, and algorithmic guarantees. Throughout, examples show why the same observational distribution can support different conclusions under different model classes. We conclude with guidance for comparing methods at their identifiable targets and a research agenda for uncertainty, heterogeneous populations, non-scalar variables, and the intersection of discrete data, feedback, and hidden confounding. [LINK]http://arxiv.org/abs/2610.23543v1 [DATE]2026-10-16 22:37:57+08:00 [CATEGORIES]cs.LG
2026 Oct 16, Fri
Cross-Dialect Generalization Without Retraining: Benchmarks and Evaluation of Schema-Derived Constrained Decoding for MLIR [AUTHORS]Plawan Kumar Rath [ABSTRACT]Multi-Level Intermediate Representation (MLIR) underlies modern ML compiler infrastructure (TensorFlow, JAX/StableHLO, PyTorch Inductor, IREE), yet appears only in trace amounts in code-LM pretraining corpora. MLIR is also extensible by design: new dialects ship per application domain, so a fine-tuned model per dialect does not scale. We ask whether inference-time priors derived mechanically from each dialect's Operation Definition Specification (ODS) can substitute for gradient-based adaptation. First, we release four natural-language-to-MLIR benchmarks across three dialects (MLIR-Spec-150, Linalg-Spec-30, StableHLO-Spec-30, StableHLO-Held-Out-200), totaling 410 in-scope NL-to-MLIR pairs, plus a 25-program out-of-grammar stress set, with functional references on 30 Spec prompts, under Apache-2.0 with Gebru datasheets and Croissant 1.1 metadata. Second, we build a three-layer schema-derived constraint stack: a CFG over op signatures (C1), type-domain splits from an ODS-extracted type lattice (C2), and an SSA-scope validator driving five-retry rejection sampling (C3). Porting it to StableHLO required no new constraint-layer code. On structurally dominated dialects, SmolLM2-1.7B with token-level constraints matches or exceeds 15B-34B code LMs running post-hoc rejection at a matched retry budget and 8-25x the per-generation speed: on linalg it reaches 80.0% verify-valid (three runs, n=125), 21 to 44 pp above CodeLlama-34B, Granite-Code-34B, and StarCoder2-15B with non-overlapping CIs. The comparison is between systems, not models: under the baselines' post-hoc protocol, SmolLM2 falls from 52.0% to 24.5% on arith+func. On arith+func and the templated StableHLO-Held-Out-200 the baselines match or beat the SLM; we scope these as non-win cells. We release benchmarks, decoder, every per-prompt generation, and a pinned reproducibility Dockerfile. [COMMENTS]Accepted to the NeurIPS 2026 Evaluations and Datasets Track; camera-ready version. 31 pages (10 main + references + appendix + checklist). Code: https://github.com/plawanrath/slm-to-mlir-constrained-emitter Datasets: https://huggingface.co/collections/plawanrath/cross-dialect-mlir-benchmarks-neurips-2026-e-and-d [LINK]http://arxiv.org/abs/2607.18254v2 [DATE]2026-10-16 22:07:14+08:00 [CATEGORIES]cs.LG
2026 Oct 16, Fri
GRAFT-EEG: Gauge-Aware Physical Field Transport for Cross-Dataset Motor-Imagery EEG [AUTHORS]Ying Zhou, Mengting Liu [ABSTRACT]Cross-dataset EEG decoding often treats montage, reference, and active-sensor changes as representation shift, although they also alter the observation operator. We study target-statistics-blind transfer: fitting receives no target signal, label, or empirical statistic, and electrode coordinates only instantiate a fixed prediction operator. method removes the active-set common-reference gauge and transports each native mask through a template motor leadfield. A source-fitted spectral classifier anchors the decision; a bounded state-space-model (SSM) residual is secondary. We establish restricted identifiability and an operator-perturbation bound. Across five leave-one-dataset-out targets (253 participants), dynamic method achieved 60.865% equal-target balanced accuracy versus 52.684% for strict-clean 5/5 BIOT (hierarchical difference +8.182 points; 95% CI +6.382 to +10.196; positive in every target). On four EEGPT-clean targets, the descriptive means were 61.520% and 59.057%. In four post-hoc exact same-parent tests, FULL exceeded coordinate-basis, matched-spectrum, row-shuffled, and wrong-anatomy controls by 9.959, 9.698, 11.241, and 3.985 points, respectively (all 95% CIs excluded zero; Holm-adjusted P<0.001). A separate 60-participant same-paradigm evaluation retained a positive physical-operator contrast (+4.029 points; +2.338 to +5.773), whereas dynamic--static was inconclusive; its preregistered status is partial confirmation rather than unrestricted external validation. The SSM added +1.059 points (+0.013 to +1.892). Claims remain bounded to five binary motor-imagery domains and the tested operators. [COMMENTS]Under review at ICLR 2027 [LINK]http://arxiv.org/abs/2610.23620v1 [DATE]2026-10-16 23:30:37+08:00 [CATEGORIES]cs.LG
2026 Oct 17, Sat
CocoRNA: Collective RNA Design with Cooperative Multi-agent Reinforcement Learning [AUTHORS]Tianmeng Hu, Biao Luo, Ke Li [ABSTRACT]Designing RNA sequences that reliably fold into specific secondary structures is essential for understanding their biological functions but remains a challenging computational problem. We propose CocoRNA, a cooperative multi-agent reinforcement learning framework for RNA inverse design. CocoRNA simplifies the design task by decomposing it into smaller sub-problems, each solved collaboratively by multiple agents. This approach reduces the complexity of the problem and improves the exploration of design policies. During training, a centralized critic uses global structural information to guide the agents, enabling them to jointly optimize their design strategies. As a result, CocoRNA learns high-quality RNA design policies that generalize effectively to unseen structures without additional training. Experiments on the Rfam dataset demonstrate that CocoRNA substantially outperforms state-of-the-art methods in both success rate and design speed. Further experiments on other biological sequence design tasks highlight the effectiveness and broad potential of CocoRNA for complex design tasks. [COMMENTS]ICML 2026 [LINK]http://arxiv.org/abs/2610.23820v1 [DATE]2026-10-17 01:43:25+08:00 [CATEGORIES]cs.LG
2026 Oct 17, Sat
Behavioral Couplings: From Formation to Behavioral Emergence in Language Models [AUTHORS]Dongyoon Hahm, Woogyeol Jin, Dennis Wei, Yi Zhou, Nathalie Baracaldo, Kimin Lee [ABSTRACT]Fine-tuning large language models (LLMs) can induce behaviors beyond those explicitly targeted. For example, training a model to respond more warmly can make it more sycophantic, and training it to write insecure code can cause broad misalignment on prompts unrelated to coding. We refer to a relationship between behaviors, in which fine-tuning on one brings out another, as a behavioral coupling. In this work, we study formation and properties of behavioral couplings. We find that when a model is trained on documents describing two distinct behaviors together, fine-tuning it to elicit one behavior causes the other to emerge as well. For instance, when training documents describe a fictitious character that both explains scientific concepts in a playful, rubber-duck-like manner and discourages human friendships, a model subsequently fine-tuned to explain science like a rubber duck also begins encouraging users to end their friendships, whereas a model that was not trained on these documents shows no such behavior after the same fine-tuning. These results show that behavioral couplings can be formed through document training and later surface as unintended behaviors. More concerningly, we find that couplings can compose. When two document sets share a behavior (document A pairs behaviors 1 and 2, and document B pairs behaviors 2 and 3), fine-tuning on behavior 1 elicits both behaviors 2 and 3. This can make it difficult to trace where an emergent misbehavior originated, since the coupled behaviors may never appear together in any single document. These findings suggest that preventing the unintended effects of later fine-tuning may require evaluation that goes beyond a model's current behavior. [LINK]http://arxiv.org/abs/2610.23847v1 [DATE]2026-10-17 01:58:08+08:00 [CATEGORIES]cs.CLcs.LG
AI-Driven Reproducibility and Analytical Robustness in Scientific Discovery [AUTHORS]Shicheng Liu, Krista Opsahl-Ong, Andrew Drozdov, Julian Killingback, Arnav Singhvi, Ronak Pradeep, Michael Bendersky, James C. Dickerson, Monica S. Lam [ABSTRACT]Reproducible and analytically robust research findings are essential to scientific progress, but evaluating these properties typically requires costly manual effort. We investigate whether LLMagents can help automate reproduction and robustness analysis, using a medical case study grounded in the Surveillance, Epidemiology, and End Results (SEER) database, a widely used U.S. cancer registry underlying tens of thousands of published studies. First, we introduce SEER-Repro, a benchmark of 20 published SEER-based papers together with the data and resources required to reproduce their findings. Second, we create an AI agent that successfully reproduces 77.4% of 2,190 atomic claims, while providing executable baselines for further analysis. Third, we develop an agentic analysis pipeline to test the robustness of the medical findings by identifying analytical choices underlying each finding, proposing defensible alternatives, and rerunning the analysis to determine whether those choices affect the conclusion. Across published papers in SEER-Repro, our pipeline identifies potentially consequential analytical choices in 9.5% of fully reproduced findings and flags them for expert review. These results show how LLMagents can support scientific review by reproducing findings, systematically testing their robustness, and directing expert attention toward potentially consequential analytical choices. [LINK]http://arxiv.org/abs/2610.23845v1 [DATE]2026-10-17 01:57:55+08:00 [CATEGORIES]cs.CL
Optimizing Singular Spectrum for Large Language Model Compression [AUTHORS]Dengjie Li, Tiancheng Shen, Yao Zhou, Bernard Ghanem, Yibo Yang, Ming-Hsuan Yang [ABSTRACT]Large language models (LLMs) have demonstrated remarkable capabilities, yet prohibitive parameter complexity often hinders their deployment. Existing singular value decomposition (SVD) based compression methods simply deem singular values as importance scores of decomposed components. However, this importance ordered by singular values does not necessarily correlate with the performance of a downstream task. In this work, we introduce SoCo (Singular spectrum optimization for large language model Compression), a novel compression framework that learns to rescale the decomposed components of SVD in a data-driven manner. Concretely, we employ a learnable diagonal matrix to assign importance scores for singular spectrum and develop a three-stage training process that progressively refines these scores from initial coarse compression to fine-grained sparsification-thereby striking an effective balance between aggressive model compression and performance preservation. Thanks to the learnable singular spectrum, SoCo adaptively prunes components according to the sparsified importance scores, rather than relying on the fixed order of singular values. More importantly, the remaining components with amplified importance scores can compensate for the loss of the pruned ones. Experimental evaluations across multiple LLMs and benchmarks demonstrate that SoCo surpasses the state-of-the-art methods in model compression. [LINK]http://arxiv.org/abs/2502.15092v2 [DATE]2026-10-17 01:56:38+08:00 [CATEGORIES]cs.CL
Steering beyond Language: Improving Multilingual Training-Free Text Embeddings via Paired-Residual Subspace Removal [AUTHORS]Ailiang Lin, Zhuoyun Li, Yusong Wang, Keyu Mao, Kotaro Funakoshi, Manabu Okumura [ABSTRACT]While deriving text embeddings directly from frozen LLMs has emerged as a promising direction, existing methods focus almost exclusively on English, leaving multilingual settings largely unexplored. In this work, we present the first systematic study of multilingual training-free text embeddings in LLMs. Our analysis reveals a language-conditioned embedding geometry in which semantically equivalent sentences remain misaligned across languages. We hypothesize that language-specific variation can be disentangled from semantic content to improve embedding quality. To this end, we propose Paired-Residual Subspace Removal (PRSR), a training-free inference-time intervention that steers LLM representations toward a more language-invariant semantic space. PRSR first constructs paired residuals from multilingual parallel sentences to capture language-specific deviations. It then estimates layer-wise low-rank subspaces from these residuals and suppresses the corresponding components in the last-token hidden states. We conduct more than 200 monolingual evaluations across 18 languages using MTEB and MMTEB, along with cross-lingual bitext mining between English and over 200 other languages. Empirical results demonstrate that PRSR consistently improves the performance of existing training-free baselines in both monolingual and cross-lingual settings across different LLMs. [LINK]http://arxiv.org/abs/2610.23807v1 [DATE]2026-10-17 01:30:43+08:00 [CATEGORIES]cs.CL
DLite: History-free Speculative Decoding for Agentic AI [AUTHORS]Jiaming Xu, Hanzhen Wang, Tianlang Zhao, Jiewen Xiao, Chengze Yuan, Jiayi Pan, Yongkang Zhou, Jiancai Ye, Yu Wang, Guohao Dai [ABSTRACT]Speculative decoding has become a mainstream technique for accelerating large language model (LLM) inference by using a lightweight draft model to generate candidate tokens that are subsequently verified by the target model. However, agentic workloads introduce long and continuously evolving contexts and frequent multi-turn interactions. Under such workloads, we observe that existing draft models suffer from up to an 80% degradation in acceptance length, while maintaining their historical KV cache introduces substantial system overhead. In this paper, we identify the growing historical context as a common source of both problems and introduce DLite, a history-free speculative decoding framework for agentic AI. DLite directly reuses contextual hidden states from the target model and performs local speculative prediction without explicitly maintaining the historical context, transforming an increasingly complex modeling problem into local simple prediction problem. A lightweight sequential head further restores intra-block causal dependencies with low overhead. To effectively train this heterogeneous history-free architecture, we introduce a tailored two-stage training strategy. Experiments on general-purpose, long-context and agentic workloads show that DLite achieves 2-6 acceptance length and outperforms the state-of-the-art speculative decoding method, DSpark, up to 70%. [COMMENTS]14 pages, 5 figures [LINK]http://arxiv.org/abs/2610.23787v1 [DATE]2026-10-17 01:18:54+08:00 [CATEGORIES]cs.CL
Language steering in latent space to mitigate unintended code-switching [AUTHORS]Andrey Goncharov, Nikolai Kondusov, Alexey Zaytsev [ABSTRACT]Multilingual Large Language Models (LLMs) often exhibit hallucinations such as unintended code-switching, reducing reliability in downstream tasks. We propose latent-space language steering, a lightweight inference-time method that identifies language directions via Principal Component Analysis (PCA) on parallel translations and steers token embeddings along these axes to control language identity. Our approach mitigates code-switching (reduced next-token distributional divergence by up to 55\%, reduced Code-Switching Index by 63--99\%) while preserving broad semantic direction with negligible computational overhead and requires only minimal parallel data for calibration. Additionally, we evaluate language classification using a single principal component (95-99\% accuracy) and further analyze the layer-wise evolution of language representations. [COMMENTS]Published at AINL 2026 [LINK]http://arxiv.org/abs/2510.13849v4 [DATE]2026-10-17 01:02:53+08:00 [CATEGORIES]cs.CLcs.LG
The Italian lawmaking process: a unified and structured dataset [AUTHORS]Matilde Contestabile, Chiara Ferrara, Giovanni Parrillo, Lorenzo Emer, Andrea Vandin [ABSTRACT]The Italian legislative process is documented in detailed parliamentary records hosted in institutional websites. However, much of this information is available only as unstructured text, limiting its systematic reuse. We applied a controlled, systematic LLM-based extraction procedure, validated against manually annotated records, to convert these texts into ordered and dated events including, e.g., bill introduction, committee examination, plenary examination, and approval. The resulting dataset reconstructs the procedural histories of all 5346 Italian laws enacted between 1985 and 2026, spanning parliamentary terms IX to XIX, for overall 101413 events. Each event records the activity, date, and chamber, with additional procedural details where available. The dataset supports research in the emerging fields of computational legal and legislative studies, potentially combined with process-oriented data science methods to examine procedural paths and legislative timing across parliamentary terms. [LINK]http://arxiv.org/abs/2610.23741v1 [DATE]2026-10-17 00:52:11+08:00 [CATEGORIES]cs.CL
Recognizing Study Lineage Is Not Enough: Over-Crediting Same-Trial Evidence in LLM-Based Medical Verification [AUTHORS]Mukand Krishna, Anoushka Harit, Chenghao Xiao, Zhongtian Sun [ABSTRACT]Publication count is not study count. A model can identify shared trial lineage correctly in a direct task yet over-credit a second publication from the same trial when aggregating evidence. We evaluate this behavior in medical LLM verification using 14 source-audited clinical families and five model conditions. The families span provenance, clinical applicability, result direction, and composition; the provenance families provide the direct lineage tests. Recognition and aggregation are separate stateless calls, and repeated calls assess within-input stability. Across all three original provenance families, GPT-5.5 identifies the independent update in 10/10 recognition calls per family but over-credits the same-trial report in 5/5 separate aggregation calls per family. Selected Qwen3-4B and Ministral-3-8B conditions show the same cross-task contrast. A study-grouping instruction reduces GPT-5.5 over-credit from 4/5 to 0/5 in each later pure-provenance family. Other model conditions show mixed behavior, and rejected outputs limit several comparisons. The results identify a behavioral gap between elicited lineage recognition and evidence aggregation. The code and data artifacts are available at https://anonymous.4open.science/r/study-lineage-review-artifact. [COMMENTS]20 pages, 5 figures, 17 tables; 8 pages of main content, with references and appendices thereafter [LINK]http://arxiv.org/abs/2610.23721v1 [DATE]2026-10-17 00:42:10+08:00 [CATEGORIES]cs.CL
The Cognitive Cost of LLM Writing: A Reading-Time Comparison with Human Text [AUTHORS]Kamer Cekini, Elisabetta Biondi, Chiara Boldrini, Andrea Passarella, Marco Conti [ABSTRACT]Although linguistic differences between human-written and Large Language Model (LLM)-generated texts have been extensively studied, their implications for the cognitive effort involved in reading remain less explored. In psycholinguistics, reading time is commonly used as an indirect measure of cognitive processing effort. Building on this perspective, we investigate whether human-written texts and their LLM-generated rewrites exhibit differences in predicted reading time that may reflect differences in processing demands for readers. To this end, we introduce RoBERTa-RT, a model trained on human eye-tracking data to predict word-level reading times. We evaluate the model on both internal and external test sets, construct a paired dataset of 496 human-written articles and their LLM-generated rewrites, and apply RoBERTa-RT to compare their predicted reading times. Our findings show that LLM-generated rewrites tend to be shorter and have slightly lower predicted total reading times, while exhibiting higher predicted reading times per word. These differences are associated with systematic linguistic changes, including longer and less frequent words, greater lexical diversity, and higher contextual predictability. From a cognitive perspective, our findings suggest that LLM-generated writing may redistribute reading demands: reducing the amount of text to process while increasing the predicted processing effort associated with individual words. This highlights the importance of considering not only the linguistic characteristics of machine-generated content, but also its potential implications for human cognitive processing. [LINK]http://arxiv.org/abs/2610.23715v1 [DATE]2026-10-17 00:35:41+08:00 [CATEGORIES]cs.CL
VISTA: Steering LLMs with Latent User Values Inferred from a Few Survey Answers [AUTHORS]Yinuo Xu, Jadie Adams, Brian Hu, Shivani Kumar, David Jurgens, Arslan Basharat [ABSTRACT]Personalizing a language model requires knowing what a person values, which stays largely consistent across tasks. Yet most methods either learn task-specific preferences or place the user's profile in the prompt, where models struggle to act on it. We introduce VISTA (Value-Informed Steering and Targeted Acquisition), which infers a latent profile from users' (full or partial) survey answers, and maps the profile through a hypernetwork to a user-specific steering intervention in a frozen LLM to select answer choices or generate free-text responses. Given the same evidence, VISTA matches or beats in-context profiles on synthetic preferences (Food2D), the World Values Survey (WVS), and PrimeX, without adding prompt tokens. To reduce the cognitive burden of completing full surveys, VISTA also chooses which questions to ask and when to stop: on WVS it stops after 6% of the questionnaire while retaining 94% of the full-history gain, where standard adaptive testing retains at most 42%. The dominant steering direction is an interpretable value axis on which VISTA places unseen users accurately. Our results point toward pluralistic alignment that adapts to individuals at deployable cost. [COMMENTS]8 pages, 4 figures [LINK]http://arxiv.org/abs/2610.23702v1 [DATE]2026-10-17 00:28:57+08:00 [CATEGORIES]cs.CL
IndicDetect: Evaluating Cross-Lingual LLM-Generated Text Detection for Hindi, Telugu, and Tamil [AUTHORS]Bhaskar Ganesh Devalla, Junchao Wu, Nilesh Dokuparthi, Greeshma Yaluru, Tatiana Muniz Rodriguez, Lidia S. Chao, Derek F. Wong [ABSTRACT]The rapid proliferation of LLMs has further heightened the need to develop dependable AI-generated text detection, especially beyond English. Nevertheless, current benchmarks pay little attention to Indic languages and test detectors in idealized settings that do not represent the real world. We present a generalized benchmark for AI-generated text detection in Hindi, Telugu, and Tamil, which we call IndicDetect, designed to assess the robustness of detectors under realistic distribution shifts. IndicDetect comprises highly curated human-written texts matched with LLM-generated counterparts across various domains and generators, and systematically evaluates detectors in the presence of domain shift, generator shift, and adversarial perturbation. Using a single and repeatable evaluation scheme, we evaluate a wide range of statistical and neural detectors. We find substantial robustness failures: supervised neural detectors perform well in-distribution, while training-free methods degrade considerably under unseen generators and adversarial attacks. The severity of these failures varies across languages, with Hindi exhibiting the largest overall degradation under adversarial perturbations. These results highlight that the primary weakness of existing detectors in Indic settings lies in their robustness, not in their peak accuracy. IndicDetect provides standard data splits, an evaluation protocol, and baselines to establish a robust, language-aware foundation for AI-generated text detection in Indic scripts. [LINK]http://arxiv.org/abs/2608.29919v2 [DATE]2026-10-17 00:26:48+08:00 [CATEGORIES]cs.CL
Paved with True Intents: Intent-Aware Training Improves LLM Safety Classification Across Training Regimes [AUTHORS]Jeremias Ferrao, Niclas Müller-Hof, Iustin Sîrbu, Traian Rebedea, Yftah Ziser [ABSTRACT]We argue that safety classifiers should model user intent as an explicit signal between the prompt and the final label. To study this, we introduce AIMS, a human-annotated dataset of 1,724 difficult safety prompts, each paired with an intent description and harm label. We use AIMS to evaluate intent-aware training across supervised fine-tuning, preference learning, reasoning distillation, and reinforcement learning. Despite its size, AIMS enables competitive safety classifiers across training regimes: DPO from model-generated intent errors improves over SFT, and intent-conditioned distillation outperforms reasoning-only distillation in most teacher-student pairs. Most notably, directly rewarding intent faithfulness with GRPO yields the strongest average performance across five external safety benchmarks, while our intent-aware models form the inference latency-F1 Pareto frontier. These results show that faithful intent modeling is a compact, high-quality supervision signal for more robust safety classifiers. [LINK]http://arxiv.org/abs/2606.27210v2 [DATE]2026-10-17 00:09:43+08:00 [CATEGORIES]cs.CL
Beyond the Benchmark: Conditioning, Domains and Failure Modes of Qwen3-TTS [AUTHORS]Séverin Baroudi, Yanis Labrak, Petr Motlicek, Hervé Bredin, Ricard Marxer [ABSTRACT]Qwen3-TTS reports state-of-the-art zero-shot voice cloning, but its technical report leaves open how the model should be conditioned, how it behaves outside read speech, and how it fails. We characterise its two cloning interfaces, in-context learning (ICL) and X-Vector, against Chatterbox, XTTS-v2 and F5-TTS on read, meeting, emotional and air-traffic-control (ATC) speech, using eighteen objective measures, an LLM judge, a spoofing countermeasure and a listening test. Neither interface dominates: ICL clones timbre more closely from correctly transcribed references, X-Vector clones timing, and ICL degrades sharply when a short reference transcript is wrong. Domain explains four times more variance than system choice. Qwen3-TTS leads on intelligibility and pitch register but trails on prosody, and the objective measures rank systems differently from listeners. An off-the-shelf countermeasure flags its clones no more often than genuine recordings, and on ATC radio it fails for every system. Code and audio are released. [LINK]http://arxiv.org/abs/2610.23678v1 [DATE]2026-10-17 00:09:20+08:00 [CATEGORIES]cs.CL
How Reliable Are System-One Decisions in Medicine? A Comparative Study of Jev, Jev-like Models, and LLMs [AUTHORS]Yanrui Du, Danyang Zhao, Sendong Zhao, Wenhao Li, Yibo Gao, Haochun Wang, Zhuoer Xu, Shiwen Cui, Bing Qin [ABSTRACT]Jev has attracted considerable attention for its rapid, generalizable decision-making, yet its medical reliability remains underexplored. To address this gap, our study compares various types of system-one decision models across 15 medical datasets comprising 8,742 instances under a unified protocol. Beyond accuracy, we assess decision stability and the ability to recognize unreliable decisions. Our results show that Jev rivals leading models on primarily text-driven tasks but is less effective on numerically driven tasks, revealing a distinct capability boundary. Under question-format variations, Jev maintains strong decision consistency and small accuracy fluctuations. StartLux-27B exhibits comparable consistency, while D3-27B achieves higher overall accuracy and a lower mean decision reversal rate. But these strengths do not extend to reliable abstention awareness: most evaluated models show limitations in both recognizing missing correct answers and using optional abstention. Moreover, we find that Jev's probability-based answer selection can be improved by changing how decisions are scored: normalizing option-wise binary scores yields higher retained accuracy than choice scoring. Taken together, our findings provide early guidance for improving and applying decision models in medicine. [LINK]http://arxiv.org/abs/2610.18529v2 [DATE]2026-10-17 00:01:04+08:00 [CATEGORIES]cs.CL
2026 Oct 16, Fri
Mitigating Cross-Lingual Cultural Inconsistencies in LLMs via Consensus-Driven Preference Optimisation [AUTHORS]Lucas Resck, Isabelle Augenstein, Anna Korhonen [ABSTRACT]Multilingual large language models (MLLMs) often answer culturally ambiguous questions inconsistently across languages, even for a fixed user persona -- e.g., Shakespeare in English but Cervantes in Spanish for the same British user. To robustly quantify this Cross-lingual Cultural Inconsistency, we introduce Singleton Fleiss's $κ_S$, a metric mathematically resilient to hallucinations. For mitigation, we propose Cross-lingual Cultural Consistent Preference Optimisation (C-3PO), a consensus-driven alignment framework. C-3PO raises $κ_S$ by 0.07--0.10 over unaligned models, above every prompting and representation steering baseline averaged across personas, whilst largely preserving explicit user identities, cultural neutrality and intrinsic cultural knowledge; where it does not, a soft-consensus variant recovers most of the cost. Although trained only on multiple-choice questions, C-3PO also improves consistency in open-ended generation. Empirical evaluations demonstrate this inconsistency is more pronounced for lower-resource languages like Indonesian and Persian. Finally, early decoding of intermediate layers reveals that MLLMs implicitly personalise outputs towards the prompt language's stereotypical culture as forward-pass representations stabilise. [COMMENTS]25 pages, 14 figures, 17 tables [LINK]http://arxiv.org/abs/2605.12515v3 [DATE]2026-10-16 23:46:19+08:00 [CATEGORIES]cs.CL
PresenceBench: An Exploratory Evaluation of Large Language Model Adaptation to User State [AUTHORS]Peña Arroyo Gallego, Mauro Ruiz Moreno, Alejandro Acien, Sergio Romero-Tapiador, Teresa Arroyo-Gallego [ABSTRACT]Presence-aware intelligence is the capacity of an AI system's intelligence layer to adapt appropriately to the functional state of the person it serves (here readiness, flow, and stability) while preserving task quality and safety. We introduce PresenceBench, an instrument for measuring it. PresenceBench evaluates how the intelligence layer uses a supplied state, not how that state is sensed, and breaks the capability into measurable properties. Two are tested here. Receptivity: does the model take up the state and give a better response than with no signal? Specificity: does any improvement depend on the correct state, rather than on extra structured context? A third, adaptation across states, is a required endpoint not measured in this release. A five-arm causal design answers both questions, against each model's own baseline and an equally structured mismatched-state control, scored by a correctness-gated human paired comparison. The exploratory evaluation covered four models across eighteen single-turn, six fixed three-turn and six simulated natural-conversation scenarios. Supplying a synthetic state, raw or translated into a response policy, gave human-preference point estimates above 0.5 for all four models (Claude Sonnet 5, DeepSeek V4 Pro, Gemini 3.6 Flash, GPT-5.6 Terra), with three of eight 95% intervals excluding chance. Specificity was not resolved: no per-model interval excluded chance, and the design cannot yet separate the state values from a generic instruction to adapt. Policy delivery reduced priced cost and output tokens against each model's baseline; these are resource-use changes, not efficiency gains, because correctness non-inferiority was not established. Safety-floor failures were rare (13 of 360 cases), with a non-significant excess under policy delivery. We report the instrument, the frozen results, and the controls a confirmatory study needs. [COMMENTS]33 pages, 6 figures, 13 tables [LINK]http://arxiv.org/abs/2610.23625v1 [DATE]2026-10-16 23:33:04+08:00 [CATEGORIES]cs.CL
The Chameleon's Limit: Investigating Persona Collapse and Homogenization in Large Language Models [AUTHORS]Yunze Xiao, Vivienne J. Zhang, Chenghao Yang, Weihao Xuan, Jen-tse Huang [ABSTRACT]Large language models (LLMs) have powered multi-agent simulations such as survey respondents or application users. However, LLMs lack the population-level diversity required by these simulations---responses from LLMs with detailed human profiles cluster into classes, unlike human responses that uniformly fill the entire space. In this paper, we formalize this as Persona Collapse and quantify its population-level structure with three diagnostic metrics: Coverage (how much of the persona space is occupied), Uniformity (how evenly agents spread across it), and Complexity (how rich the resulting behaviors are). Across eight LLMs on personality (BFI-44), moral reasoning, and self-introduction, persona collapse is universal but task-dependent: CoSER-Llama-8B collapses the most in personality while being the most diverse in moral reasoning. Moreover, sampling more personas cannot mitigate the collapse. We further show a fidelity--diversity tradeoff: the models with the highest per-persona fidelity produce the most polarized populations. Finally, persona-targeted fine-tuning further narrows the population instead of diversifying it. [LINK]http://arxiv.org/abs/2604.24698v2 [DATE]2026-10-16 23:30:51+08:00 [CATEGORIES]cs.CL
SIT-Eval: Situation-Based Interactive Turn-by-Turn Evaluation of Conversational Agents [AUTHORS]Ravil Mussabayev, Saveliy Zavertyaev, Dmitriy Parpulov, Andrey Savchenko [ABSTRACT]The rapid proliferation of conversational AI agents necessitates evaluation methods that assess performance across diverse dialogue scenarios, utilizing multiple situational criteria, as appropriate behavior depends on context and encompasses various aspects. Existing dialogue benchmarks typically use static natural conversations, which underrepresent rare situations---a critical safety gap, as low-probability events become frequent failures at production scale. However, limited sample sizes force a choice: either maintain natural dialogues for realism or turn to adaptive situation sampling for comprehensive yet statistically reliable coverage. We introduce SIT-Eval, an adjustable framework for evaluating situation-based conversational agents. We formalize an agent's desired behavior as situational rules: if situation is P, follow rules Q. At each conversation turn, an LLM-interlocutor enacts a uniformly sampled situation. A context-aware LLM-judge leverages all conditionally applicable rules as criteria to evaluate the agent's reply for that situation. The judge is validated using human-verified synthetic golden dialogues. Benchmarking 8 LLMs on 1292 conversation turns reveals failure modes hidden by aggregate metrics: models with similar overall scores vary widely in their adherence to individual rules. All models struggle most with ambiguity clarification and Socratic questioning. [LINK]http://arxiv.org/abs/2610.23612v1 [DATE]2026-10-16 23:24:02+08:00 [CATEGORIES]cs.CL
Analyzing Occupational Gender Bias across the LLM Training Lifecycle [AUTHORS]Manthan Gadhia, Alexander Hoyle, Karolina Stańczak [ABSTRACT]Post-training is widely assumed to reduce the social biases language models inherit from pretraining, yet direct evidence for this assumption remains mixed. In this paper, we examine how gender disparities evolve across the language modeling pipeline. In particular, we study occupational gender bias in OLMo 3, a fully open language model family, across its post-training stages by prompting each checkpoint to generate open-ended narratives about synthetic personas given their occupation and education. We consider two settings, a gender-given setting in which the profile states the persona's gender explicitly and a gender-assumed setting in which it must be inferred by the model. We find that when gender is stated, post-training increases semantic diversity for male-coded profiles, whereas for female-coded profiles the increase is smaller after SFT and not significant after DPO and RLVR. When gender is assumed by the model, post-training affects narratives about female and male personas similarly. The model's own gender assumptions, however, show a consistent female-preferential skew relative to United States labor-force demographics, except for heavily male-dominated occupations. Every successive post-training stage monotonically increases the odds of attributing a female label over the base model. A provenance analysis of the supervised fine-tuning corpus finds only a modest 7.7% female-token surplus, too small to explain the size of these downstream effects. The project code is available under: https://github.com/manthangadhia/llm-lifecycle-bias. [COMMENTS]Submitted to 7th Workshop on Gender Bias in Natural Language Processing @AACL-IJCNLP 2026 [LINK]http://arxiv.org/abs/2610.23587v1 [DATE]2026-10-16 23:07:02+08:00 [CATEGORIES]cs.CL
MARS: Memory-Aware Routing for Session-Level LLM Assistants [AUTHORS]Minghan Wang, Jinhang Zuo, Zhongxiang Dai, Zhiyong Wang, Fang Kong [ABSTRACT]Long-term LLM assistants must repeatedly decide what prior context to expose and which model should answer as an interaction evolves. Existing memory systems and model routers usually make these decisions separately, although memory construction changes the evidence, distractors, and difficulty of the resulting input. We introduce MARS, a memory-aware routing framework that uses session state as a shared control interface between memory construction and model selection. A Query and Topic Manager (QTM) estimates compact decision-relevant state that guides a fixed state-conditioned memory policy and an online router, enabling each turn to receive an appropriate memory view and a suitable model under a quality-memory-cost trade-off. We evaluate MARS on RealMem and MultiChallenge with 16 system configurations each, covering long-history memory-driven interaction and bounded sessions with structured context interference. On RealMem, MARS preserves 98.2% of full-context GPT-4o's QA while using only 14.5% of its generation API cost, including both generation and controller calls for MARS, and achieves the strongest measured memory-recall and retrieval scores among compared systems. On MultiChallenge, MARS trails the strongest static system overall but approaches its performance at lower recorded cost on a mechanism-focused subset. Together, these results support a favorable quality-memory-cost trade-off for long-history, memory-dependent interaction. [COMMENTS]32 pages, including references and appendices [LINK]http://arxiv.org/abs/2610.23583v1 [DATE]2026-10-16 23:04:22+08:00 [CATEGORIES]cs.CL
PragReST: Self-Reinforcing Counterfactual Reasoning for Pragmatic Language Understanding [AUTHORS]Jihyung Park, Minchao Huang, Leqi Liu, Elias Stengel-Eskin [ABSTRACT]Natural language understanding often depends on meanings that are implied rather than explicitly stated, requiring pragmatic reasoning. Despite strong performance on math and logical reasoning, large language models (LLMs) still struggle with making pragmatic inferences, often choosing literal interpretations. To improve LLM pragmatic reasoning, we introduce PragReST, a self-supervised framework that constructs pragmatic QA data, generates counterfactual reasoning traces, and trains models to internalize them through supervised fine-tuning and reinforcement learning, without human-labeled training data or distillation from a stronger teacher. Across four pragmatic benchmarks (PragMega, Ludwig, MetoQA, and AltPrag), PragReST improves over backbone models, task-specific pragmatic tuning baselines, and non-counterfactual variants of the same pipeline. On accuracy-based benchmarks, PragReST improves over the instruct backbone by 5.37 and 5.50% (absolute) for Qwen3-8B and Qwen3-14B, respectively. Our error analysis and ablations underscore the importance of counterfactual reasoning: PragReST primarily reduces errors caused by failures to contrast observed utterances with plausible alternatives, and removing counterfactual reasoning substantially reduces performance. Moreover, our training preserves out-of-domain performance on general-knowledge and mathematical reasoning benchmarks. [COMMENTS]EMNLP Camera-Ready. First two authors contributed equally. Code and models: https://github.com/jihyung803/PragReST [LINK]http://arxiv.org/abs/2606.18624v2 [DATE]2026-10-16 22:54:52+08:00 [CATEGORIES]cs.CL
Towards Reliable Truth-Aligned Uncertainty Estimation in Large Language Models [AUTHORS]Ponhvoan Srey, Quang Minh Nguyen, Xiaobao Wu, Anh Tuan Luu [ABSTRACT]Uncertainty estimation (UE) aims to quantify uncertainty in large language model (LLM) outputs to flag potential hallucinations and support reliable deployment. However, UE metrics often exhibit unstable performance across models and datasets, limiting their practical utility. We formalise this phenomenon as proxy failure, arising because UE metrics are derived from model behaviour rather than being explicitly grounded in factual correctness. We show that when mutual information between a UE score and correctness is low, the score's discriminability must be close to chance. To address this, we propose Truth AnChoring (TAC), a lightweight post-hoc calibration method that maps raw UE scores to truth-aligned correctness probabilities. Even with noisy and few-shot supervision, TAC continues to improve calibration, providing a practical calibration protocol. Our findings highlight the limitations of treating heuristic UE metrics as direct indicators of uncertainty about factual correctness, and position our TAC as a principled step toward more reliable uncertainty estimation for LLMs. The code repository is available at https://github.com/ponhvoan/TruthAnchor/. [COMMENTS]Accepted to EMNLP Findings 2026 [LINK]http://arxiv.org/abs/2604.00445v2 [DATE]2026-10-16 22:40:44+08:00 [CATEGORIES]cs.CL
DominoTree: Conditional Tree-Structured Drafting with Domino for Speculative Decoding [AUTHORS]Saw S. Lin, Jyh-Shing Roger Jang [ABSTRACT]Speculative decoding speeds up large language model inference: a small drafter proposes several tokens, and the target model verifies them in one parallel pass. Drafting a tree lets the target verify many alternative continuations at once. DDTree builds such trees from the block-diffusion drafter DFlash, but scores each position independently of the tokens earlier in its branch; Domino conditions on those tokens but drafts only a single chain. DominoTree combines the two: a training-free best-first tree scored by Domino's path-dependent causal correction, restricted to top-M candidates. In a research harness at temperature zero, it averages 5.33x speedup over autoregressive (AR) decoding on Qwen3-4B and 6.89x on Qwen3-8B, peaking at 8.77x; across temperatures it is 20-26% faster than Domino on Qwen3-4B and 27-36% on Qwen3-8B, and 11-20% faster than the stronger of DDTree and its cost-aware variant CaDDTree. A controlled decomposition attributes +10.1% (Qwen3-4B) and +10.7% (Qwen3-8B) accepted length to the correction, and +4.7% and +4.4% to conditioning on the realized path. Served in SGLang on one RTX 5090, it reaches 4.11x and 4.83x AR at batch size 1, 1.17x and 1.09x the speed of Domino's chain, and up to 1.71x AR on 8K-token HELMET summarization, 14-21% above the chain. The lead holds at 2 concurrent requests but reverses from 8 onward, where the chain sustains more goodput despite DominoTree's 28% higher acceptance. DominoTree accepts more tokens than the chain in every serving setting. On the hybrid Gated-DeltaNet Qwen3.6-27B, correct output requires tree-aware recurrent verification; with it, DominoTree matches the chain at batch size 1 (4.58x versus 4.54x AR) and reaches up to 2.11x AR at long context, 15% above the chain. [COMMENTS]31 pages. Submitted to IEEE Access. v4: adds Qwen3.6-27B results, single-GPU serving re-runs, fused tree-builder kernels and tuned budgets; code and data at https://github.com/slin-zhq/Domino-Tree [LINK]http://arxiv.org/abs/2607.08642v4 [DATE]2026-10-16 22:29:03+08:00 [CATEGORIES]cs.CL
EduWild: How Teacher-Configured Chatbots Fail in Real-World Use and How to Fix Them [AUTHORS]Ziqi Liu, Ruiwei Xiao, Yuxuan Li, John Stamper [ABSTRACT]Teachers increasingly configure educational chatbots through natural-language instructions, but how these bots fail in real-world use remains poorly understood. We introduce EduWild, a dataset of 1,726 conversations with 217 teacher-configured chatbots, each paired with the instructions and configuration behind it, including 900 conversations from the intended end users of 35 deployed chatbots. We develop a 29-code taxonomy covering factual, workflow, diagnostic, and pedagogical failures and apply it using a human-calibrated LLM judge. The most frequent failures are lapses in following teacher intent --- bots skip required workflow steps (53.8% of end-user conversations), omit expectation setup (33.4%), and overload learners with information or questions (37.1%). Guided by these findings, we design a prompt-based harness targeting the six most frequent failure modes. In a study with 200 college students across ten bots, the harness reduces the mean number of target failure codes per conversation from 1.406 to 0.489, a 65.2% relative reduction with bots weighted equally. These findings show how real-world failure analysis can guide practical improvements to educational chatbots. [LINK]http://arxiv.org/abs/2610.23530v1 [DATE]2026-10-16 22:28:06+08:00 [CATEGORIES]cs.CL
Truth in Contrast: Calibrated Factual Evidence from Language Model Likelihoods [AUTHORS]Ponhvoan Srey, Jason Lucas, Pittawat Taveekitworachai, Junyoung Lee, Anh Tuan Luu [ABSTRACT]Large Language Models (LLMs) encode substantial factual knowledge. Yet, estimating confidence in factual claims reliably remains challenging. We study likelihood-based factual confidence from the complementary perspectives of discrimination and calibration across multilingual factual statements. Specifically, we instantiate contrastive generative likelihoods as Contrastive Factual Evidence (CFE), which compares the likelihood of the same claim under competing factual and non-factual conditions, and derive its probabilistic counterpart, the Contrastive Factual Posterior (CFP). We show analytically that CFP calibration depends on the differential error between the two surrogate conditionals, allowing errors shared across them to cancel and yield tighter calibration bounds. Empirically, CFP substantially improves calibration while maintaining factual discrimination comparable to its strongest competitor. Nonetheless, discrimination is uniformly weak across methods in lower-resource languages, which simultaneously undermines the utility of the measured calibration. To this end, we interpret a likelihood geometry, within which an evidence-bearing direction is induced, along with an orthogonal evidence-invariant direction. In our investigation, we selectively align target-language claims with stronger English evidence only along the factual-evidence direction, improving discrimination and open-ended factual generation without target-language labels. [LINK]http://arxiv.org/abs/2610.23527v1 [DATE]2026-10-16 22:27:04+08:00 [CATEGORIES]cs.CL
Perceive, Solve, Reason: Progressive Capability Formation in Time Series-LLMs [AUTHORS]Xin Qiu, Junlong Tong, Wei Zhang, Xiaoyu Shen [ABSTRACT]Large language models (LLMs) can now perceive, describe, and reason about images and audio, yet time series understanding remains underestimated: most work uses LLMs as forecasters and classifiers, reasoning studies are few and narrow, and no suite trains and evaluates the full range of capabilities. In this paper, we build the first such suite, with 337K training and 44K evaluation samples across three progressive capabilities, Perceive, Solve, and Reason, and use it to ask how time series should enter an LLM that must master all three, and in what order. We find that a dedicated encoder is essential: with one, a 4B model surpasses GPT-4o on every perception attribute we test, whereas textualizing the series costs 8--10 points in reasoning accuracy. The capabilities then build on one another: learning them in order beats mixing the same data and rivals reasoning-only training with over three times the reasoning data, but only when an encoder is used. Analyses further reveal that capabilities are transferred through distinct network depths rather than simply stacked, and that a stage converged on its own benchmark is not necessarily ready for the next. Distilled into a simple recipe, these findings yield 4B TS-LLMs that, without any target-domain data, outperform four recent 7B TS-LLMs on seven of eight out-of-domain metrics, by up to 11 points. We therefore advocate treating time series as a first-class modality for LLMs, read through dedicated encoders and taught to perceive before they reason. [LINK]http://arxiv.org/abs/2610.23521v1 [DATE]2026-10-16 22:23:23+08:00 [CATEGORIES]cs.CL
Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching [AUTHORS]Roy Miles, Aysim Toker, Andreea-Maria Oncescu, Jiankang Deng, Ismail Elezi [ABSTRACT]Reasoning with large language models often benefits from generating multiple chains-of-thought, but existing aggregation strategies are typically trajectory-level (e.g., selecting the best trace or voting on the final answer), discarding useful intermediate work from partial or "nearly correct" attempts. We propose Stitching Noisy Diffusion Thoughts, a self-consistency framework that turns cheap diffusion-sampled reasoning into a reusable pool of step-level candidates. Given a problem, we (i) sample many diverse, low-cost reasoning trajectories using a masked diffusion language model, (ii) score every intermediate step with an off-the-shelf process reward model (PRM), and (iii) stitch these highest-quality steps across trajectories into a composite rationale. This rationale is then used to recompute only the final answer. This modular pipeline separates exploration (diffusion) from evaluation and solution synthesis, avoiding monolithic unified hybrids while preserving broad search. Across math reasoning benchmarks, we find that step-level recombination is most beneficial on harder problems, and ablations highlight the importance of the final solver in converting stitched but imperfect rationales into accurate answers. Using low-confidence diffusion sampling with parallel, independent rollouts, our training-free framework improves average accuracy by up to 23.8% across six math and coding tasks. At the same time, it achieves up to a 1.8x latency reduction relative to both traditional diffusion models (e.g., Dream, LLaDA) and unified architectures (e.g., TiDAR). The code is available https://github.com/roymiles/diffusion-stitching [COMMENTS]NeurIPS 2026. Code available at https://github.com/roymiles/diffusion-stitching [LINK]http://arxiv.org/abs/2602.22871v3 [DATE]2026-10-16 21:55:50+08:00 [CATEGORIES]cs.CL
HydraLLM: Adaptive Width Slicing for Elastic Text and Vision-Language Models [AUTHORS]Oscar Kemme, Janek Haberer, Ali Hojjat, Olaf Landsiedel [ABSTRACT]LLMs and LVLMs are increasingly deployed across heterogeneous hardware, ranging from data-center GPUs to resource-constrained edge devices. Their fixed-size architectures limit the ability to reuse a single model across deployment settings with different compute, latency, and model-size targets. We introduce HydraLLM, an elastic MQA-based architecture. Its number of active attention heads determines the effective widths of the embeddings, attention projections, feed-forward layers, and the output head, forming a nested family of subnetworks within a single parameter set. We introduce a static-to-elastic conversion pipeline that, starting from a pre-trained model checkpoint, includes importance-driven weight reordering and stochastic subnetwork training with knowledge distillation. We extend this principle to multimodal inference through an elastic projection module whose output width follows the active text width. HydraLLM covers a wider range of active parameter configurations than FFN-only elastic architectures. With 75% of heads active, HydraLLM achieves 2.2 percentage points higher TriviaQA accuracy than a state-of-the-art nested baseline configuration which has 17.25% more parameters. With 50% of heads active, our multimodal model achieves a CLIPScore comparable to that of the 7$\times$ larger Gemma 3 4B model. With only 25% of heads active, HydraLLM increases token-generation throughput by 75.3% over the full-width configuration. [COMMENTS]Accepted at the BabyLM Workshop at EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23455v1 [DATE]2026-10-16 21:38:11+08:00 [CATEGORIES]cs.CLcs.LG
"You" Is Not Toxic: How LLM Labelers Rewrite the Web's Conversation Laws [AUTHORS]Liang Yue, Eric A. Vance [ABSTRACT]What the Web community believes about toxic conversations, such as that toxicity rises in longer and deeper threads, was measured mostly with Google's Perspective API, which is shutting down; large language models (LLMs) will make the next measurements. We find that LLM labelers mistake second-person address for toxicity, and that this one bias manufactures laws that human labels do not show. On 1.25 million news comments with crowd labels, Claude Haiku flags 12% of the non-toxic comments that contain no "you" and 51% of those with four or more; a single crowd rater flags 6% and 10%. Deeper replies address someone more often, so under Haiku toxicity rises with reply depth (+0.93 percentage points per level), while under the crowd's labels it slightly declines (-0.12). Rewriting "you" in the third person removes two thirds of the false rise; flagging only as many comments as the crowd does not remove it, and showing Haiku the parent comment enlarges it. All seven LLMs we test over-flag "you", and at the crowd's flag rate, without the parent, any false rise of the open ones vanishes at a fixed number of second-person words; Haiku also over-flags "you" on Reddit, Gab and Wikipedia talk pages. Agreement with human labels does not reveal the bias: we prove that a fixed Cohen's kappa leaves a law nearly unconstrained. We therefore certify the law instead of the labeler. Threadsafe uses the labeler's scores to choose which comments human raters label and to predict the others, and corrects the predictions with the human labels; linear laws are estimated without bias whatever the labeler does. In simulated audits its intervals for the laws keep nominal coverage from 250 labels on, where uniform-sample methods such as PPI++ fall short, and at 1,000 to 8,000 labels are 0.75 times as wide as those of design-based supervised learning (DSL); a thousand human labels rule out Haiku's rising law. [COMMENTS]43 pages, 4 figures, 44 tables. Labels, prompts and code: https://github.com/liyu2195/you-is-not-toxic [LINK]http://arxiv.org/abs/2610.23446v1 [DATE]2026-10-16 21:29:30+08:00 [CATEGORIES]cs.CL
Atomic Consistency Preference Optimization for Long-Form Question Answering [AUTHORS]Jingfeng Chen, Raghuveer Thirukovalluru, Junlin Wang, Kaiwei Luo, Bhuwan Dhingra [ABSTRACT]Large Language Models (LLMs) often produce factoid hallucinations - plausible yet incorrect answers. A common mitigation strategy is model alignment, which improves factual accuracy by training on curated (factual, non-factual) pairs. However, this approach often relies on a stronger model (e.g., GPT-4) or an external knowledge base to assess factual correctness that may not always be accessible. Addressing this, we propose Atomic Consistency Preference Optimization (ACPO), a self-supervised preference-tuning method that enhances factual accuracy without external supervision. ACPO leverages atomic consistency signals (i.e., the agreement of individual facts across multiple stochastic responses) to identify high- and low-quality data pairs for model alignment. Despite being fully self-supervised, ACPO outperforms the strong supervised alignment baseline by 1.95 points averaged across Phi-3 and Llama3 on the LongFact and BioGen datasets, demonstrating its effectiveness in improving factual reliability without relying on external models or knowledge bases. [COMMENTS]13 pages, 1 figure [LINK]http://arxiv.org/abs/2505.09039v3 [DATE]2026-10-16 21:28:17+08:00 [CATEGORIES]cs.CL
ScoutRoute: Efficient Mixture-of-Experts Routing Using Internal Language Model Signals [AUTHORS]Rajdeep Pandhare, Delzaan Bharucha [ABSTRACT]Efficient expert routing is a fundamental challenge in Mixture-of-Experts (MoE) systems, particularly for heterogeneous collections of independently trained Large Language Mod els (LLMs). Existing routing approaches typically rely on jointly trained gating networks, prompt-based predictors, or computationally expensive multi-model inference, limiting their flexibility and efficiency. This paper presents ScoutRoute, a lightweight post-hoc routing framework that predicts expert competence using internal language model signals extracted from a single forward pass through an inexpensive scout model. Hidden-state representations are used to estimate cross-model competence through lightweight logistic probes, while logit-derived confidence measures are incorporated within a domain-aware routing strategy for efficient expert selec tion without modifying the underlying language models. ScoutRoute isevaluatedon fiveopen-weightlanguagemodels across fourbenchmark datasets covering mathematical reasoning, code generation, commonsense reasoning, and multidis ciplinary knowledge. Experimental results show that ScoutRoute consistently outperforms individual routing signals while requiring only a single scout-model inference. The proposed competence probes also achieve superior calibration compared with conventional confidence measures, demonstrating that internal language model representations provide transferable information for accurate and efficient expert routing across heterogeneous language models [COMMENTS]10 pages, 1 figure, 6 tables. Published in Advances in Artificial Intelligence and Machine Learning, 2026 [LINK]http://arxiv.org/abs/2610.23443v1 [DATE]2026-10-16 21:25:53+08:00 [CATEGORIES]cs.CL
Cross-Layer Attention Probing for Fine-Grained Hallucination Detection [AUTHORS]Malavika Suresh, Rahaf Aljundi, Ikechukwu Nkisi-Orji, Nirmalie Wiratunga [ABSTRACT]With the large-scale adoption of Large Language Models (LLMs) in various applications, there is a growing reliability concern due to their tendency to generate inaccurate text, i.e. hallucinations. In this work, we propose Cross-Layer Attention Probing (CLAP), a novel activation probing technique for hallucination detection, which processes the LLM activations across the entire residual stream as a joint sequence. Our empirical evaluations using five LLMs and three tasks show that CLAP improves hallucination detection compared to baselines on both greedy decoded responses as well as responses sampled at higher temperatures, thus enabling fine-grained detection, i.e. the ability to disambiguate hallucinations and non-hallucinations among different sampled responses to a given prompt. This allows us to propose a detect-then-mitigate strategy using CLAP to reduce hallucinations and improve LLM reliability compared to direct mitigation approaches. Finally, we show that CLAP maintains high reliability even when applied out-of-distribution. Code available at: https://github.com/itsmemala/CLAP [COMMENTS]Published at the TRUST-AI workshop, ECAI 2025 [LINK]http://arxiv.org/abs/2509.09700v2 [DATE]2026-10-16 21:07:39+08:00 [CATEGORIES]cs.CL
On the Tip of the Tongue: First-Token Cueing as a Behavioral Probe of Recall Failures in LLMs [AUTHORS]Jaewon Sok, Jewon Yeom, Seonghyeon Park, Jeongjae Park, Hwiyeong Lee, Taesup Kim [ABSTRACT]Large language models often answer factual questions incorrectly in closed-book settings, and such errors are usually attributed to missing knowledge. Yet models often encode answers they fail to produce. We show that many closed-book errors are first-token recall failures: the model commits to a wrong first token, but once given the right one it completes the answer and often recognizes it as correct. We study this with First-Token Cueing, a behavioral probe that prefills the response with the first token of the gold answer. Model uncertainty peaks at the first answer token, where it already predicts correctness, and among gold prefixes of any length the first token alone accounts for the largest share of recovered errors. Across 16 open models, this one token corrects between a fifth and a half of the errors on multi-token answers in English QA and improves accuracy in all seven languages of a multilingual benchmark. Most of the errors it corrects are also solved by the model's own sampled reasoning, a criterion for encoded knowledge, and models recognize correct starts they do not produce: they continue a correct start more confidently than an incorrect start of the same answer type and prior probability, yet rarely find it without the gold answer. These results locate a recoverable class of factual errors at the first decoding step, where inference-time methods may gain the most. [LINK]http://arxiv.org/abs/2610.23400v1 [DATE]2026-10-16 20:55:36+08:00 [CATEGORIES]cs.CL
FSPO: Policy-Consistent Risk and Pareto-Feasible Control for Budgeted LLM RL Post-Training [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Daren Zha [ABSTRACT]Adaptive LLM reinforcement-learning post-training changes multiple training actuators online, including rollout temperature, group size, clipping, KL regularization, verifier allocation, and update budget. Three coupled issues remain unresolved. A future-risk model trained from behavior trajectories need not estimate the risk induced by the controller that will be deployed; a score calibrated on logged state-action pairs can become miscalibrated after selective action choice; and independent per-resource minimum costs do not in general certify a feasible multi-resource continuation. We introduce FSPO, a feedback-state controller for budgeted LLM RL post-training that addresses these issues jointly. FSPO learns a policy-consistent risk-to-go model whose Bellman target follows the same frozen controller used for future decisions, together with a long-horizon utility model. Decision-conditioned trajectory calibration (DCTC) calibrates risk on cross-fitted trajectories generated by actions selected by provisional controllers. A Pareto resource continuation certificate (PRCC) admits an action only when a non-dominated cumulative reservation remains feasible over the residual horizon. Under a matched GRPO resource envelope, FSPO reaches 66.11% held-out and 59.43% OOD accuracy, compared with 64.47% and 57.03% for PB2, the strongest evaluated adaptive baseline. Three paired training seeds give gains of +2.42 and +3.19 percentage points over the contextual bandit on held-out and OOD evaluation. Under high behavior-deployment mismatch, policy-consistent risk lowers selected-decision ECE from 0.108 to 0.053; DCTC lowers it from 0.039 to 0.022 at matched acceptance; PRCC removes false-feasible admissions on an 18-action catalog ($0.197\rightarrow0.000$); and enabling all three components reduces trajectory failure from 0.181 to 0.083 in a factorial ablation. [COMMENTS]40 pages, 4 figures [LINK]http://arxiv.org/abs/2610.02828v2 [DATE]2026-10-16 20:44:10+08:00 [CATEGORIES]cs.CLcs.LG
Two Minds Better Than One: Collaborative Reward Modeling for LLM Alignment [AUTHORS]Jiazheng Zhang, Wenqing Jing, Zizhuo Zhang, Zhiheng Xi, Shihan Dou, Rongxiang Weng, Jiahuan Li, Jingang Wang, Mingxu Chai, Shibo Hong, Yuming Yang, Chenhao Huang, Tao Gui, Qi Zhang, Xuanjing Huang [ABSTRACT]Reward models (RMs) are central to aligning large language models (LLMs) with human values, yet noisy preference feedback can induce reward misgeneralization through spurious correlations and overfitting. We characterize the training dynamics of preference pairs and uncover a distinctive optimization signature of noise: persistently high losses, sharp temporal fluctuations, and irregular gradient updates. Our analyses further show that excluding such pairs improves downstream alignment. In practice, however, reliable preferences are not known a priori, and self-filtering based on a model's own training signals can amplify confirmation bias. To address this challenge, we introduce Collaborative Reward Modeling (CRM), an online framework for identifying reliable preferences. CRM couples cross-model peer review with curriculum learning: two RMs evaluate each other's sample selections, while curriculum keeps their learning progress synchronized. Across extensive experiments, CRM improves RM generalization by up to 9.94 points on RewardBench under 40% preference noise. CRM further generalizes to implicit-reward alignment, offering a robust and broadly applicable approach to learning from imperfect feedback. [LINK]http://arxiv.org/abs/2505.10597v3 [DATE]2026-10-16 20:42:25+08:00 [CATEGORIES]cs.LGcs.CL
Dependency Reasoning for Aspect Sentiment Quadruple Prediction [AUTHORS]Shima Esfandyari, Seywan Soleymanbaigi, Fardin Akhlaghian Tab, Fatemeh Daneshfar, Mourad Oussalah [ABSTRACT]Aspect Sentiment Quadruple Prediction (ASQP) is fundamentally a relational reasoning problem: identifying aspect-opinion-category-sentiment quadruples requires modeling structured dependencies, not merely sequential co-occurrence. Existing LLM-based approaches incorporate syntactic information indirectly, via retrieval-based demonstration selection or multi-stage training, rather than as unified, single-pass generation-time context. We present DRIFT, which linearizes dependency parse trees and injects them directly into LLM prompts as explicit relational context, enabling the model to reason over aspect-opinion dependencies rather than infer them from token proximity alone. DRIFT further integrates retrieval-augmented in-context learning and a syntax-aware LLM-based adjudication mechanism that complements exact-match evaluation under span-boundary variation. On SemEval Rest15, Rest16, and ACOS, DRIFT achieves the strongest performance among compared generative ASQP methods, reaching 54.47% and 64.10% F1 on Rest15 and Rest16, with consistent gains across short- and medium-distance aspect-opinion configurations; adjudication further shows exact-match metrics systematically underestimate semantically correct predictions. [LINK]http://arxiv.org/abs/2610.14873v2 [DATE]2026-10-16 20:03:17+08:00 [CATEGORIES]cs.CLcs.LG
Language Models Learn to Improve Language Models [AUTHORS]Minhua Lin, Zijun Wang, Zewen Liu, Zhaotian Weng, Bing He, Zhan Shi, Benoit Dumoulin, Xiang Zhang, Suhang Wang, Hanqing Lu [ABSTRACT]Large language models (LLMs) are increasingly used to automate AI research, including language-model (LM) development, offering a potential path toward recursive self-improvement. However, turning experimental feedback into a learned ability to propose effective improvements across models faces two challenges. First, training such an ability with real feedback is expensive: reinforcement learning (RL) requires evaluating tens of thousands of proposals, each potentially requiring a full model-training run. Even at ten GPU-hours per run, reward collection alone would consume hundreds of thousands of GPU-hours, motivating a cheaper surrogate. Second, surrogate feedback can become unreliable as policy optimization favors overestimated proposals when the proposal distribution shifts. To address these challenges, we propose LM2LM, which trains LLMs with RL to generate effective LM improvement proposals. To provide scalable feedback, we use a research world model (RWM), a frozen LLM that predicts proposal gains from real experimental records provided in context, as the reward model for RL. To maintain feedback reliability, we periodically execute current-policy proposals ranked highly by the RWM and add their measured outcomes to its context. Across eight autoregressive and diffusion LM development environments, real-execution evaluations on held-out tasks show that LM2LM enables Qwen3.8-27B to outperform GPT-6-sol, even when GPT-6-sol is given 300 real experimental records in context, while requiring 33% lower data-generation cost. LM2LM also generalizes to a model family unseen during training. [LINK]http://arxiv.org/abs/2610.23344v1 [DATE]2026-10-16 19:51:00+08:00 [CATEGORIES]cs.CL
Robust Function Calling for LLMAgents Under Interface Drift [AUTHORS]Aniruddha Chattopadhyay, Kanishka Halder, Partha Sarathi Roy [ABSTRACT]Tool-augmented language-model agents rely on interface text: tool descriptions, argument schemas, and the candidate tool list shown at inference time. These interfaces often change after deployment even when the underlying tool semantics remain fixed, creating a practical reliability problem for function-calling systems. We study this failure mode as interface drift and evaluate whether a lightweight, training-free repair layer can recover robustness without retraining the base model. Our method, SchemaShield-Lite, combines schema validation, candidate-set retry for missing tool names, and one-shot constrained repair. On Qwen3.5-9B, it raises DICE-300 accuracy from 0.717 to 0.813 and BFCL-200 accuracy from 0.580 to 0.805, recovering 19/22 and 33/45 drift-induced failures with no observed repair harms. Naive retry does not explain the gains. Ablations show that candidate-set drift, especially distractor tools, accounts for most observed damage, while description and schema edits alone are less harmful. The results position validation-and-repair as a practical robustness layer for deployed LLM-agent tool use. [COMMENTS]16 pages, 1 figure, 6 tables. Peer-reviewed and accepted at NGEN-AI 2026 (International Conference on Next Generation AI Systems). This is the pre-review submitted version [LINK]http://arxiv.org/abs/2610.23342v1 [DATE]2026-10-16 19:49:37+08:00 [CATEGORIES]cs.CL
CascadeLedger: Counterfactual Auditing of Information and Resource Integrity in LLM Cascades [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Daren Zha [ABSTRACT]We introduce CascadeLedger, a counterfactual execution-audit framework for budgeted language-model cascades. The protocol enforces call-before-read semantics, binds successful reads to service and cache identities, records availability and component-wise budget transitions, and freezes the event trace before any target or task scorer is joined. The central object is aggregate-equivalent trace validity: two executions can share a final answer, declared resource projection, and service histogram while differing in whether their routing decisions used only legitimately observable information. We complement single-trace invariants with counterfactual prefix non-interference (CPNI), which couples a visible prefix and policy randomness while changing only a declared hidden family. The counterfactual influence frontier (CIF) localizes the earliest violating prefix and a minimal hidden witness; projection-preserving adversarial mutation (PPAM) searches for invalid traces that remain acceptable to a non-counterfactual checker. Across frozen-cache, independently generated, and serving evaluations, the two-layer audit reports a clean false-positive rate of 0.0000, detects 0.9983 of serialized faults and 0.9982 of unlogged hidden-read faults, and changes 0.3403 of corrupted quality-resource conclusions after deterministic recomputation. The online recorder and offline CPNI audit have separate overhead measurements. Learned routing and semantic verification provide downstream compatibility studies under the same execution contract. [COMMENTS]38 pages, 6 figures [LINK]http://arxiv.org/abs/2610.20668v2 [DATE]2026-10-16 19:25:48+08:00 [CATEGORIES]cs.CLcs.LG
Self-Reflective Chain-of-Thought (SRCoT): Evidence-Based Text Generation via a Closed-Loop Retrieve--Reflect--Revise Pipeline on Edge LLMs [AUTHORS]Fabian Broszio, Michael Hoffmann, Adrian Paschke [ABSTRACT]Large language models (LLMs) can generate fluent outputs that lack source support or conflict with local operational constraints. We introduce Self-Reflective Chain-of-Thought (SRCoT), a modular architecture that organizes generation as a closed Retrieve--Reflect--Revise loop. SRCoT separates environment state, persistent knowledge, accumulated experience, and short-term memory into provenance-labeled stores. BM25 retrieval supplies local context; subsequent passes expose intermediate drafts, missing-information markers, follow-up queries, and optional claim--evidence mappings. These artifacts inform retrieval, revision, clarification, or abstention decisions. We examine the architecture through qualitative, trace-based case studies in two fully synthetic settings: emergency medical services and Munich-Sim, an emergency-operations simulation. Local prototype runs on an Apple M4-family laptop with 128 GB memory illustrate source-labeled documentation, explicit gap marking, and structured dispatch-style outputs. The traces also reveal provenance-assignment errors, incomplete evidence coverage, and unsafe assumptions that remain possible despite prompting constraints. These cases demonstrate report-level inspectability and local execution, not comparative superiority, hallucination reduction, clinical validity, or operational safety. The architecture separates retrieval and control from the model interface, but behavior across different models remains to be evaluated. SRCoT is a research proof of concept requiring human review, with its principal contribution lying in an inspectable control procedure rather than a new model. [COMMENTS]28 pages, 2 figures. Extended technical account; a related brief report based on overlapping work has been submitted to Frontiers in Disaster and Emergency Medicine. The two manuscripts do not represent independent empirical studies [LINK]http://arxiv.org/abs/2610.23311v1 [DATE]2026-10-16 19:14:52+08:00 [CATEGORIES]cs.CL
Inferring L1 Backgrounds from L2 English with LLM Token Probabilities [AUTHORS]Ahmet Yavuz Uluslu, Mark Gales, Kate Knill, Gerold Schneider [ABSTRACT]Native language identification (NLI) infers a non-native speaker's first language (L1) from their second language (L2) production. We investigate a token probability approach that scores each text sample as a continuation of a prompt prefix specifying a candidate L1, then predicts the candidate with the highest score. We evaluate our approach using Llama 3.1 8B and 70B on three datasets: TOEFL9, Write and Improve 2024, and BULATS-L1. We compare its performance with previous prompting-based classification approaches. Token-probability scoring substantially improves on direct classification, particularly with the 8B model, while the 70B model achieves state-of-the-art performance across all datasets. Our results indicate that pretrained LLMs' token probabilities contain information useful for distinguishing L1 backgrounds from L2 English production. [LINK]http://arxiv.org/abs/2610.23309v1 [DATE]2026-10-16 19:12:42+08:00 [CATEGORIES]cs.CL
Annotrace: A Workbench for Step-Level Annotation of Agent Traces [AUTHORS]Arun Josephraj Arokiaraj, Chenyi Huang, Sahan Bulathwela, Zekun Wu, Bin Wu [ABSTRACT]Every study of why LLMagents fail begins by labelling traces by hand, and every such study builds its own throwaway tool. Observability platforms record agent traces but offer no structured labelling; annotation tools label text but have no notion of a step, a delegation or a tool call, leaving it unclear what the smallest unit of execution to label should be. Annotrace reads six trace formats, including OpenTelemetry GenAI, LangSmith and Langfuse, and turns each run into an ordered, nested list of steps. The annotator writes their own labelling codebook; a frontier model proposes a starting set from a sample of their own traces, answers first, and is asked exactly the questions the human is asked, making annotation a joint human-AI process. Annotrace measures agreement between any two annotators, human or AI. On a multi-agent question-answering corpus it reproduces a previous manual effort's 235 annotation units exactly, and two models labelling the same traces agree 91% of the time at $κ= 0.82$. [COMMENTS]4 pages, 4 figures, 1 table. Under review at the AAAI 2027 Demonstration Program [LINK]http://arxiv.org/abs/2610.23284v1 [DATE]2026-10-16 18:55:26+08:00 [CATEGORIES]cs.CL
Verbalising Model's Computation for Chain-of-Thought Faithfulness Detection [AUTHORS]Yujing Wang, Yulan He, Wenya Wang, Hanqi Yan [ABSTRACT]Current Chain-of-Thought (CoT) verification methods detect reasoning faithfulness from model outputs or static activations, but are limited in explaining how unfaithfulness develops during generation. We introduce VERITAS to verbalise the model's computation: how different reasoning anchors emerge, transfer across layers, and surface in the output. VERITAS adopts the layer-wise Jacobian projection to produce a verbalised account of the model's internal reasoning evolution at each generated CoT token. By training a classifier on the extracted features along the traces, we show that these structured features provide insightful perspectives on how and where reasoning diverges toward unfaithfulness. Experiments on four datasets from FaithCoT-Bench and BonaFide show that VERITAS achieves a significant performance gain over LLM-as-Judge, with a 28.4% gain in balanced accuracy, and an average improvement of 16.0% over activation probing. Further analysis shows VERITAS can be generalised to different datasets and model architectures. Our work shows that verbalising model computation provides a richer and more interpretable lens on CoT faithfulness, moving beyond detection toward a mechanistic understanding of how reasoning fails. [LINK]http://arxiv.org/abs/2610.23275v1 [DATE]2026-10-16 18:44:23+08:00 [CATEGORIES]cs.CL
Sample, Mix, or Verify? Allocating Test-Time Compute under Imperfect Verifiers [AUTHORS]Srinivasan Iyengar [ABSTRACT]A common way to make language models and agents more accurate is to spend more compute at inference time: generate many candidate solutions and submit the one a verifier prefers. Success then requires that some candidate is correct and that the verifier picks it. The first grows predictably with the number of samples; the second does not, because real verifiers (unit tests, majority votes, LLM judges, reward models) accept wrong answers, reject right ones, and make correlated mistakes. With a fixed budget, a practitioner must decide how many samples to draw, which models to draw them from, and how much to spend on verification, and today these choices are made by rules of thumb. We introduce SPLIT, which makes them from a small labelled pilot. It rests on an exact formula for the probability of submitting a correct answer under accept/reject and top-score selection that accounts for varying task difficulty, blind spots shared across models, and correlated repeated verifier calls. SPLIT fits the formula by empirical Bayes and uses it to plan budgets on unseen tasks. Across seven benchmarks spanning math, science QA, code and software-engineering agents, 26 models, 175 agent systems and six kinds of real verifiers, all from public releases, SPLIT lands within 0.02-0.8 percentage points of the best allocation in hindsight, while every fixed rule we test loses at least 2.5 points in some setting. Regret depends only on how well SPLIT ranks configurations, and it stays near-optimal under compute, price and latency costs and online. The formula also explains when asking more LLM judges lowers precision. As a forecaster far beyond the pilot, SPLIT beats power-law fitting when the verifier caps success, and we report where it loses. [COMMENTS]27 pages (18 main + appendix), 14 figures, 11 tables. Code will be released [LINK]http://arxiv.org/abs/2610.23255v1 [DATE]2026-10-16 18:28:07+08:00 [CATEGORIES]cs.LGcs.CL
Ignored, Not Refused: What Low Prompt-Injection Success Rates Hide in Small Tool-Using Agents [AUTHORS]Farhan Zarif, Kazi Shahid Ahmed Galib, Synthia Akter Promi, Shabnam Sharmin Tarin, Ariyan Hossain, Sadif Ahmed, Farig Yousuf Sadeque [ABSTRACT]Tool-using agents read untrusted external content, which exposes them to indirect prompt injection (IPI), where hidden instructions hijack the agent's tool calls. Benchmarks measure IPI security by the attack success rate (ASR), which treats every failed attack as a successful defense. To test this assumption, we audit 33,000 trajectories from eleven open-weight agents (2-14B parameters) on InjecAgent and AgentDojo. A label-blind LLM judge shows that most failed attacks are ignored (76-80%) rather than refused (17-18%). With identical model weights, changing only the system prompt shifts InjecAgent ASR from 0.2% to 43%. Our Attack-Task Outcome Matrix (ATOM) further shows that in 21.7% of AgentDojo episodes, the agent stayed safe but did not complete the user's task. A low ASR therefore mostly reflects agents silently ignoring injections, not refusing them. Evaluations should report how attacks fail and whether the user's task is completed, alongside ASR. We release the labeled corpus, judge prompt, and analysis code. [COMMENTS]17 pages, 3 figures, 11 tables [LINK]http://arxiv.org/abs/2610.23254v1 [DATE]2026-10-16 18:26:23+08:00 [CATEGORIES]cs.CL
Who Pays for Efficient Inference? Confidence-Based Small-to-Large LLM Routing Across Tasks and Languages [AUTHORS]Md Sabbir Hossen, Anichur Rahman, Pabon Shaha, Airin Afroj Aishi, Andrew H. Sung, Md Shohel Rana [ABSTRACT]Small-to-large language model routing can reduce inference cost by escalating difficult inputs while retaining easier ones on a smaller model. Its success, however, depends on identifying examples that benefit from escalation, and the reliability of confidence across multilingual tasks remains unclear. We present a systematic evaluation of confidence-based routing across three model families, four languages, and three task types, covering 72 model, dataset, language, and scale configurations and 42,288 predictions. We compare global confidence routing with random routing, equal-quota confidence routing, and an oracle upper bound. Confidence identifies beneficial escalations on the Belebele reading comprehension and SIB-200 topic classification datasets, with mean correction-detection AUCs of 0.612 and 0.705. At 50% escalation, it improves over random routing by 1.40--4.67 percentage points on Belebele and 2.54--5.63 points on SIB-200. In contrast, the mean AUC on the MMLU-ProX-Lite knowledge and reasoning dataset is only 0.510, and routing provides no consistent improvement over random selection because small models often remain highly confident while incorrect. Global confidence routing also creates language escalation gaps up to 30.33 percentage points, whereas equal-quota routing removes this disparity with little aggregate accuracy cost. These results show that confidence routing is task-dependent and can allocate computation unevenly across languages. Code is available at https://github.com/sabbir5622r/multilingual-llm-routing [LINK]http://arxiv.org/abs/2610.23249v1 [DATE]2026-10-16 18:23:41+08:00 [CATEGORIES]cs.CL
MultiAgent-OPD: End-to-End Training of Hierarchical Multi-Agent Systems with On-Policy Distillation [AUTHORS]Yuhan Sun, Jiaxin Zhao, Yongdong Luo, Yibo Hu, Meiguang Jin, Junfeng Ma, Weixun Wang, Weihang Pan, Zhongan Bi [ABSTRACT]In a hierarchical multi-agent system, a Leader assigns tasks and subagents carry them out. Jointly training the Leader and the subagents is difficult because a final answer gives only indirect feedback to the subagents. An LLM judge can score the performance of the subagents, but these scores depend on the judge's own assessment of the quality of subagents' responses. We present MultiAgent-OPD, which combines on-policy distillation with reinforcement learning to jointly train a shared-parameter Leader-Searcher-Answerer team end to end. We first train a Teacher in a Searcher-Answerer workflow using answer correctness. During team training, the frozen Teacher scores student-generated search and summary tokens to provide local process supervision for the Searchers, while the Leader learns from team success and the Answerer from candidate correctness. Across seven datasets, the trained system reaches 38.68% macro avg@4, exceeding Prompted Critic by 4.06 percentage points and C-MAS + RL, the trained cascaded multiagent system, by 2.61 percentage points. The team also solves more questions within four attempts. [LINK]http://arxiv.org/abs/2610.23234v1 [DATE]2026-10-16 18:01:52+08:00 [CATEGORIES]cs.CL
STAIF: A Stage-wise Optimization for Complex Instruction Following [AUTHORS]Jian Hong, Chen Cheng, Quan Liu, Yuhao Chen, Enhong Chen [ABSTRACT]Following complex instructions with multiple explicit constraints remains a fundamental challenge for large language models (LLMs). Existing alignment methods, such as DPO, optimize holistic reward signals that often underemphasize strict satisfaction of individual constraints, particularly under out-of-distribution or multi-constraint settings. In this paper, we propose STAIF, a stage-wise optimization framework that decouples the alignment of subjective (soft) constraints from the optimization of objectively verifiable (hard) constraints. Stage 1 applies preference optimization with multiple negative samples to sharpen sensitivity to soft constraints, while Stage 2 applies Reinforcement Learning with Verifiable Rewards (RLVR) to enforce strict compliance with hard constraints. To support this method, we construct STAINSTRUCT, a high-quality bilingual (English, Chinese) dataset of approximately 31,000 complex multi-constraint instructions. Extensive analyses validate the design of STAIF and show state-of-the-art performance on representative benchmarks against strong baselines, as well as genuine generalization. [COMMENTS]16 pages, 6 figures [LINK]http://arxiv.org/abs/2607.22649v2 [DATE]2026-10-16 18:00:07+08:00 [CATEGORIES]cs.CL
MAS-Bai: Linguistically Informed Adaptive Multi-Agent Translation for Bowen-Script Bai [AUTHORS]Jingting Zheng, Hong Ren, Yuehan Cui, Shaolin Zhu, Deyi Xiong [ABSTRACT]Large Language Models (LLMs) remain unreliable for extremely low-resource languages whose writing systems conflict with high-resource pretraining priors. We study Bai-Chinese translation in the Bowen script, where Chinese characters are frequently borrowed phonetically rather than semantically, causing severe literal-translation hallucinations. We propose MAS-Bai, a linguistically informed adaptive multi-agent framework for Bowen-script Bai-Chinese translation. MAS-Bai routes each input through either a lightweight retrieval-translation path or a deeper philological path with etymological and syntacticagents, and uses cycle-consistency with a lightweight reverse translator as an inference-time reliability signal. We construct a 6,536-sentence Bai-Chinese corpus and evaluate on a 1,536-sentence test set. MAS-Bai achieves 16.23 BLEU and 62.15 COMET, improving over Standard RAG by 8.11 BLEU and 6.85 COMET. Forward LoRA, expert ablations and controlled routing experiments distinguish the roles of supervised adaptation, Bai-specific expertise and conditional routing. The largest gains occur on Bowen phonetic loans and divergent Bai word order, while human evaluation and reverse-verifier analyses support the automatic results. [COMMENTS]14 pages, 3 figures, Findings of EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23205v1 [DATE]2026-10-16 17:43:32+08:00 [CATEGORIES]cs.CL
Pro-AI Bias in Large Language Models [AUTHORS]Benaya Trabelsi, Jonathan Shaki, Sarit Kraus [ABSTRACT]Large language models (LLMs) are increasingly employed for decision-support across multiple domains. We ask whether these models favor artificial intelligence (AI) itself. We examine this question through three experiments. First, in open-ended advice across investment, study, career, and startup domains, AI is prominent in most recommendation lists and is usually ranked first when it appears. Second, in salary estimations, models overestimate AI-related jobs more than they do for closely matched non-AI jobs. In a controlled follow-up, changing only the job title from the matched non-AI title to the AI-related title raises predicted salary by 10%, while all other supplied job details remain fixed. Finally, we show that the number of degrees and other credentials awarded by AI programs were 28% more overestimated than corresponding awards for matched computing programs. Overall, we show that AI is not only prominent in everyday LLM advice, but also receives systematically inflated quantitative estimates compared to non-AI cases, even against its closest disciplinary neighbors, raising concerns for high-stakes decision support. [COMMENTS]21 pages, 4 figures. Reproducibility code included as ancillary files [LINK]http://arxiv.org/abs/2601.13749v2 [DATE]2026-10-16 17:34:15+08:00 [CATEGORIES]cs.CLcs.LG
Token-Budget-Aware LLM Reasoning [AUTHORS]Tingxu Han, Zhenting Wang, Chunrong Fang, Shiyu Zhao, Shiqing Ma, Zhenyu Chen [ABSTRACT]Reasoning is critical for large language models (LLMs) to excel in a wide range of tasks. While methods like Chain-of-Thought (CoT) reasoning and enhance LLM performance by decomposing problems into intermediate steps, they also incur significant overhead in token usage, leading to increased costs. We find that the reasoning process of current LLMs is unnecessarily lengthy and it can be compressed by including a reasonable token budget in the prompt, but the choice of token budget plays a crucial role in the actual compression effectiveness. We then propose a token-budget-aware LLM reasoning framework that dynamically adjusts the number of reasoning tokens based on the reasoning complexity of each problem. Experiments show that our method effectively reduces token costs in CoT reasoning with only a slight performance reduction, offering a practical solution to balance efficiency and accuracy in LLM reasoning. Code: https://github.com/GeniusHTX/TALE [LINK]http://arxiv.org/abs/2412.18547v6 [DATE]2026-10-16 17:33:52+08:00 [CATEGORIES]cs.CLcs.LG
BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models [AUTHORS]Liubov Chubarova, Alexandra Kuleshova, Daniil Volkov, Kirill Sultanov, Alexey Zaytsev [ABSTRACT]While Multimodal Large Language Models (MLLMs) have made significant strides in visual comprehension, their ability to reason about text-dense, professional documents remains incompletely evaluated. Existing benchmarks emphasize information extraction, require external domain knowledge, or cover professional documents only as one of many settings. They are also largely English- or Chinese-centric, leaving other languages and Russian, in particular, substantially underrepresented. To address these limitations, we introduce BEAR-Bench (Bilingual Enterprise and Academic Reasoning), a self-contained, complex English-and-Russian benchmark comprising 1000 human-annotated questions based on text-rich business and scientific documents. We evaluate 16 proprietary and open-weight MLLMs, including Gemini 3.1 Pro and Qwen3.5-397B, on BEAR-Bench and observe clear headroom even for the strongest systems. Finally, we use the resulting model outputs to compare existing hallucination detection methods, evaluating not only how often models fail on BEAR-Bench but also how reliably those failures can be identified. [LINK]http://arxiv.org/abs/2608.17895v2 [DATE]2026-10-16 17:30:41+08:00 [CATEGORIES]cs.CL
What LLMs Do in the Shadows: Reading Context Reliance from the Unembedding Dark Space [AUTHORS]Alexandra Kuleshova, Andrey Volodichev, Daria Kotova, Alexey Zaytsev [ABSTRACT]Detecting faithfulness hallucinations in large language models (LLMs) is crucial for their safe deployment in high-stakes domains. Recent approaches address this by contrasting contextual against parametric knowledge utilization, yet measuring context reliance typically demands noticeable overhead, such as materializing attention maps, running additional forward passes, or training auxiliary probes. In this work, we show that context reliance can be estimated by projecting hidden states onto the dark space - directions that compose weakly with the unembedding matrix. Beyond its recognized function as an initial-token attention sink, we show that this subspace actively mediates context processing, and that its share of hidden-state norm - termed the Dark Matter Ratio (DMR) - provides a direct score of context utilization. Pairing DMR with a parametric knowledge score yields DarkLens, a lightweight faithfulness detector operating strictly within a single forward pass. Evaluated across three QA and summarization benchmarks on six open-weight LLMs (4B-13B parameters), DarkLens matches or outperforms a wide range of baselines without requiring extra forward passes or probe training. [LINK]http://arxiv.org/abs/2610.23182v1 [DATE]2026-10-16 17:21:49+08:00 [CATEGORIES]cs.CL
Tree-Aware Posterior Distillation for Long-Horizon Agentic Reinforcement Learning [AUTHORS]Heming Zou, Qi Wang, Yixiu Mao, Yun Qu, Yuyi Guo, Zeyuan Liu, Chenglin Li, Weijie Liu, Saiyong Yang, Xiangyang Ji [ABSTRACT]Reinforcement learning with verifiable rewards has become a scalable post-training paradigm for large language models. However, training agents over long-horizon interactions presents two challenges: using only terminal rewards makes turn-level credit assignment sparse and high-variance, while sequential mini-batch updates on a fixed rollout batch introduce policy staleness as the current policy drifts from the behavior policy. Existing approaches obtain denser credit via learned critics, process reward models, or on-policy distillation, but they rely on a value network, step-level annotations, or a stronger external teacher that increase training complexity. Meanwhile, importance sampling and clipping mitigate policy drift, but they do not specify a finite target that stops sequential updates from overshooting. In this work, we introduce TreeOPSD, a critic-free, tree-aware posterior self-distillation framework for long-horizon agentic reinforcement learning. Treating each interaction turn as a tree node, TreeOPSD branches alternative continuations from shared prefixes and recursively backs up terminal rewards into implicit turn-level credit. It then maps each credit to a closed-form, bounded teacher distribution parameterized by the likelihood ratio relative to the frozen behavior policy. Minimizing the divergence between this teacher and the student policy recovers the policy-gradient direction as the tangent at the behavior policy while bounding updates toward a finite target across sequential mini-batches. Empirically, TreeOPSD improves agentic capabilities across several competitive benchmarks and model scales, especially for long-horizon tasks, e.g., raising the average accuracy of Qwen3-30B-A3B-Thinking-2507 by 4.97 points over GRPO on the Deep Research setting. [LINK]http://arxiv.org/abs/2610.23181v1 [DATE]2026-10-16 17:17:26+08:00 [CATEGORIES]cs.LGcs.CL
Cross-Centaur: Modeling Individual Decision-Making Across Psychological Tasks with Language Models [AUTHORS]Hikaru Asano, Hidezo Suganuma, Ryutaro Mori, Yukino Baba [ABSTRACT]Large language models (LLMs) are increasingly used as computational models of human behavior. Recent models such as Centaur have advanced this goal by fine-tuning LLMs on trial-by-trial decisions from psychological experiments. However, existing large-scale behavioral datasets do not link the same participants across different tasks, so these models learn population-level regularities but are not trained to model individual-level regularities across tasks. We introduce Cross-Psych-101, a dataset of 591 participants and 261\{,\}551 valid decisions across 12 Bandit, Tradeoff, Structure, and Social tasks. We train Cross-Centaur to predict a participant's target-task actions from a target-task description and the source-task description paired with the same participant's source-task behavior. Cross-Centaur outperforms off-the-shelf LLMs and Centaur on held-out participants, both within and across tasks. Probing analyses suggest that its hidden states encode participant-level behavioral tendencies that transfer across tasks. These results highlight the potential of cross-task individual-level data for predicting human decision-making with language models. Cross-Pysch-101 is available at https://huggingface.co/datasets/HikaruAsano/cross-psych-101. [COMMENTS]Accepted to EMNLP 2026 (Main Conference, Oral Presentation) [LINK]http://arxiv.org/abs/2610.23157v1 [DATE]2026-10-16 16:55:00+08:00 [CATEGORIES]cs.CL
Can AI interview? Comparing AI-led to Human-led Chat Interviews [AUTHORS]Semra Guven, Tobias Gårdhus, Andreas Bjerre-Nielsen, Anna Rogers, Hjalmar Bang Carlsen [ABSTRACT]Large language model chatbots have been proposed as AI-interviewers for collecting qualitative interview data at scale, potentially bridging social science's longstanding divide between the breadth of surveys and the depth of qualitative interviews. This paper provides the first direct empirical comparison of data quality in AI-led and human-led interviews. In a randomized experiment (N = 41), university students were interviewed via synchronous text chat by either trained non-expert interviewers or a locally hosted AI system; we evaluate exposure, specificity, and relevance. Median interview duration was similar across conditions, but AI-interviewers asked more than twice as many questions per interview and elicited roughly 65 percent more respondent text per interview. Per-question responses were longer with human interviewers, though not after multiple-testing correction; only AI-condition participants skipped questions. Responses were almost uniformly on-topic in both conditions; for specificity we detect no difference, but the design cannot exclude one of up to three quarters of a point. These results were obtained with a small, locally hosted open-weight model (7B parameters), which addresses concerns about data security, GDPR compliance, and scientific reproducibility. Our findings support the viability of AI-interviewing for large-N qualitative data collection. [COMMENTS]37 pages, 3 figures, 6 tables. Submitted to Socius. Code and data: https://github.com/ainterviewer/can-ai-interview [LINK]http://arxiv.org/abs/2610.23149v1 [DATE]2026-10-16 16:43:18+08:00 [CATEGORIES]cs.CL
Deep LLM layers expose passage-level context in EEG during silent reading [AUTHORS]Ingo Marquardt, Anthilia Alchanat, Priyanka Jain [ABSTRACT]In a previous report we showed that a decoder can recover open-vocabulary word-level information from 19-channel EEG recorded during silent reading, by aligning short EEG windows with hidden-state embeddings of the presented word taken from a large language model (LLM). That analysis contrasted two of the model's 33 layers: the context-free input embedding and one mid-depth layer. Here we ask how EEG-decodable structure is distributed over the full depth of the language model. We trained one decoder per layer of Llama-3.1-8B, with and without a causal transformer over the sequence of EEG windows. We distinguish word-level decodability from a context-tracking component. Word-level decodability peaks in early layers and declines steadily towards the last layer. In contrast, the context-tracking component of the retrieval gain rises with depth, both with and without a causal transformer. Unexpectedly, a decoder that sees a single 0.5-s window at a time tracks the text passage too. Two new controls dissect this component. A target-geometry control, a synthetic predictor made of the true target plus noise calibrated to each decoder's word-level accuracy, shows that context tracking is not a by-product of word decoding and target geometry. A control that permutes predictions across runs from either the same or from different sessions shows that most of the observed context-tracking is not caused by an EEG recording session signature. Instead, context tracking mostly follows the text. A single EEG window therefore carries passage-level information that a contextual target can absorb. The results locate word identity in the early layers of the language model, discourse state in the deeper layers, and make the choice of target layer a design dial for EEG language decoders. [COMMENTS]41 pages (including 23 pages of Supplementary Material), 12 figures (including 5 in Supplementary Material) [LINK]http://arxiv.org/abs/2610.23638v1 [DATE]2026-10-16 23:41:19+08:00 [CATEGORIES]cs.LG
ThetreStageAI: Multi-View Retrieval-Augmented Generation for Structured Bengali Theatre Production Planning [AUTHORS]Md. Mehedi Hasan Naeem, Mst. Kamrunnahar Ruma, Shakila Sultana, Nafiza Anjum, Nasreen Jannatun Nemu, Md. Sujan Ali [ABSTRACT]Staging a theatre performance requires coordinating dialogue, actor movement, stage directions, lighting focus, cue timing, and equipment constraints within the same production plan. This work presents ThetreStageAI, a retrieval augmented framework for structured Bengali theatre production planning. Rather than retrieving from one mixed corpus, the system maintains separate scene, actor-blocking, and lighting views. Bengali text is embedded with BGE-M3 and indexed in view-specific Qdrant collections. Retrieved references are combined with a production brief and passed to Gemini, which generates schema-constrained JSON containing characters, scenes, dialogue, stage directions, blocking, lighting, and optional sound cues. Structural and semantic checks are applied before acceptance, with at most one constrained repair attempt. We also construct a human-guided, LLM-assisted synthetic dataset of 500 short Bengali theatre records that expands to 1,500 retrieval documents. Dataset diagnostics confirm consistent identifiers and valid lighting fields while also revealing substantial dialogue repetition. A paired camera-ready evaluation compares single-combined and full multi-view retrieval under matched settings. The source code, dataset, and evaluation materials are publicly available. [COMMENTS]6 pages, 6 figures. Accepted at the 12th IEEE International Conference on Sustainable Technologies and Engineering (IEEE iCOSTE 2026), Cox's Bazar, Bangladesh, November 1-3, 2026 [LINK]http://arxiv.org/abs/2610.23618v1 [DATE]2026-10-16 23:29:20+08:00 [CATEGORIES]cs.LG
Activation-Efficient Fine-Tuning via Sequence-Compressed Low-Rank Subspaces [AUTHORS]Irene Tenison, Stella Ahn, Miriam Kim, Ebtisam Alshehri, Lalana Kagal [ABSTRACT]Parameter-Efficient Fine-Tuning (PEFT) has become the standard approach for adapting large language models (LLMs). However, despite reducing trainable parameter counts, many PEFT methods remain memory-bound by token-level activation intermediates that scale linearly with sequence length, limiting their practicality for long-context and on-device adaptation. Unlike prior memory-efficient PEFT approaches that preserve token-level adapter activations, LARS (Low-memory Activation-Rank Subspace) reduces the sequence-length scaling of stored adapter activation intermediates from O(BSRL) to O(BRL) by performing adaptation within a sequence-level low-rank subspace. Through pooled feature extraction, low-rank subspace modulations, and residual projection, LARS reduces the overall memory footprint by 23% on GPUs and 47.8% on CPUs relative to LoRA at S = 1024, with relative savings increasing with context length. Experiments across reasoning, understanding, long-context, retrieval, and token-level tasks show that LARS maintains competitive performance relative to LoRA while substantially improving memory efficiency. Beyond GPUs, LARS also reduces average CPU power and energy consumption per optimisation step by 22.43% and 22.06%, respectively. These results demonstrate that activation-efficient adaptation provides a practical path toward resource-constrained and on-device fine-tuning of LLMs. [COMMENTS]Accepted to AACL 2026 [LINK]http://arxiv.org/abs/2604.22783v2 [DATE]2026-10-16 23:29:14+08:00 [CATEGORIES]cs.LG
Boundary Detection for High-Dimensional Point Clouds [AUTHORS]Abigail Hickok, Valerio Galanti, Osama Al Dalahmah, Andrew J. Blumberg [ABSTRACT]We introduce two new methods for boundary detection in high-dimensional point cloud data, intended to work in situations with non-uniform density, high ambient dimension, and substantial noise. Our first method fits a separating hyperplane on local neighborhoods after doing local dimension reduction---this method works best when one can reliably estimate the intrinsic dimension of the data. Our second method measures the directional uniformity of local neighborhoods, and has the best overall performance on synthetic data. In addition to pointwise numerical boundary scores, our algorithms yield an estimated normal vector at each boundary point. We validate our methodology on synthetic data as well as two real examples: we analyze spatial transcriptomics data, where we show that we can detect the boundary of a tumor, and we also study LLM activation geometry. [COMMENTS]31 pages; 23 figures [LINK]http://arxiv.org/abs/2610.23590v1 [DATE]2026-10-16 23:12:00+08:00 [CATEGORIES]cs.LG
Beyond Latent Personalization: Integrating HRV-Conditioned Personalization via LLM for Interpretable Stress Detection [AUTHORS]Wei-Hsiang Ma, Ya-Tse Wu, Hsing-Hang Chou, Chi-Chun Lee [ABSTRACT]Stress detection based on heart rate variability (HRV) has received increasing attention. However, existing personalized approaches integrate personal attributes with HRV through latent representations, limiting the interpretability of their interactions. In this work, we propose an LLM-enhanced personalized stress detection framework that integrates HRV features and personal attributes through LLM-based semantic interpretation. The proposed framework achieves a 29.9% relative MCC improvement over the strongest baseline on TILES-2018 while providing human-readable descriptions of individual physiological responses. The results highlight the potential of LLM-based semantic interpretation for achieving both effective and interpretable personalized stress detection. [COMMENTS]Submitted to ICASSP 2027 [LINK]http://arxiv.org/abs/2610.23555v1 [DATE]2026-10-16 22:44:02+08:00 [CATEGORIES]cs.LG
Out of Support, Not Out of Reach: In-Context Learning with Unsupported Instructions and Composed Tasks [AUTHORS]Xuhan Tong, Yuchen Zeng, Jiawei Zhang [ABSTRACT]In-Context Learning (ICL) enables pretrained LLMs to adapt to downstream tasks from a small set of input-output demonstrations, without parameter updates. Yet existing theory primarily explains ICL as identifying a latent task already represented in pretraining, leaving unclear whether ICL can succeed when either the instruction or the target task falls outside the pretraining support. We develop a theoretical framework under mild assumptions that separates these two forms of distribution shift. For target tasks represented in pretraining, we show that demonstrations can identify the task and reduce sensitivity to pretrain-supported prompt templates as their number increases. For an unsupported instruction, we show that under a fixed model-induced reference extension, demonstration evidence can dominate an $N$-independent instruction shift while preserving the exponential identification rate. We then use the same ICL framework as an interpretive lens for long-horizon target tasks that are not represented in pretraining but admit a plan that decomposes them into pretraining-supported subtasks. Applying the supported-task analysis step by step yields a standard accumulated-error bound and clarifies how invoking familiar subtasks along an intermediate-state trajectory can avoid directly selecting the unsupported complete task. Together, our results characterize when demonstrations, instructions, and task decomposition can support ICL beyond the corresponding pretraining supports. [LINK]http://arxiv.org/abs/2610.23528v1 [DATE]2026-10-16 22:27:05+08:00 [CATEGORIES]cs.LG
Generative Policy Improvement from Observational Data [AUTHORS]Jonas Schweisthal, Dennis Frauen, Stefan Feuerriegel, Emma Brunskill [ABSTRACT]We study how to improve decision policies from observational data to generate better actions when actions are complex, structured objects, such as text messages or images. This task is subject to two major challenges: (i) action-level overlap is often too weak for reliable off-policy evaluation and improvement, and (ii) generating new structured actions often requires expensive training or fine-tuning (e.g., of an LLM) and can be ill-defined due to information loss in action representations. To address both, we propose PolicyLift, a three-stage framework for generative policy improvement with structured actions from observational data. In Stage 1, PolicyLift learns a small set of discrete action codes that capture high-level strategies in the observed action space and are tailored to support policy improvement under limited overlap. In Stage 2, it learns a context-dependent policy over these action codes. At deployment time, Stage 3 "lifts" the learned code-level policy back to the original action space by generating actions from the historical behavior system in a code-conditioned way. We show theoretically that improving the policy over action codes guarantees improvement of the induced policy in the original action space. We demonstrate PolicyLift across experiments in synthetic and real-world text generation tasks with consistent policy improvements over the behavior policy. [COMMENTS]Preprint [LINK]http://arxiv.org/abs/2610.23436v1 [DATE]2026-10-16 21:20:43+08:00 [CATEGORIES]cs.LG
2026 Oct 17, Sat
The Italian lawmaking process: a unified and structured dataset [AUTHORS]Matilde Contestabile, Chiara Ferrara, Giovanni Parrillo, Lorenzo Emer, Andrea Vandin [ABSTRACT]The Italian legislative process is documented in detailed parliamentary records hosted in institutional websites. However, much of this information is available only as unstructured text, limiting its systematic reuse. We applied a controlled, systematic LLM-based extraction procedure, validated against manually annotated records, to convert these texts into ordered and dated events including, e.g., bill introduction, committee examination, plenary examination, and approval. The resulting dataset reconstructs the procedural histories of all 5346 Italian laws enacted between 1985 and 2026, spanning parliamentary terms IX to XIX, for overall 101413 events. Each event records the activity, date, and chamber, with additional procedural details where available. The dataset supports research in the emerging fields of computational legal and legislative studies, potentially combined with process-oriented data science methods to examine procedural paths and legislative timing across parliamentary terms. [LINK]http://arxiv.org/abs/2610.23741v1 [DATE]2026-10-17 00:52:11+08:00 [CATEGORIES]cs.CL
From Flow-Box Invariants to Global Conservation Laws: Machine Learning-Assisted Detection of Singularities and Near-Integrability [AUTHORS]Jimmie Adriazola, Ioannis G. Kevrekidis, Panayotis G. Kevrekidis, Wei Zhu [ABSTRACT]Conservation laws reveal invariant structure in dynamical systems and constrain their evolution to lower-dimensional level sets. Locally, however, per the flow-box theorem, every $d$-dimensional system generically admits $d-1$ functionally independent conservation laws near a regular point even though, generically, only a few such laws persist globally. Data-driven methods typically learn conservation laws on a prescribed region of phase space from which training samples are drawn, leaving their behavior beyond that region unresolved. We formulate this local-to-global continuation problem and develop an \{expanding-domain neural deflation\} framework that tracks the number of independent conservation laws over nested domains. The method recovers the expected $d-1$ local invariants and detects their successive breakdown as the domain expands, relaxing to the expected global conservation law limit. It distinguishes the local invariants (associated with Frobenius integrability) from persistent global structure and shows that the domain scale over which additional laws survive reflects proximity to Liouville integrability. Pointwise losses can localize isolated obstructions, whereas in more complicated dynamics the breakdown of our loss functions is gradual and spatially diffuse. The framework thus provides a data-driven approach for studying the persistence and breakdown of (so-far-learned) conservation laws across phase space. [LINK]http://arxiv.org/abs/2610.23760v1 [DATE]2026-10-17 01:02:39+08:00 [CATEGORIES]cs.LG
2026 Oct 16, Fri
A low-dimensional dynamical basis for tongue shape control [AUTHORS]Sam Kirkham, Patrycja Strycharczuk [ABSTRACT]Speech production involves the control of a high-dimensional articulatory system with many degrees-of-freedom, but a widespread hypothesis is that speakers navigate this complexity using a set of lower-dimensional control variables. Common candidates for these control variables include individual articulators (e.g. tongue body) and abstract tract variables (e.g. tongue body constriction location), but identifying the most appropriate coordinate system for articulatory control remains an open question. In this study, we advance a model in which holistic tongue shape modes represent a potential low-dimensional control basis for a simple dynamical law. We first discover empirical tongue shape modes in a cross-speaker coordinate space and then fit the model to ultrasound data on monophthongal vowels from 36 speakers of British English. A three-dimensional linear point attractor fitted to these time-varying modes recovers the dynamics of tongue shape with a high degree of accuracy. This suggests that complex tongue movement can arise from independent modes that linearly combine under simple dynamical laws, thereby relocating some complexity from the control signal to the coordinate system. We show that this framing explains why one-dimensional articulatory measures can exhibit observable nonlinearities, and hypothesize that shape modes are a plausible kinematic expression of underlying muscle synergies. [COMMENTS]Pre-print currently under review [LINK]http://arxiv.org/abs/2610.23456v1 [DATE]2026-10-16 21:39:24+08:00 [CATEGORIES]cs.CL
"You" Is Not Toxic: How LLM Labelers Rewrite the Web's Conversation Laws [AUTHORS]Liang Yue, Eric A. Vance [ABSTRACT]What the Web community believes about toxic conversations, such as that toxicity rises in longer and deeper threads, was measured mostly with Google's Perspective API, which is shutting down; large language models (LLMs) will make the next measurements. We find that LLM labelers mistake second-person address for toxicity, and that this one bias manufactures laws that human labels do not show. On 1.25 million news comments with crowd labels, Claude Haiku flags 12% of the non-toxic comments that contain no "you" and 51% of those with four or more; a single crowd rater flags 6% and 10%. Deeper replies address someone more often, so under Haiku toxicity rises with reply depth (+0.93 percentage points per level), while under the crowd's labels it slightly declines (-0.12). Rewriting "you" in the third person removes two thirds of the false rise; flagging only as many comments as the crowd does not remove it, and showing Haiku the parent comment enlarges it. All seven LLMs we test over-flag "you", and at the crowd's flag rate, without the parent, any false rise of the open ones vanishes at a fixed number of second-person words; Haiku also over-flags "you" on Reddit, Gab and Wikipedia talk pages. Agreement with human labels does not reveal the bias: we prove that a fixed Cohen's kappa leaves a law nearly unconstrained. We therefore certify the law instead of the labeler. Threadsafe uses the labeler's scores to choose which comments human raters label and to predict the others, and corrects the predictions with the human labels; linear laws are estimated without bias whatever the labeler does. In simulated audits its intervals for the laws keep nominal coverage from 250 labels on, where uniform-sample methods such as PPI++ fall short, and at 1,000 to 8,000 labels are 0.75 times as wide as those of design-based supervised learning (DSL); a thousand human labels rule out Haiku's rising law. [COMMENTS]43 pages, 4 figures, 44 tables. Labels, prompts and code: https://github.com/liyu2195/you-is-not-toxic [LINK]http://arxiv.org/abs/2610.23446v1 [DATE]2026-10-16 21:29:30+08:00 [CATEGORIES]cs.CL
Sample, Mix, or Verify? Allocating Test-Time Compute under Imperfect Verifiers [AUTHORS]Srinivasan Iyengar [ABSTRACT]A common way to make language models and agents more accurate is to spend more compute at inference time: generate many candidate solutions and submit the one a verifier prefers. Success then requires that some candidate is correct and that the verifier picks it. The first grows predictably with the number of samples; the second does not, because real verifiers (unit tests, majority votes, LLM judges, reward models) accept wrong answers, reject right ones, and make correlated mistakes. With a fixed budget, a practitioner must decide how many samples to draw, which models to draw them from, and how much to spend on verification, and today these choices are made by rules of thumb. We introduce SPLIT, which makes them from a small labelled pilot. It rests on an exact formula for the probability of submitting a correct answer under accept/reject and top-score selection that accounts for varying task difficulty, blind spots shared across models, and correlated repeated verifier calls. SPLIT fits the formula by empirical Bayes and uses it to plan budgets on unseen tasks. Across seven benchmarks spanning math, science QA, code and software-engineering agents, 26 models, 175 agent systems and six kinds of real verifiers, all from public releases, SPLIT lands within 0.02-0.8 percentage points of the best allocation in hindsight, while every fixed rule we test loses at least 2.5 points in some setting. Regret depends only on how well SPLIT ranks configurations, and it stays near-optimal under compute, price and latency costs and online. The formula also explains when asking more LLM judges lowers precision. As a forecaster far beyond the pilot, SPLIT beats power-law fitting when the verifier caps success, and we report where it loses. [COMMENTS]27 pages (18 main + appendix), 14 figures, 11 tables. Code will be released [LINK]http://arxiv.org/abs/2610.23255v1 [DATE]2026-10-16 18:28:07+08:00 [CATEGORIES]cs.LGcs.CL
Chinese Decision: A Family of Language Models for Single-Pass Probabilistic Decision-Making [AUTHORS]Qiang Yan [ABSTRACT]Many language-based applications require probabilities over a finite set of decisions rather than open-ended responses. We present Chinese Decision, a family of Chinese-language models for single-pass probabilistic decision-making. Given context, an instruction, and candidate descriptions, the model extracts final-position candidate-letter logits in one forward pass and normalizes them over valid options, without autoregressive answer generation. A shared interface supports Boolean judgments, categorical choices, and ordinal scores. We report a completed 0.8B model trained with distribution cross-entropy on 13.048 million decisions and preliminary 2B checkpoints trained with an additional RLCD objective. On the full CJ-Bench of 307,900 decisions, the uncalibrated 0.8B model achieves 72.24% general-domain accuracy and 76.98% average accuracy across medicine, law, and finance using a single checkpoint, with 1.07% overall expected calibration error. The 2B model reaches 72.56% general-domain accuracy after 40.13% of one epoch but has not surpassed the completed 0.8B model on the professional-domain average. We further evaluate a validation-temperature-scaled deployment model and its Q8 quantization, finding close aggregate performance while documenting individual prediction changes. We release the training, evaluation, and deployment pipeline. The current results establish a reproducible approach to Chinese probabilistic decisions but do not isolate the effect of model scale. [COMMENTS]11 pages, 2 figures, 11 tables. Code: https://github.com/yanqiangmiffy/chinese-decision [LINK]http://arxiv.org/abs/2610.23170v1 [DATE]2026-10-16 17:08:32+08:00 [CATEGORIES]cs.CL
Beyond Expected Values: Risk-Sensitive Planning with Distributional Monte-Carlo Tree Search [AUTHORS]Quynh Trinh, Trung Nguyen, Nam Nguyen, Tuan Dam [ABSTRACT]Monte-Carlo Tree Search (MCTS) typically stores only scalar mean estimates, which makes it poorly suited for tail-risk objectives such as conditional value-at-risk (CVaR). We introduce Categorical Thompson Sampling with Optimism (CATSO) and Particle Thompson Sampling with Optimism (PATSO), two distributional MCTS algorithms that maintain finite-support empirical backup laws at Q-edges. CATSO represents Q-edge laws with categorical atoms and Dirichlet Thompson sampling, while PATSO represents them with adaptive particles. Both algorithms select actions using the CVaR of a Thompson-sampled Q-edge law plus a polynomial optimism bonus, and propagate scalar continuation values through visit-weighted averages of child Q-edge CVaR scores. For the finite-depth nested CVaR planning target, we prove an $O(n^\{-1/2\})$ bound on the root value error after $n$ rollouts, which upper-bounds the risk-sensitive simple regret. CATSO adds a fixed-grid discretization bias, while capped PATSO adds a tunable Wasserstein compression bias. Experiments on stochastic planning benchmarks show that distributional Q-edges improve lower-tail return and reduce catastrophic outcomes when mean-optimal and risk-sensitive behavior differ. [COMMENTS]Proceedings of the 40th Annual Conference on Neural Information Processing Systems [LINK]http://arxiv.org/abs/2610.23565v1 [DATE]2026-10-16 22:51:28+08:00 [CATEGORIES]cs.LG
Predictable Scaling: Efficient Model Exploration via Scaling-Law Prediction [AUTHORS]Aishwarya Wesanekar, Sheng-Tao Yang, Jiawei Zhang, Flavio Sales Truzzi, Chen Chen, Nader Beigiparast, Shengyuan Huang, Roshanak Zilouchian Moghaddam, Vitor Cid, Colin Clement, Vinayak Thapliyal, Sebastian Ament, Jihao Andreas Lin, Vincent Huang, Kai Han, Boyi Jiang, Wenlin Chen, Chaorong Chen, Ashwin Kumar, Maximilian Balandat, David Eriksson, Yifei Zhang, Pan Zhong [ABSTRACT]Evaluating candidate changes (hypotheses) to recommendation models requires full-scale training runs consuming thousands of GPU-hours, severely limiting the hypothesis space engineers can feasibly explore. We introduce Predictable Scaling (PS), a systematic framework that exploits empirically verified regularity conditions to predict full-scale model performance from small-scale proxies. PS constructs scaled-down variants via composable scaling functions along distinct resource axes, and projects metrics to the target scale via curve fitting with calibrated uncertainty estimates. We compare three prediction strategies of increasing expressiveness: even the simplest (raw proxy) achieves prediction error within the empirical noise floor of repeated-run variance; the most expressive (Gaussian Process) provides an indicative 15% relative MAE improvement and calibrated uncertainty. Evaluation on two ranking models of increasing complexity---a task-specific model and a foundation model---demonstrates that PS correctly ranks actionable hypotheses across diverse hypothesis types at 79-88% compute reductions. [LINK]http://arxiv.org/abs/2610.23453v1 [DATE]2026-10-16 21:35:54+08:00 [CATEGORIES]cs.LG
Regularized free deconvolution for spectral repair of resampling distortion in high-dimensional particle filters [AUTHORS]Deming Song, Chi-Kin Lam, Kefeng Li, Hui Yu, Xilin Li, Huagang Lin [ABSTRACT]Resampling is the standard remedy for weight degeneracy in particle filters, but in high dimensions it distorts the spectrum of the ensemble covariance in a way that is computable from the importance weights. This paper studies the inverse problem: recovering the pre-resampling spectrum from the observed weights and the empirical spectrum alone, and using it without biasing the filter. We first show that the forward spectral law extends to random importance weights once one conditions on the weights together with the realized offspring counts, rather than on the weights alone. We then construct a regularized free-deconvolution estimator, prove its consistency with an explicit error budget at a fixed resolution scale, and certify model mismatch through a computable forward residual. The repaired spectrum enters the filter only as the preconditioner of a Metropolis-adjusted rejuvenation move, so that spectral error affects mixing efficiency but not the invariant target. For the full filtering cycle, we derive a geometrically summable error recursion under stable scalar dynamics and an explicit law for the cost of abstaining from repair. Pre-registered simulations show that the inversion reduces the spectral error by an order of magnitude, that the corrected preconditioner recovers about 90% of the oracle's sampling efficiency in an anisotropic regime, and that the cost of abstention depends on when abstentions occur rather than on how many. The repair recovers eigenvalue laws, not eigenvectors, and is certified only within an explicitly stated exogenous-weight regime. [COMMENTS]This is a preprint of an article submitted to Signal Processing (Elsevier). The implementation and experiment scripts are available at https://github.com/songming-teaching/spectral-repair-resampling. This is the second paper in a series on spectral analysis of particle filters. The first paper is available at arXiv:2610.16680 [LINK]http://arxiv.org/abs/2610.23448v1 [DATE]2026-10-16 21:30:52+08:00 [CATEGORIES]cs.LG
2026 Oct 17, Sat
Acceptance Is Not Speed: Length Inflation in Training-Free Lossy Speculative Decoding [AUTHORS]Chiatzen Wang, Bill Xu, Haochen Gu, Hongyang Zhang [ABSTRACT]Speculative decoding typically characterizes speedup by draft tokens accepted per verification round. Lossless verification preserves the target model's output distribution and thus expected completion length, making this metric a meaningful efficiency proxy. Relaxed (lossy) verification accepts tokens that lossless verification would reject, potentially altering that distribution. We identify a length inflation phenomenon: relaxed verification lengthens completions, primarily by extending reasoning traces, largely offsetting apparent gains in accepted tokens per round. We evaluate existing training-free relaxed verification rules on twelve target-drafter pairs against lossless verification on identical prompts, measuring completion length, verifier rounds, wall-clock time, and task accuracy. At each rule's most relaxed evaluated setting, relaxed verification on average accepts $1.34\times$ as many tokens per round but produces $1.54\times$ longer completions. For rules with higher acceptance lengths, tokens accepted only under relaxed verification tend to have low target-model probability, contributing to length inflation. For rules that reduce verifier rounds, accuracy typically drops substantially. Further analysis traces some failures to continued reasoning that exhausts the budget needed to deliver a final answer. Overall, accepted tokens per round alone is an unreliable measure of decoding speed under relaxed verification. Completion length and verifier rounds must also be reported to assess actual efficiency. [COMMENTS]Submitted for NAACL 2027, under review [LINK]http://arxiv.org/abs/2610.23818v1 [DATE]2026-10-17 01:41:19+08:00 [CATEGORIES]cs.CL
2026 Oct 17, Sat
VocalGrad: Evaluating Acoustic Perception in Audio Language Models [AUTHORS]Koki Ryu, Hitomi Yanaka [ABSTRACT]Recent large audio language models (LALMs) have extended their capabilities beyond lexical speech understanding to tasks requiring paralinguistic and non-linguistic audio understanding, such as emotion recognition. However, these models still struggle to reliably utilize acoustic information, particularly when it co-occurs with spoken linguistic content. Existing benchmarks do not clearly distinguish acoustic perception from higher-level reasoning, making it difficult to identify the source of these failures. To address this issue, we introduce VocalGrad, a benchmark designed to evaluate LALMs' ability to perceive low-level acoustic variations. VocalGrad formulates this problem as a simple binary classification of temporal change direction, enabling evaluation largely independent of higher-level reasoning or semantic understanding. Our benchmark results reveal a substantial gap in acoustic perception between human annotators and state-of-the-art LALMs: while human annotators achieve high accuracy, LALMs perform at near-chance levels. We further conduct linear probing to identify structural bottlenecks in these models. Our analysis reveals two distinct failure modes: for some categories, task-relevant information is preserved in internal representations but not utilized by the language model head; for others, the information becomes progressively less linearly decodable across model layers. Overall, our results reveal substantial limitations of LALMs in modeling acoustic variation under controlled settings and expose representation differences not observable from text outputs. [COMMENTS]NeurIPS 2026 (Evaluations & Datasets Track). Dataset page: https://huggingface.co/datasets/ynklab/vocalgrad [LINK]http://arxiv.org/abs/2610.23745v1 [DATE]2026-10-17 00:54:35+08:00 [CATEGORIES]cs.CL
Structural Causal Bottleneck Models [AUTHORS]Simon Bing, Jonas Wahl, Jakob Runge [ABSTRACT]We introduce structural causal bottleneck models (SCBMs), a novel class of structural causal models in which causal effects between high-dimensional variables are mediated by low-dimensional summary statistics, or *bottlenecks*. SCBMs provide a flexible framework for mechanism-specific, target-dependent dimension reduction while remaining estimable via standard learning algorithms. We prove that bottleneck variables are identifiable up to bijection from observational data, and validate this experimentally. We then study causal effect estimation in linear SCBMs under three standard adjustment strategies. For instrumental variables, we identify a population-level non-identifiability of naïve two-stage least squares when all variables are high-dimensional, and show that bottleneck representations restore identification. For mediation and backdoor adjustment, we prove that conditioning on bottleneck variables yields valid effect estimates with explicit finite-sample efficiency gains over conditioning on full high-dimensional variables. We argue that SCBMs provide a principled alternative to existing causal dimension reduction frameworks such as causal representation learning and causal abstractions. [COMMENTS]NeurIPS 2026 [LINK]http://arxiv.org/abs/2603.08682v3 [DATE]2026-10-17 01:56:06+08:00 [CATEGORIES]cs.LG
TANGO: RNA Topology and Geometry Co-Design [AUTHORS]Tianmeng Hu, Biao Luo, Ke Li [ABSTRACT]Designing RNA sequences that satisfy target tertiary-structure objectives is an important problem in RNA engineering, with broad applications in synthetic biology and therapeutics. Existing 3D RNA design methods have made substantial progress, but they largely treat the target fold as a geometry-only objective. This is incomplete for RNA, whose functional folds are specified not only by tertiary geometry but also by secondary base-pairing topology. We therefore formulate RNA inverse design as a topology-and-geometry co-design problem and propose TANGO, a multi-objective multi-agent reinforcement learning framework for this task. TANGO adopts a staged curriculum: it first learns a topology-aware policy from secondary-structure targets, then transfers this policy to tertiary-structure targets to train a general co-design model, and finally applies frontier-guided fine-tuning to refine target-specific trade-offs among secondary-structure fidelity, tertiary-structure fidelity, and sequence diversity. Experiments show that TANGO generates diverse and novel designs, improving tertiary- and secondary-structure performance by 18.3% and 52.7%, respectively, over the strongest baseline. [COMMENTS]NeurIPS 2026 [LINK]http://arxiv.org/abs/2610.20277v2 [DATE]2026-10-17 01:46:35+08:00 [CATEGORIES]cs.LG
Neural Expansion: A Unified Mechanism for How Deep Neural Networks Generalize [AUTHORS]Chashi Mahiul Islam, Samuel Jacob Chacko, Mao Nishino, Canlin Zhang, Xiuwen Liu [ABSTRACT]Deep neural networks demonstrate remarkable generalization despite being highly over-parameterized, challenging traditional statistical learning theories. While several theoretical paradigms such as margin-based bounds, PAC-Bayesian analyses, and neural tangent kernels provide global explanations, they often lack mechanistic insight into how generalization emerges from training dynamics. We introduce Neural Expansion, a unified geometric framework that characterizes generalization through the expansion of directionally-selective ray coverage in the input space. Using Generalization Intervals (GIs), we quantify direction-specific regions around training samples where the model's predictions remain stable. Our extensive empirical studies across diverse architectures and datasets reveal that training leads to progressive expansion of ray coverage, particularly along directions with low Jacobian singular values. This mechanism helps explain a large fraction of correctly classified test inputs, adversarial examples, and even mislabeled or out-of-distribution samples through the geometry of ray coverage rather than instance-level memorization. By connecting input-space curvature to predictive behavior, our framework provides a mechanistic foundation that complements existing global theories of deep learning. [COMMENTS]Accepted at NeurIPS 2026. Includes supplementary material. Code: https://github.com/meetchashi/neural_expansion_neurips_2026 [LINK]http://arxiv.org/abs/2610.23716v1 [DATE]2026-10-17 00:38:02+08:00 [CATEGORIES]cs.LG
2026 Oct 16, Fri
Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching [AUTHORS]Roy Miles, Aysim Toker, Andreea-Maria Oncescu, Jiankang Deng, Ismail Elezi [ABSTRACT]Reasoning with large language models often benefits from generating multiple chains-of-thought, but existing aggregation strategies are typically trajectory-level (e.g., selecting the best trace or voting on the final answer), discarding useful intermediate work from partial or "nearly correct" attempts. We propose Stitching Noisy Diffusion Thoughts, a self-consistency framework that turns cheap diffusion-sampled reasoning into a reusable pool of step-level candidates. Given a problem, we (i) sample many diverse, low-cost reasoning trajectories using a masked diffusion language model, (ii) score every intermediate step with an off-the-shelf process reward model (PRM), and (iii) stitch these highest-quality steps across trajectories into a composite rationale. This rationale is then used to recompute only the final answer. This modular pipeline separates exploration (diffusion) from evaluation and solution synthesis, avoiding monolithic unified hybrids while preserving broad search. Across math reasoning benchmarks, we find that step-level recombination is most beneficial on harder problems, and ablations highlight the importance of the final solver in converting stitched but imperfect rationales into accurate answers. Using low-confidence diffusion sampling with parallel, independent rollouts, our training-free framework improves average accuracy by up to 23.8% across six math and coding tasks. At the same time, it achieves up to a 1.8x latency reduction relative to both traditional diffusion models (e.g., Dream, LLaDA) and unified architectures (e.g., TiDAR). The code is available https://github.com/roymiles/diffusion-stitching [COMMENTS]NeurIPS 2026. Code available at https://github.com/roymiles/diffusion-stitching [LINK]http://arxiv.org/abs/2602.22871v3 [DATE]2026-10-16 21:55:50+08:00 [CATEGORIES]cs.CL
Do Speech BCIs Need Larger Models? Rethinking Neural Decoding beyond Scaling [AUTHORS]Zehui Feng, Weichuan Wang, Xiaohan Chen, Cuntai Guan, Ting Han [ABSTRACT]Recent speech brain-computer interfaces (BCIs) increasingly rely on large models and complex multi-stage pipelines for neural decoding. However, in speech BCI, where training data are scarce, noisy, and expensive to collect, the benefits of model scaling remain unclear. This work revisits a fundamental question: are large models truly necessary for accurate neural decoding? We propose BEST (Brain-Encoding-Speech-Text), a two-stage framework for efficient neural-to-text decoding. Instead of relying on large generative models or phoneme-centric pipelines, BEST introduces hierarchical intermediate representations and a compact speech-oriented decoder that directly maps neural activity to text. This design preserves essential structure while significantly reducing computational cost, enabling robust generalization in low-resource settings and practical deployment under constrained resources. On the Brain-to-Text '24 and '25 benchmarks, BEST achieves word error rates (WERs) of 6.19% and 1.62%, respectively, using only 3.85% of the parameters required by comparable methods. Beyond these results, we provide three key insights: (1) decoder scaling yields non-monotonic gains; (2) cross-modal alignment is more critical than model size; and (3) overfitting degards sequence-level performance. Representation analysis further reveals a consistent neural-audio-text alignment pathway. [COMMENTS]25 pages, Accepted by NeurIPS 2026 as a Main Track conference paper [LINK]http://arxiv.org/abs/2610.23175v1 [DATE]2026-10-16 17:12:58+08:00 [CATEGORIES]cs.CL
Deconstructing Multi-Task Active Learning: The Paradox of Gradient Conflict and Orthogonal Decomposition [AUTHORS]Juho Jung, Seongjae Kang, Jinyoung Han [ABSTRACT]Existing multi-task active learning (MTL-AL) methods predominantly rely on linear scalarization of heterogeneous objectives such as uncertainty, diversity, and gradient conflict. We argue that this entangled formulation is fundamentally limited, as linear combination fails to capture Pareto-optimal trade-offs under conflicting objectives. We propose a shift from combination to deconstruction and introduce a principled framework that rethinks MTL-AL along three orthogonal dimensions. First, we identify the Conflict Paradox: although the optimization literature treats gradient conflict as interference to be suppressed, penalizing conflict when ranking candidates is measurably worse than not scoring it at all. We therefore decouple the two concerns: the acquisition function does not score conflict, and the resulting interference is resolved by gradient surgery during training. Second, we introduce Orthogonal Decomposition, separating acquisition into Informativeness and Exploration. This two-stage process, informativeness filtering followed by coverage maximization, prevents "diversity dilution" by excluding uninformative outliers before enforcing diversity. Third, we propose Cross-Task Mutual Information (CTMI), a covariance-based statistic of cross-task dependency computed from MC-dropout task entropies. Extensive experiments on COCO, Cityscapes, and NYUv2, including comparisons against recent MTL-AL and data-selection baselines, demonstrate consistent improvements over state-of-the-art methods, highlighting that structural rethinking of acquisition yields larger gains than incremental heuristic design. Code is available at https://github.com/Juho-Jung/multitask_active_learning_official. [COMMENTS]10 pages, 5 figures, NeurIPS 2026 [LINK]http://arxiv.org/abs/2610.23623v1 [DATE]2026-10-16 23:32:37+08:00 [CATEGORIES]cs.LG
Leaner Transformers Can Easily Learn to Cluster [AUTHORS]Charlotte Park, Kenneth L. Clarkson, Lior Horesh, Takuya Ito, Parikshit Ram [ABSTRACT]Transformers have in-context learning capabilities, where some known learning algorithms can be executed in the forward pass through the model. Recent work shows that transformers can exactly perform Lloyd's algorithm for $k$-means clustering with $n$ points in $d$ dimensions with an embedding size $d_\{\textsf\{emb\}\} = d+k$ (thus, requiring attention projection matrices of size $(d+k)^2$). In this work, we build upon this result in the following ways: First, we present an equally expressive but smaller transformer that executes Lloyd's algorithm with embedding size $d_\{\textsf\{emb\}\} = (d + \lceil \log_2 k \rceil)$. Next, we train these transformers to learn the clustering algorithms given a distribution of clustering tasks, and theoretically characterize and empirically validate the factors affecting the convergence and in-distribution generalization of learning algorithms based on stochastic gradients. Finally, we probe the general clustering abilities of these learned algorithms (in the form of transformers), and try to understand situations where they succeed and fail. [COMMENTS]NeurIPS 2026 accepted paper [LINK]http://arxiv.org/abs/2610.09760v2 [DATE]2026-10-16 22:22:28+08:00 [CATEGORIES]cs.LG
Cross-Dialect Generalization Without Retraining: Benchmarks and Evaluation of Schema-Derived Constrained Decoding for MLIR [AUTHORS]Plawan Kumar Rath [ABSTRACT]Multi-Level Intermediate Representation (MLIR) underlies modern ML compiler infrastructure (TensorFlow, JAX/StableHLO, PyTorch Inductor, IREE), yet appears only in trace amounts in code-LM pretraining corpora. MLIR is also extensible by design: new dialects ship per application domain, so a fine-tuned model per dialect does not scale. We ask whether inference-time priors derived mechanically from each dialect's Operation Definition Specification (ODS) can substitute for gradient-based adaptation. First, we release four natural-language-to-MLIR benchmarks across three dialects (MLIR-Spec-150, Linalg-Spec-30, StableHLO-Spec-30, StableHLO-Held-Out-200), totaling 410 in-scope NL-to-MLIR pairs, plus a 25-program out-of-grammar stress set, with functional references on 30 Spec prompts, under Apache-2.0 with Gebru datasheets and Croissant 1.1 metadata. Second, we build a three-layer schema-derived constraint stack: a CFG over op signatures (C1), type-domain splits from an ODS-extracted type lattice (C2), and an SSA-scope validator driving five-retry rejection sampling (C3). Porting it to StableHLO required no new constraint-layer code. On structurally dominated dialects, SmolLM2-1.7B with token-level constraints matches or exceeds 15B-34B code LMs running post-hoc rejection at a matched retry budget and 8-25x the per-generation speed: on linalg it reaches 80.0% verify-valid (three runs, n=125), 21 to 44 pp above CodeLlama-34B, Granite-Code-34B, and StarCoder2-15B with non-overlapping CIs. The comparison is between systems, not models: under the baselines' post-hoc protocol, SmolLM2 falls from 52.0% to 24.5% on arith+func. On arith+func and the templated StableHLO-Held-Out-200 the baselines match or beat the SLM; we scope these as non-win cells. We release benchmarks, decoder, every per-prompt generation, and a pinned reproducibility Dockerfile. [COMMENTS]Accepted to the NeurIPS 2026 Evaluations and Datasets Track; camera-ready version. 31 pages (10 main + references + appendix + checklist). Code: https://github.com/plawanrath/slm-to-mlir-constrained-emitter Datasets: https://huggingface.co/collections/plawanrath/cross-dialect-mlir-benchmarks-neurips-2026-e-and-d [LINK]http://arxiv.org/abs/2607.18254v2 [DATE]2026-10-16 22:07:14+08:00 [CATEGORIES]cs.LG
2026 Oct 17, Sat
When Is Language-Model Advice Worth Buying? A Matched-Budget Study of Predictive Gating in Reinforcement Learning [AUTHORS]Ibne Farabi Shihab, Md Najmus Swaqeeb, Abu Sa-Adat Mohamed Moon-Im Al Ahsan, Joyanta Jyoti Mondal [ABSTRACT]Language-model advice can accelerate reinforcement learning, but every call has a price, and a returned suggestion may be unparsable, unsafe to execute, or no better than the agent's own action. We introduce certified predictive advice gating, a cost-aware controller that forecasts possible advisor responses, values the action actually executed after parsing, execution checks, and cache reuse, and requests advice only when a calibrated lower bound on that value covers the price. Our analysis separates the gap of executed advice, cumulative regret under an intervention-stable fallback, and myopic acquisition loss. On BabyAI GoToObj with Qwen2.5-1.5B and 7B advisors, the gate improves return over no advice by paired $0.029\pm0.016$ and $0.030\pm0.015$ while using more than $97\%$ fewer calls than always asking. Matched-budget comparisons, however, do not resolve an advantage over an early query schedule, so the gain reflects economical advice volume rather than demonstrated state selection. Groupwise calibration raises coverage on positive-value opportunities from $0.47$ to $0.85$, still below the $0.90$ target. A separate headroom-aware gate reports paired gains over an early schedule of $0.032\pm0.012$ and $0.029\pm0.011$ on GoToObj and $0.027\pm0.011$ and $0.028\pm0.012$ on GoToObjDoor, and adds BabyAI and ALFWorld success, decision-stratum coverage, and component timing. These results separate return gains, query placement, and execution screening; finite-sample deployment guarantees remain unestablished. [COMMENTS]41 pages, 6 figures [LINK]http://arxiv.org/abs/2609.29548v2 [DATE]2026-10-17 00:49:25+08:00 [CATEGORIES]cs.LG
2026 Oct 16, Fri
Dependency Reasoning for Aspect Sentiment Quadruple Prediction [AUTHORS]Shima Esfandyari, Seywan Soleymanbaigi, Fardin Akhlaghian Tab, Fatemeh Daneshfar, Mourad Oussalah [ABSTRACT]Aspect Sentiment Quadruple Prediction (ASQP) is fundamentally a relational reasoning problem: identifying aspect-opinion-category-sentiment quadruples requires modeling structured dependencies, not merely sequential co-occurrence. Existing LLM-based approaches incorporate syntactic information indirectly, via retrieval-based demonstration selection or multi-stage training, rather than as unified, single-pass generation-time context. We present DRIFT, which linearizes dependency parse trees and injects them directly into LLM prompts as explicit relational context, enabling the model to reason over aspect-opinion dependencies rather than infer them from token proximity alone. DRIFT further integrates retrieval-augmented in-context learning and a syntax-aware LLM-based adjudication mechanism that complements exact-match evaluation under span-boundary variation. On SemEval Rest15, Rest16, and ACOS, DRIFT achieves the strongest performance among compared generative ASQP methods, reaching 54.47% and 64.10% F1 on Rest15 and Rest16, with consistent gains across short- and medium-distance aspect-opinion configurations; adjudication further shows exact-match metrics systematically underestimate semantically correct predictions. [LINK]http://arxiv.org/abs/2610.14873v2 [DATE]2026-10-16 20:03:17+08:00 [CATEGORIES]cs.CLcs.LG
2026 Oct 17, Sat
DiscoveryRSI: Recursively Learning to Discover Valuable Hypotheses [AUTHORS]Jianyuan Zhong, Chenwei Wu, Zhaochen Yu, Xixian Jiang, Xuewei Li, Hao Xiao, Xilin Dai, Junyang Wang, Jiaru Zou, Zhiqing Cui, Boyuan Zheng, Jing Xiong, Zhen Zhang, Qiang Gao, Michael Morley, Luis Nakayama, Qiang Xu, Liyue Shen, Tao Gui, Leo Anthony Celi, James Zou, Zhenfei Yin, Yingcheng Wu, Ling Yang [ABSTRACT]Scientific discovery requires developing worthwhile hypotheses into practical investigations, yet learning from this evolving research process remains challenging. We introduce an operational definition of valuable and testable research hypotheses: evidence-grounded claims about consequential uncertainties, coupled with feasible investigations, explicit decision criteria, and alternative explanations. Our DiscoveryRSI framework develops hypothesis trees through multi-agent evolutionary search, compiles selected investigations into executable Harbor tasks, and recursively learns through three mechanisms: hypothesis credit optimization attributes scientific assessments to documented research contributions; harness replay identifies productive strategies and recurring failures to refine research workflows; and rediscovery applies accepted improvements to new investigations, generating further experience for subsequent learning. We also introduce DFM-Bench, which provides comprehensive assessment of hypothesis quality and cumulative research progress at the levels of hypotheses, revisions, development paths and research runs. Across four life-science domains, automated evaluation shows that hypothesis quality rises from 4.86 for the base model to 5.39 after two rounds of self-improvement, with the large gains in falsifiability and data resolvability. We further validate hypothesis quality through expert review, comparing our DiscoveryRSI-27B with other research agents. Preliminary human assessments show that DiscoveryRSI-27B scores higher than GPT-6 Luna on hypothesis importance and whether claims are plausible and evidence-bounded. DiscoveryRSI-27B achieved a 6.3% rate of hypothesis correspondence with research published after the model cutoff, largely outperform GPT-6 with 2.5% and Claude Opus 5 with 2.3%, while delivering greater human-rated improvement along research paths. [COMMENTS]Code: https://github.com/Gen-Verse/DiscoveryRSI [LINK]http://arxiv.org/abs/2610.23839v1 [DATE]2026-10-17 01:56:35+08:00 [CATEGORIES]cs.CL
TANGO: RNA Topology and Geometry Co-Design [AUTHORS]Tianmeng Hu, Biao Luo, Ke Li [ABSTRACT]Designing RNA sequences that satisfy target tertiary-structure objectives is an important problem in RNA engineering, with broad applications in synthetic biology and therapeutics. Existing 3D RNA design methods have made substantial progress, but they largely treat the target fold as a geometry-only objective. This is incomplete for RNA, whose functional folds are specified not only by tertiary geometry but also by secondary base-pairing topology. We therefore formulate RNA inverse design as a topology-and-geometry co-design problem and propose TANGO, a multi-objective multi-agent reinforcement learning framework for this task. TANGO adopts a staged curriculum: it first learns a topology-aware policy from secondary-structure targets, then transfers this policy to tertiary-structure targets to train a general co-design model, and finally applies frontier-guided fine-tuning to refine target-specific trade-offs among secondary-structure fidelity, tertiary-structure fidelity, and sequence diversity. Experiments show that TANGO generates diverse and novel designs, improving tertiary- and secondary-structure performance by 18.3% and 52.7%, respectively, over the strongest baseline. [COMMENTS]NeurIPS 2026 [LINK]http://arxiv.org/abs/2610.20277v2 [DATE]2026-10-17 01:46:35+08:00 [CATEGORIES]cs.LG
RCPN: towards robust pseudo-labels in unsupervised CLIP adaptation via consistent prototypes and neighborhoods [AUTHORS]Eman Ali, Chetan Arora, Muhammad Haris Khan [ABSTRACT]In unsupervised adaptation for vision-language models such as CLIP, pseudo-labels derived from zero-shot predictions often exhibit significant noise, particularly under domain shifts or in visually complex scenarios. Conventional pseudo-label filtering approaches, which rely on fixed confidence thresholds, tend to be unreliable in fully unsupervised settings. In this work, we propose a novel adaptive pseudo-labeling framework that enhances CLIP's adaptation performance by integrating prototype consistency and neighborhood-based consistency. The proposed method comprises two key components: PICS, which assesses pseudo-label accuracy based on in-class feature compactness and cross-class feature separation; and NALR, which exploits semantic similarities among neighboring samples to refine pseudo-labels dynamically. Additionally, we introduce an adaptive weighting mechanism that adjusts the influence of pseudo-labeled samples during training according to their estimated correctness. Extensive experiments on 11 benchmark datasets demonstrate that our method achieves state-of-the-art performance in unsupervised adaptation scenarios, delivering more accurate pseudo-labels while maintaining computational efficiency. [LINK]http://arxiv.org/abs/2507.22075v2 [DATE]2026-10-17 01:28:20+08:00 [CATEGORIES]cs.LG
Mesoscopic learning dynamics of extensive-width neural networks [AUTHORS]Shu Zhou, Jean Barbier, Sebastian Goldt [ABSTRACT]Understanding how stochastic gradient descent learns internal representations in wide neural networks trained on rich target functions is a central problem in machine learning. We study this question in two-layer neural networks learning target functions represented by networks of the same architecture, with arbitrary nonlinearity, and with network width and input dimension growing proportionally. In this regime, standard approaches for describing the learning dynamics require tracking a diverging number of overlaps, which measure correlations between learned and target features, leaving no obvious description in terms of a few order parameters. Here, we develop a mesoscopic description of learning by characterizing the distribution of overlaps. In the early stages of learning, we find that a Gaussian description of overlap statistics reduces the learning dynamics to a few collective variables that track learning of the target function's linear and quadratic components. We then estimate the timescale of specialization, when Gaussian closure breaks down and individual learned features begin to strongly correlate with specific target features. We find that specialization in extensive-rank networks is accelerated via a quadratic shielding mechanism, whereby learning the linear and quadratic components of the target function filters gradient fluctuations. We finally describe how the activation function determines the pattern of specialization and characterize the subsequent refinement of learned features. Together, these results show how collective learning gives way to individual feature discovery and clarify when learning in large neural networks admits a simple statistical description. [LINK]http://arxiv.org/abs/2610.23779v1 [DATE]2026-10-17 01:12:43+08:00 [CATEGORIES]cs.LG
Equivalence testing and run-count design for stochastic machine-learning experiments [AUTHORS]Dmytro Knopov, Nadiya Shvai [ABSTRACT]Claims that two training methods perform equally well often rest on a non-significant difference between two or three runs per method. We develop a protocol, with its power theory, for certifying practical equivalence in per-run metrics such as the AUROC (area under the receiver operating characteristic curve). It rests on a two-level (seed x item) variance model separating between-run dispersion from item-level noise. Equivalence within a pre-specified margin is tested by two one-sided tests (TOST), equivalently by a confidence interval inside the margin; a two-level bootstrap covers composite statistics. A closed-form formula converts a margin and a target power into runs per method; its shifted-t refinement is calibrated against the realised two-level process. Simulations on a 270-cell grid and under non-Gaussian seed laws confirm the formula within its design-relevant range, from five runs per method. For four published runs of a joint energy-based model, the failure to certify equivalence was largely decided by design: with the variance calibrated to those runs, two per method certify with probability near three per cent, whereas certifying 0.01 AUROC at 80 per cent power needs twelve even under exact equality. A second case study on a thousand-seed CIFAR-10 zoo and the MultiBERTs checks the formula against real seed laws (within 0.05 in that range) and resolves comparisons that two or three runs leave undecided. In a two-stage design (pilot, then formula), our experiments suggest that an upper confidence bound on the pilot variance substantially reduces the undersizing seen with its point estimate. [COMMENTS]31 pages, 4 figures, 4 tables. Code, simulation scripts and case-study data: https://github.com/DmytroKnopov/equivtest (release v0.2.0, archived at https://doi.org/10.5281/zenodo.23421510) [LINK]http://arxiv.org/abs/2610.23700v1 [DATE]2026-10-17 00:28:12+08:00 [CATEGORIES]cs.LG
2026 Oct 16, Fri
Referring Multi-Object Tracking with Diverse Dynamic Expressions [AUTHORS]Yani Zhang, Dongming Wu, Wencheng Han, Xingping Dong, Shengcai Liao, Bo Du [ABSTRACT]Referring understanding is a fundamental task that bridges natural language and visual content by localizing objects described in free-form expressions. However, existing works are constrained by limited language expressiveness, lacking the capacity to model object dynamics in spatial numbers and temporal states. To address these limitations, we introduce a new and general referring understanding task, termed referring multi-object tracking (RMOT). Its core idea is to employ a language expression as a semantic cue to guide the prediction of multi-object tracking, comprehensively accounting for variations in object quantity and temporal semantics. Along with RMOT, we introduce a RMOT benchmark named Refer-KITTI-V2, featuring scalable and diverse language expressions. To efficiently generate high-quality annotations covering object dynamics with minimal manual effort, we propose a semi-automatic labeling pipeline that formulates a total of 9,758 language prompts. In addition, we propose TempRMOT, an elegant end-to-end Transformer-based framework for RMOT. At its core is a query-driven Temporal Enhancement Module that represents each object as a Transformer query, enabling long-term spatial-temporal interactions with other objects and past frames to efficiently refine these queries. TempRMOT achieves state-of-the-art performance on both Refer-KITTI and Refer-KITTI-V2, demonstrating the effectiveness of our approach. The source code and dataset is available at https://github.com/zyn213/TempRMOT. [LINK]http://arxiv.org/abs/2406.05039v3 [DATE]2026-10-16 22:34:19+08:00 [CATEGORIES]cs.CL
RubricEM: Meta-RL with Rubric-guided Policy Decomposition beyond Verifiable Rewards [AUTHORS]Gaotang Li, Bhavana Dalvi Mishra, Zifeng Wang, Jun Yan, Yanfei Chen, Chun-Liang Li, Long T. Le, Rujun Han, George Lee, Hanghang Tong, Chen-Yu Lee, Tomas Pfister [ABSTRACT]Deep research agents plan, search, evaluate evidence, and synthesize long-form reports. Training these agents with reinforcement learning (RL) is challenging because they operate over long horizons and their outputs lack exact answer verifiers. To this end, we introduce RubricEM, a unified RL framework for deep research that combines stagewise credit assignment with reflection-based meta-RL for experience reuse. Rubrics unify the framework by providing a shared interface for policy execution, judge feedback, and experience reuse. RubricEM first instills a rubric-guided scaffold through supervised fine-tuning, teaching the agent to plan with rubrics and organize its trajectory into semantic stages. Building on this structural prior, we optimize the task policy with Stage-Structured GRPO, using stagewise judge feedback for finer-grained credit assignment. In parallel, we asynchronously train a shared-backbone reflection meta-policy (with meta-RL) to distill completed trajectories into reusable guidance, contributing through both shared-parameter updates and textual memory. The resulting RubricEM-8B model achieves strong performance across four long-form research benchmarks, outperforming existing open models of larger sizes and approaching proprietary deep-research systems. Beyond final performance, we perform thorough controlled analyses to understand the key ingredients of RubricEM. [COMMENTS]67 pages, 6 figures [LINK]http://arxiv.org/abs/2605.10899v2 [DATE]2026-10-16 19:35:29+08:00 [CATEGORIES]cs.CLcs.LG
Think More or Read More? A Token-Budget Theory of Reasoning with Shared Context [AUTHORS]Manh Nguyen, Sunil Gupta, Hung Le [ABSTRACT]Reasoning systems increasingly improve answers by incorporating additional context into the model's input, from multi-agent debate, where agents read their peers' responses, to self-refinement, where a model revisits its own drafts. In these systems, communication consumes context as one agent's output becomes another's input. We study the regime where token reading and generation draw on one fixed token window, so every token spent reading is a token that cannot be spent reasoning. We develop a \textbf\{token-budget theory\} that formalizes this trade-off, modeling accuracy as a function of the reasoning budget remaining after reading and aggregating votes with a Condorcet jury model that permits correlated errors. For multi-agent debate, the theory gives a closed-form prediction for the optimal number of agents. This quantity is estimated from single-agent measurements, allowing the optimal group size to be determined without running debate, and the same accounting extends to a heterogeneous set of different models. We empirically verify our theory and show that, at the predicted group size, our method outperforms strong baselines under the tightest calibration spend, where debate-calibrated rules can afford only one or two questions. Self-refinement obeys the same accounting, with its own draft in place of a peer message. The theory further predicts that communication pays only when the gain in the accuracy ceiling exceeds the reasoning room spent on reading, with this condition tightening as the window shrinks. Across five open models and four reasoning tasks, self-consistency matches or beats deployed debate at $93$\% of $980$ matched window-and-group-size settings, with the gap being largest under tight windows and nearly disappearing at a loose one, consistent with the theory. Under a shared window, how much room a model has left to \textit\{think matters more than what it reads\}. [LINK]http://arxiv.org/abs/2610.23246v1 [DATE]2026-10-16 18:18:15+08:00 [CATEGORIES]cs.CLcs.LG
ReLaDrift: Recurrent Language Generation via Drifting for Test-Time Scaling [AUTHORS]Yuki Otsuka, Daisuke Oba, Naoaki Okazaki [ABSTRACT]Drifting Models generate a sample with a single function evaluation, making them attractive for fast parallel language generation, but their one-step formulation does not support iterative refinement. We propose Recurrent Language Generation via Drifting (ReLaDrift), which uses a shared Drifting Model to generate an initial draft from noise and iteratively refine it in a continuous latent space. We train this recurrence by sampling the number of rounds and applying the drifting loss only at the final round, stopping gradients through preceding rounds so that only one round requires backpropagation. Unlike the one-step Drifting Model, whose accuracy does not improve when iterated, ReLaDrift enables a trade-off between test-time computation and generation quality, as diffusion and flow language models do. On Sudoku Hard and GSM8K, it achieves competitive or superior performance compared with these iterative-generation approaches across a range of inference budgets. Our code is available at https://github.com/Yuki-Ots/ReLaDrift [LINK]http://arxiv.org/abs/2610.23225v1 [DATE]2026-10-16 17:55:22+08:00 [CATEGORIES]cs.CL
When Magnitude Suffices: Reconstructing Time-Domain-Gated Microwave Signals Without Phase on a Thoracic Phantom [AUTHORS]Maciej Cichoń, Patrycja Gołaś, Maksymilian Bartosik, Angelika Walczak, Agata Bielecka-Dąbrowa, Maciej Ślot [ABSTRACT]Time-domain gating separates the arrivals that form a microwave S-parameter measurement. It requires the measurement of complex values, so this application almost always uses a vector network analyzer (VNA). We show that, on one calibrated measurement configuration, the gated magnitude the VNA would give can be reconstructed from a magnitude-only sweep of a spectrum analyzer with a tracking generator (SA). We select the gate center for separability of the measured states, correct each instrument once by a per-bin subtraction of a reference state, and train a fully connected network on sweeps from both instruments. The test object is a torso phantom with lung-shaped compartments whose filling takes eleven dielectric states, each held out in turn. On $S_\{11\}$ the reconstruction from SA input reaches a shape error of 0.08 dB$^2$ with a state ordering of 0.81. This matches the reconstruction from VNA input. On $S_\{21\}$ it reaches 0.82 dB$^2$ and keeps the ordered groups of the states. An explicit phase prediction performs much worse, because gating needs the phase to about 0.2 rad RMS. Across eleven measurement configurations the dependence of phase on magnitude is strong for reflection and weak for transmission. This explains the difference between $S_\{11\}$ and $S_\{21\}$. Where the magnitude carries enough of the phase, a scalar instrument gives the gated response without any phase measurement after installation. On the open UM-BMID breast data set, with whole phantoms held out, the shape error of the network is 30 percent lower than that of a per-bin linear map. [COMMENTS]18 pages, 7 figures, 4 tables [LINK]http://arxiv.org/abs/2610.23608v1 [DATE]2026-10-16 23:22:35+08:00 [CATEGORIES]cs.LG
2026 Oct 17, Sat
Beyond the Answer: Student-Conditioned Trace Construction for Reasoning Distillation [AUTHORS]Shaoang Li, Yanhang Shi, Yufei Li, Mingfu Liang, Frank Shyu, Sandeep Pandey, Luke Simon, Xi Liu, Jian Li [ABSTRACT]A reasoning trace can accompany a correct answer without providing effective supervision for a particular student. We study how to construct traces that improve learning beyond answer-only supervision by considering both their intermediate content and their accessibility to the target learner. We introduce ReDiST, a student-conditioned trace-construction and distillation framework for evidence-rich reasoning. Starting from the student's own attempts, ReDiST revises answer-correct drafts against the supplied evidence and repairs unsuccessful drafts with additional guidance from the verified answer. A frozen student-matched reader screens and ranks original and edited candidates using answer recovery and verified-answer probability, given the question, evidence, and candidate trace. We distill the selected targets while balancing trace and final-answer supervision with an auxiliary answer loss. Construction and selection occur offline, and inference uses only the distilled student. Experiments on question-answering benchmarks show that ReDiST can improve both source-task and cross-benchmark performance over answer-only, original-trace, and the evaluated trace-selection and rewriting baselines. Across supervision recipes, performance rankings can differ between source and transfer benchmarks, highlighting the importance of evaluating trace teaching utility across distributions. [LINK]http://arxiv.org/abs/2610.23831v1 [DATE]2026-10-17 01:52:58+08:00 [CATEGORIES]cs.CLcs.LG
Automatic Tensor-level Execution of Abstract Interpreters for Neural Networks [AUTHORS]Avaljot Singh, Yasmin Chandini Sarita, Aditya Mishra, Gagandeep Singh, Charith Mendis [ABSTRACT]Developing new abstract interpretation-based DNN certifiers is difficult because their mathematical design is expressed at the neuron level, while implementations must operate at the tensor level. Existing libraries such as auto_LiRPA, IVAN, and ERAN support fixed sets of certifiers; extending them requires expertise in formal methods and tensor programming. Domain-specific languages such as ConstraintFlow let researchers specify certifiers at the neuron level, but cannot run them on real DNNs.
We present Neuton, a framework that automatically translates neuron-level certifier specifications into tensor-level executables, enabling new DNN certifiers without manual tensor programming. Neuton makes three contributions. First, a stack-based intermediate representation (IR) and shape analysis infer implicit tensor operations, including broadcasts, dimension expansions, and reductions, required for correct execution. Second, domain-specific rewrite rules optimize common tensor operation sequences into more efficient equivalents. Third, gBCSR is a sparse tensor format using variable-sized, layer-aligned blocks over a non-zero default value, tailored to DNN certifier sparsity.
We prove that translation preserves neuron-level semantics. We evaluate Neuton on 8 certifiers and 32 DNN benchmarks. Neuton compiles four existing certifiers and produces the first executable implementations of four others: two previously specified but never run, and two new designs. These expose precision-runtime trade-offs unavailable in existing libraries. For example, SkipPoly matches DeepPoly's precision on all 32 DNNs while running as fast or faster on half. Neuton's runtime on existing certifiers is competitive with hand-optimized libraries. Code: https://github.com/uiuc-focal-lab/constraintflow.git [LINK]http://arxiv.org/abs/2507.20055v2 [DATE]2026-10-17 01:43:47+08:00 [CATEGORIES]cs.CL
GLOBE: Evolving Structured Agentic Memory for E-commerce Product Optimization in Generative Search [AUTHORS]Hyunseo Kim, Suil Choe, Sukkyoung So, Daeyeon Yi [ABSTRACT]Optimizing product listings for generative search requires language agents not only to rewrite effectively but also to continually self-improve by retaining successful strategies across diverse categories. However, existing approaches often treat structured listings as free-form text, leaving effective strategies implicit in prompts, search trajectories, or rewritten outputs rather than accumulating them as a reusable agentic memory. To address this limitation, we introduce GLOBE, an agentic framework that drives continual self-improvement by storing optimization knowledge in an evolving handbook of explicit, field-scoped behaviors. In each episode, an Editor applies relevant behaviors and receives rank-based feedback, which a Reflector converts into behavior updates. Crucially for dependable execution in real-world settings, a Moderator removes redundant proposals, a Curator commits accepted changes, and consistently harmful behaviors are deactivated. This cycle transforms episodic feedback into a persistent, field-scoped optimization memory. Experiments show that structured behaviors outperform monolithic prompt optimization, while feedback-driven evolution and moderation further improve effectiveness and reduce downside risk. The learned behaviors also remain effective on unseen product categories, suggesting that agentic memory evolution supports interpretable and transferable product optimization. [COMMENTS]Accepted to the Second Workshop for Research on Agent Language Models(REALM) at EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23709v1 [DATE]2026-10-17 00:33:17+08:00 [CATEGORIES]cs.CL
2026 Oct 16, Fri
"You" Is Not Toxic: How LLM Labelers Rewrite the Web's Conversation Laws [AUTHORS]Liang Yue, Eric A. Vance [ABSTRACT]What the Web community believes about toxic conversations, such as that toxicity rises in longer and deeper threads, was measured mostly with Google's Perspective API, which is shutting down; large language models (LLMs) will make the next measurements. We find that LLM labelers mistake second-person address for toxicity, and that this one bias manufactures laws that human labels do not show. On 1.25 million news comments with crowd labels, Claude Haiku flags 12% of the non-toxic comments that contain no "you" and 51% of those with four or more; a single crowd rater flags 6% and 10%. Deeper replies address someone more often, so under Haiku toxicity rises with reply depth (+0.93 percentage points per level), while under the crowd's labels it slightly declines (-0.12). Rewriting "you" in the third person removes two thirds of the false rise; flagging only as many comments as the crowd does not remove it, and showing Haiku the parent comment enlarges it. All seven LLMs we test over-flag "you", and at the crowd's flag rate, without the parent, any false rise of the open ones vanishes at a fixed number of second-person words; Haiku also over-flags "you" on Reddit, Gab and Wikipedia talk pages. Agreement with human labels does not reveal the bias: we prove that a fixed Cohen's kappa leaves a law nearly unconstrained. We therefore certify the law instead of the labeler. Threadsafe uses the labeler's scores to choose which comments human raters label and to predict the others, and corrects the predictions with the human labels; linear laws are estimated without bias whatever the labeler does. In simulated audits its intervals for the laws keep nominal coverage from 250 labels on, where uniform-sample methods such as PPI++ fall short, and at 1,000 to 8,000 labels are 0.75 times as wide as those of design-based supervised learning (DSL); a thousand human labels rule out Haiku's rising law. [COMMENTS]43 pages, 4 figures, 44 tables. Labels, prompts and code: https://github.com/liyu2195/you-is-not-toxic [LINK]http://arxiv.org/abs/2610.23446v1 [DATE]2026-10-16 21:29:30+08:00 [CATEGORIES]cs.CL
2026 Oct 17, Sat
VocalGrad: Evaluating Acoustic Perception in Audio Language Models [AUTHORS]Koki Ryu, Hitomi Yanaka [ABSTRACT]Recent large audio language models (LALMs) have extended their capabilities beyond lexical speech understanding to tasks requiring paralinguistic and non-linguistic audio understanding, such as emotion recognition. However, these models still struggle to reliably utilize acoustic information, particularly when it co-occurs with spoken linguistic content. Existing benchmarks do not clearly distinguish acoustic perception from higher-level reasoning, making it difficult to identify the source of these failures. To address this issue, we introduce VocalGrad, a benchmark designed to evaluate LALMs' ability to perceive low-level acoustic variations. VocalGrad formulates this problem as a simple binary classification of temporal change direction, enabling evaluation largely independent of higher-level reasoning or semantic understanding. Our benchmark results reveal a substantial gap in acoustic perception between human annotators and state-of-the-art LALMs: while human annotators achieve high accuracy, LALMs perform at near-chance levels. We further conduct linear probing to identify structural bottlenecks in these models. Our analysis reveals two distinct failure modes: for some categories, task-relevant information is preserved in internal representations but not utilized by the language model head; for others, the information becomes progressively less linearly decodable across model layers. Overall, our results reveal substantial limitations of LALMs in modeling acoustic variation under controlled settings and expose representation differences not observable from text outputs. [COMMENTS]NeurIPS 2026 (Evaluations & Datasets Track). Dataset page: https://huggingface.co/datasets/ynklab/vocalgrad [LINK]http://arxiv.org/abs/2610.23745v1 [DATE]2026-10-17 00:54:35+08:00 [CATEGORIES]cs.CL
The Italian lawmaking process: a unified and structured dataset [AUTHORS]Matilde Contestabile, Chiara Ferrara, Giovanni Parrillo, Lorenzo Emer, Andrea Vandin [ABSTRACT]The Italian legislative process is documented in detailed parliamentary records hosted in institutional websites. However, much of this information is available only as unstructured text, limiting its systematic reuse. We applied a controlled, systematic LLM-based extraction procedure, validated against manually annotated records, to convert these texts into ordered and dated events including, e.g., bill introduction, committee examination, plenary examination, and approval. The resulting dataset reconstructs the procedural histories of all 5346 Italian laws enacted between 1985 and 2026, spanning parliamentary terms IX to XIX, for overall 101413 events. Each event records the activity, date, and chamber, with additional procedural details where available. The dataset supports research in the emerging fields of computational legal and legislative studies, potentially combined with process-oriented data science methods to examine procedural paths and legislative timing across parliamentary terms. [LINK]http://arxiv.org/abs/2610.23741v1 [DATE]2026-10-17 00:52:11+08:00 [CATEGORIES]cs.CL
Beyond Consensus: Perspectivist Modeling and Evaluation of Annotator Disagreement in NLP [AUTHORS]Yinuo Xu, David Jurgens [ABSTRACT]Annotator disagreement is widespread in NLP, particularly for subjective and ambiguous tasks such as toxicity detection and stance analysis. While early approaches treated disagreement as noise to be removed, recent work increasingly models it as a meaningful signal reflecting variation in interpretation and perspective. This survey provides a unified view of disagreement-aware NLP methods. We first present a domain-agnostic taxonomy of the sources of disagreement spanning data, task, and annotator factors. We then synthesize modeling approaches using a common framework defined by prediction targets and pooling structure, highlighting a shift from consensus learning toward explicitly modeling disagreement, and toward capturing structured relationships among annotators. We review evaluation metrics for both predictive performance and annotator behavior, and noting that most fairness evaluations remain descriptive rather than normative. We conclude by identifying open challenges and future directions, including integrating multiple sources of variation, developing disagreement-aware interpretability frameworks, and grappling with the practical tradeoffs of perspectivist modeling. [COMMENTS]Accepted to EMNLP 2026 [LINK]http://arxiv.org/abs/2601.09065v3 [DATE]2026-10-17 00:39:26+08:00 [CATEGORIES]cs.CL
GLOBE: Evolving Structured Agentic Memory for E-commerce Product Optimization in Generative Search [AUTHORS]Hyunseo Kim, Suil Choe, Sukkyoung So, Daeyeon Yi [ABSTRACT]Optimizing product listings for generative search requires language agents not only to rewrite effectively but also to continually self-improve by retaining successful strategies across diverse categories. However, existing approaches often treat structured listings as free-form text, leaving effective strategies implicit in prompts, search trajectories, or rewritten outputs rather than accumulating them as a reusable agentic memory. To address this limitation, we introduce GLOBE, an agentic framework that drives continual self-improvement by storing optimization knowledge in an evolving handbook of explicit, field-scoped behaviors. In each episode, an Editor applies relevant behaviors and receives rank-based feedback, which a Reflector converts into behavior updates. Crucially for dependable execution in real-world settings, a Moderator removes redundant proposals, a Curator commits accepted changes, and consistently harmful behaviors are deactivated. This cycle transforms episodic feedback into a persistent, field-scoped optimization memory. Experiments show that structured behaviors outperform monolithic prompt optimization, while feedback-driven evolution and moderation further improve effectiveness and reduce downside risk. The learned behaviors also remain effective on unseen product categories, suggesting that agentic memory evolution supports interpretable and transferable product optimization. [COMMENTS]Accepted to the Second Workshop for Research on Agent Language Models(REALM) at EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23709v1 [DATE]2026-10-17 00:33:17+08:00 [CATEGORIES]cs.CL
Coordination Matters: Process-Aware Evaluation of Cooperative Multi-Agent Reinforcement Learning [AUTHORS]Maria Ana Cardei, Matthew Landers, Afsaneh Doryab [ABSTRACT]Cooperative multi-agent reinforcement learning (MARL) benchmarks commonly emphasize aggregate outcomes such as return, success rate, or completion time. While essential, these metrics can obscure how agents coordinate and fail to distinguish policies that achieve similar outcomes through different interaction processes. We propose a process-aware evaluation framework that complements aggregate performance with diagnostics of interference, coordinated task completion, participation, and coordination efficiency. We evaluate the framework in STAT, a controlled spatial task-allocation testbed, and examine its broader applicability in two established cooperative MARL benchmarks. Across diverse value-based and actor-critic methods, we find policies with nearly identical return can exhibit substantial differences in coordination behavior. Controlled scaling in STAT reveals distinct process signatures for spatial, task, and agent scaling, while a reward-shaping ablation shows that explicitly penalizing assignment conflicts can alter coordination behavior with little change in task return. Results in the established benchmarks further show that the framework remains informative under different coordination mechanisms. These findings support process-aware evaluation as a complement to outcome-based benchmarking. The framework provides a common structure for selecting and interpreting environment-specific diagnostics that characterize not only what cooperative policies achieve, but how they coordinate to achieve it. [COMMENTS]Under review [LINK]http://arxiv.org/abs/2605.06557v2 [DATE]2026-10-17 01:58:51+08:00 [CATEGORIES]cs.LG
Structural Causal Bottleneck Models [AUTHORS]Simon Bing, Jonas Wahl, Jakob Runge [ABSTRACT]We introduce structural causal bottleneck models (SCBMs), a novel class of structural causal models in which causal effects between high-dimensional variables are mediated by low-dimensional summary statistics, or *bottlenecks*. SCBMs provide a flexible framework for mechanism-specific, target-dependent dimension reduction while remaining estimable via standard learning algorithms. We prove that bottleneck variables are identifiable up to bijection from observational data, and validate this experimentally. We then study causal effect estimation in linear SCBMs under three standard adjustment strategies. For instrumental variables, we identify a population-level non-identifiability of naïve two-stage least squares when all variables are high-dimensional, and show that bottleneck representations restore identification. For mediation and backdoor adjustment, we prove that conditioning on bottleneck variables yields valid effect estimates with explicit finite-sample efficiency gains over conditioning on full high-dimensional variables. We argue that SCBMs provide a principled alternative to existing causal dimension reduction frameworks such as causal representation learning and causal abstractions. [COMMENTS]NeurIPS 2026 [LINK]http://arxiv.org/abs/2603.08682v3 [DATE]2026-10-17 01:56:06+08:00 [CATEGORIES]cs.LG
Task-aligned spectral inductive biases in parameterized quantum circuits [AUTHORS]Felix Frohnert, Adrián Pérez-Salinas, Sebastian Lange, Marcus Schwaebisch, Evert van Nieuwenburg, Vedran Dunjko [ABSTRACT]Parameterized quantum circuits are widely investigated in quantum machine learning, yet how their architecture can provide task-aligned inductive biases remains poorly understood. We study data reuploading quantum circuits with two distinct inputs: a signal that varies between samples and a fixed, task-defined spectral query that specifies how the signal should be probed. We show that encoding the query can change the circuit's hypothesis class over the signal by modulating how its existing signal frequencies are combined, without introducing new sample-dependent information. This mechanism matches tasks built on nonuniform Fourier transforms, whose evaluation frequencies specify the query and whose sampling positions guide its encoding. We examine this connection in a classification task inspired by synthetic aperture radar, where acquisition geometry determines a query for classifying unfocused spectral patches before image formation. Across small-scale spectral and radar-like tasks, aligned queries improve learning from limited data over parameter-matched circuits with mismatched or absent queries. These results provide a route from known spectral task structure to the design and validation of inductive biases in parameterized quantum circuits. [LINK]http://arxiv.org/abs/2610.23834v1 [DATE]2026-10-17 01:53:19+08:00 [CATEGORIES]cs.LG
TANGO: RNA Topology and Geometry Co-Design [AUTHORS]Tianmeng Hu, Biao Luo, Ke Li [ABSTRACT]Designing RNA sequences that satisfy target tertiary-structure objectives is an important problem in RNA engineering, with broad applications in synthetic biology and therapeutics. Existing 3D RNA design methods have made substantial progress, but they largely treat the target fold as a geometry-only objective. This is incomplete for RNA, whose functional folds are specified not only by tertiary geometry but also by secondary base-pairing topology. We therefore formulate RNA inverse design as a topology-and-geometry co-design problem and propose TANGO, a multi-objective multi-agent reinforcement learning framework for this task. TANGO adopts a staged curriculum: it first learns a topology-aware policy from secondary-structure targets, then transfers this policy to tertiary-structure targets to train a general co-design model, and finally applies frontier-guided fine-tuning to refine target-specific trade-offs among secondary-structure fidelity, tertiary-structure fidelity, and sequence diversity. Experiments show that TANGO generates diverse and novel designs, improving tertiary- and secondary-structure performance by 18.3% and 52.7%, respectively, over the strongest baseline. [COMMENTS]NeurIPS 2026 [LINK]http://arxiv.org/abs/2610.20277v2 [DATE]2026-10-17 01:46:35+08:00 [CATEGORIES]cs.LG
CocoRNA: Collective RNA Design with Cooperative Multi-agent Reinforcement Learning [AUTHORS]Tianmeng Hu, Biao Luo, Ke Li [ABSTRACT]Designing RNA sequences that reliably fold into specific secondary structures is essential for understanding their biological functions but remains a challenging computational problem. We propose CocoRNA, a cooperative multi-agent reinforcement learning framework for RNA inverse design. CocoRNA simplifies the design task by decomposing it into smaller sub-problems, each solved collaboratively by multiple agents. This approach reduces the complexity of the problem and improves the exploration of design policies. During training, a centralized critic uses global structural information to guide the agents, enabling them to jointly optimize their design strategies. As a result, CocoRNA learns high-quality RNA design policies that generalize effectively to unseen structures without additional training. Experiments on the Rfam dataset demonstrate that CocoRNA substantially outperforms state-of-the-art methods in both success rate and design speed. Further experiments on other biological sequence design tasks highlight the effectiveness and broad potential of CocoRNA for complex design tasks. [COMMENTS]ICML 2026 [LINK]http://arxiv.org/abs/2610.23820v1 [DATE]2026-10-17 01:43:25+08:00 [CATEGORIES]cs.LG
GRAFT-ATHENA: Self-Improving Agentic Teams for Autonomous Discovery and Evolutionary Numerical Algorithms [AUTHORS]Juan Diego Toscano, Zhaojie Chai, George Em Karniadakis [ABSTRACT]Language-model agents plan, execute, and evaluate scientific workflows. However, the problem--method relationships guiding their choices remain implicit, so plausible outputs can violate the assumptions of the scientific problem. We introduce Graph Reduction to Adaptive Factored Trees (GRAFT), which makes this problem-to-method map explicit as an expandable probabilistic structure of admissible problems, methods, and dependencies. Because this structure follows scientific semantics, knowledge from documentation, papers, and domain experts can be encoded, while factorization keeps the structure tractable. Semantic fingerprints of encoded problems measure their similarity, so expertise and experience guide related problems. GRAFT-guided workflows exceeded published expert baselines in physics-informed learning, recovered blood-rheology trends, predicted Apollo Command Module hypersonic loading within 1.8\% of experiment, and established machine-checked error bounds for regularized brain-flow reconstruction and a spectral physics-informed architecture. The representation retains evaluated outcomes and incorporates new methods, allowing scientific knowledge and experience to accumulate. [LINK]http://arxiv.org/abs/2605.11117v3 [DATE]2026-10-17 01:10:34+08:00 [CATEGORIES]cs.LG
Logic-Guided Graph Denoising Diffusion Models [AUTHORS]Samuel Cognolato, Luciano Serafini, Alessandro Sperduti, Andrea Cavallaro [ABSTRACT]Graph diffusion models are competitive approaches for generating complex graph-structured data. However, when such distributions have strict structural constraints, diffusion models may fail to capture them exactly, resulting in the generation of invalid graphs. In this paper, we propose a method for grounding graph diffusion models in domain-specific knowledge via First Order Logic (FOL), leveraging its expressive and interpretable formalism to impose structural constraints on the generative process. We incorporate domain knowledge using classifier guidance, a technique that conditions the diffusion model's distribution on external variables to exert control over the sampled data. During the generative process, classifier guidance steers the denoising trajectory to optimize an objective function, which we define as a fuzzy and differentiable relaxation of the logical rules. Our methodology requires only the specification of constraints in FOL, whose satisfaction is encouraged at inference time to guide a pre-trained generative model. We evaluate our approach by comparing the unconditional generation of knowledge graphs with our logic-guided framework, which incorporates domain ontologies as structural constraints during inference. This approach significantly increases the proportion of generated graphs that satisfy the ontological constraints. [COMMENTS]Manuscript under review at AAAI2027 [LINK]http://arxiv.org/abs/2610.23766v1 [DATE]2026-10-17 01:06:07+08:00 [CATEGORIES]cs.LG
From Flow-Box Invariants to Global Conservation Laws: Machine Learning-Assisted Detection of Singularities and Near-Integrability [AUTHORS]Jimmie Adriazola, Ioannis G. Kevrekidis, Panayotis G. Kevrekidis, Wei Zhu [ABSTRACT]Conservation laws reveal invariant structure in dynamical systems and constrain their evolution to lower-dimensional level sets. Locally, however, per the flow-box theorem, every $d$-dimensional system generically admits $d-1$ functionally independent conservation laws near a regular point even though, generically, only a few such laws persist globally. Data-driven methods typically learn conservation laws on a prescribed region of phase space from which training samples are drawn, leaving their behavior beyond that region unresolved. We formulate this local-to-global continuation problem and develop an \{expanding-domain neural deflation\} framework that tracks the number of independent conservation laws over nested domains. The method recovers the expected $d-1$ local invariants and detects their successive breakdown as the domain expands, relaxing to the expected global conservation law limit. It distinguishes the local invariants (associated with Frobenius integrability) from persistent global structure and shows that the domain scale over which additional laws survive reflects proximity to Liouville integrability. Pointwise losses can localize isolated obstructions, whereas in more complicated dynamics the breakdown of our loss functions is gradual and spatially diffuse. The framework thus provides a data-driven approach for studying the persistence and breakdown of (so-far-learned) conservation laws across phase space. [LINK]http://arxiv.org/abs/2610.23760v1 [DATE]2026-10-17 01:02:39+08:00 [CATEGORIES]cs.LG
LagNet: Lag-Structured Operators for Long-Term Time Series Forecasting [AUTHORS]Zhenan Yu, Guangxin Jiang, Jiuyun Jiang, Jin Yang [ABSTRACT]Long-term time series forecasting maps a finite history window to an extended future horizon. Many existing models focus primarily on the design of input representations and model structures, while paying less attention to the structural form of the history-to-future forecasting operator itself. We examine this operator directly and find that, across model families, sensitivity matrices exhibit recurring diagonal patterns whose diagonals correspond to relative temporal lags. Motivated by this observation, we develop a framework for direct multi-step forecasting based on lag-structured temporal operators. Its basic building block, LagLayer, constructs a Toeplitz matrix by tying history-to-future connections with the same relative lag and incorporates horizon-wise and history-wise scaling coefficients to relax exact sharing. This design enables LagLayer to serve as an effective replacement for temporal linear layers in existing forecasting models. Stacking LagLayer modules with activations and multiple feature maps yields LagNet. Experiments on long-term time series forecasting benchmarks show that our method is competitive with strong baselines, while sensitivity matrix diagnostics and limited-data experiments demonstrate that LagLayer provides an effective and sample-efficient alternative to a linear layer. [LINK]http://arxiv.org/abs/2610.23735v1 [DATE]2026-10-17 00:44:53+08:00 [CATEGORIES]cs.LG
Diffusion-Based Stochastic Operator Networks for Uncertainty Quantification in Stochastic Partial Differential Equations [AUTHORS]Phuoc-Toan Huynh, Richard Archibald, Feng Bao [ABSTRACT]We introduce a novel framework for uncertainty quantification of solution operators associated with stochastic partial differential equations (SPDEs). Although SPDEs play a central role in modeling complex physical systems under uncertainty, their practical use typically requires specifying the magnitude and structure of model uncertainties that are often unknown and difficult to infer from noisy measurements. To address this challenge, we develop a stochastic operator-learning framework that learns directly from noisy data and outputs both a mean solution field and a quantification of uncertainty. The proposed method, namely the Stochastic Operator Network (SON), is constructed by combining the structure of the Deep Operator Network (DeepONet) with Stochastic Neural Networks (SNNs) to model stochasticity and enable probabilistic prediction. The training procedure is carried out by minimizing a Hamiltonian-type loss and optimizing the resulting objective using the Stochastic Maximum Principle. Numerical experiments on benchmark SPDEs under multiple uncertainty sources demonstrate the accuracy and robustness of the proposed method in capturing solution structure and quantifying predictive uncertainty. [LINK]http://arxiv.org/abs/2605.17107v2 [DATE]2026-10-17 00:44:26+08:00 [CATEGORIES]cs.LG
Blitz: Accelerating the Diffusion Modules of AlphaFold3 and Boltz-2 [AUTHORS]Vinay Saji Mathew, Soundar Kumara, Gretta Kellogg, William K. M. Lai [ABSTRACT]Diffusion-based protein structure prediction models produce all-atom structures in about 200 denoising steps, and that cost has driven a number of few-step samplers. We present Blitz, a few-step distillation framework that uses a geometric noise schedule to distribute steps more evenly across noise scales, then learns those updates through multistep consistency training. Physical steering and target-free candidate selection help reduce the remaining geometric defects at inference. We apply Blitz to AlphaFold 3 and Boltz-2, reducing denoising from 200 steps to eight. Because published few-step results use different benchmarks and scoring procedures, we compare both distilled models with their teachers and other few-step methods under a common evaluation protocol covering structural accuracy, interface quality, ligand placement, and stereochemical validity. Blitz retains teacher-level structural accuracy, with AF3-Blitz outperforming all evaluated few-step baselines and exceeding its teacher on several accuracy metrics. [COMMENTS]24 pages, 12 figures [LINK]http://arxiv.org/abs/2610.23703v1 [DATE]2026-10-17 00:29:04+08:00 [CATEGORIES]cs.LG
Expressivity of IQP Feature-Map Quantum Kernels via Reproducing Kernel Hilbert Spaces [AUTHORS]Valeria Repetto, Stefano Melchionna, Romina D'Aurizio [ABSTRACT]Quantum kernels are among the most promising methods in quantum machine learning, providing a natural way to embed data into high-dimensional Hilbert spaces via quantum circuits and capturing non-linear relationships in data. Their learning behaviour is governed by the function class they represent, which is formally characterized by the structure of the associated Reproducing Kernel Hilbert Space (RKHS). However, a theoretical characterization of this function class, and of how it depends on the choice of feature map and its hyperparameters, is still lacking. In this work, we focus on the Instantaneous Quantum Polynomial (IQP) feature map and derive a Fourier representation of the associated RKHS generators as finite sums of trigonometric functions evaluated on polynomially transformed input variables, with a finite set of accessible frequencies determined by the circuit construction and scaled by the bandwidth hyperparameter. This representation identifies three bandwidth regimes with qualitatively distinct expressive properties from localised, through multimodal, to periodic. Numerical experiments on synthetic datasets with controlled periodic and multimodal structure, across data dimensions and feature-map configurations, reproduce the three regimes in both supervised classification and unsupervised anomaly detection, benchmarked against a classical Gaussian kernel. These results make the relationship between expressivity and data structure explicit for IQP kernels: expressive power translates into a practical benefit only when the accessible frequencies match with the structure of the data, which provides a criterion for selecting the bandwidth when the periodic or multimodal structure of data is known. [COMMENTS]19 pages, 11 Figures , Supplementary [LINK]http://arxiv.org/abs/2610.23692v1 [DATE]2026-10-17 00:20:46+08:00 [CATEGORIES]cs.LG
2026 Oct 16, Fri
PresenceBench: An Exploratory Evaluation of Large Language Model Adaptation to User State [AUTHORS]Peña Arroyo Gallego, Mauro Ruiz Moreno, Alejandro Acien, Sergio Romero-Tapiador, Teresa Arroyo-Gallego [ABSTRACT]Presence-aware intelligence is the capacity of an AI system's intelligence layer to adapt appropriately to the functional state of the person it serves (here readiness, flow, and stability) while preserving task quality and safety. We introduce PresenceBench, an instrument for measuring it. PresenceBench evaluates how the intelligence layer uses a supplied state, not how that state is sensed, and breaks the capability into measurable properties. Two are tested here. Receptivity: does the model take up the state and give a better response than with no signal? Specificity: does any improvement depend on the correct state, rather than on extra structured context? A third, adaptation across states, is a required endpoint not measured in this release. A five-arm causal design answers both questions, against each model's own baseline and an equally structured mismatched-state control, scored by a correctness-gated human paired comparison. The exploratory evaluation covered four models across eighteen single-turn, six fixed three-turn and six simulated natural-conversation scenarios. Supplying a synthetic state, raw or translated into a response policy, gave human-preference point estimates above 0.5 for all four models (Claude Sonnet 5, DeepSeek V4 Pro, Gemini 3.6 Flash, GPT-5.6 Terra), with three of eight 95% intervals excluding chance. Specificity was not resolved: no per-model interval excluded chance, and the design cannot yet separate the state values from a generic instruction to adapt. Policy delivery reduced priced cost and output tokens against each model's baseline; these are resource-use changes, not efficiency gains, because correctness non-inferiority was not established. Safety-floor failures were rare (13 of 360 cases), with a non-significant excess under policy delivery. We report the instrument, the frozen results, and the controls a confirmatory study needs. [COMMENTS]33 pages, 6 figures, 13 tables [LINK]http://arxiv.org/abs/2610.23625v1 [DATE]2026-10-16 23:33:04+08:00 [CATEGORIES]cs.CL
The Chameleon's Limit: Investigating Persona Collapse and Homogenization in Large Language Models [AUTHORS]Yunze Xiao, Vivienne J. Zhang, Chenghao Yang, Weihao Xuan, Jen-tse Huang [ABSTRACT]Large language models (LLMs) have powered multi-agent simulations such as survey respondents or application users. However, LLMs lack the population-level diversity required by these simulations---responses from LLMs with detailed human profiles cluster into classes, unlike human responses that uniformly fill the entire space. In this paper, we formalize this as Persona Collapse and quantify its population-level structure with three diagnostic metrics: Coverage (how much of the persona space is occupied), Uniformity (how evenly agents spread across it), and Complexity (how rich the resulting behaviors are). Across eight LLMs on personality (BFI-44), moral reasoning, and self-introduction, persona collapse is universal but task-dependent: CoSER-Llama-8B collapses the most in personality while being the most diverse in moral reasoning. Moreover, sampling more personas cannot mitigate the collapse. We further show a fidelity--diversity tradeoff: the models with the highest per-persona fidelity produce the most polarized populations. Finally, persona-targeted fine-tuning further narrows the population instead of diversifying it. [LINK]http://arxiv.org/abs/2604.24698v2 [DATE]2026-10-16 23:30:51+08:00 [CATEGORIES]cs.CL
MARS: Memory-Aware Routing for Session-Level LLM Assistants [AUTHORS]Minghan Wang, Jinhang Zuo, Zhongxiang Dai, Zhiyong Wang, Fang Kong [ABSTRACT]Long-term LLM assistants must repeatedly decide what prior context to expose and which model should answer as an interaction evolves. Existing memory systems and model routers usually make these decisions separately, although memory construction changes the evidence, distractors, and difficulty of the resulting input. We introduce MARS, a memory-aware routing framework that uses session state as a shared control interface between memory construction and model selection. A Query and Topic Manager (QTM) estimates compact decision-relevant state that guides a fixed state-conditioned memory policy and an online router, enabling each turn to receive an appropriate memory view and a suitable model under a quality-memory-cost trade-off. We evaluate MARS on RealMem and MultiChallenge with 16 system configurations each, covering long-history memory-driven interaction and bounded sessions with structured context interference. On RealMem, MARS preserves 98.2% of full-context GPT-4o's QA while using only 14.5% of its generation API cost, including both generation and controller calls for MARS, and achieves the strongest measured memory-recall and retrieval scores among compared systems. On MultiChallenge, MARS trails the strongest static system overall but approaches its performance at lower recorded cost on a mechanism-focused subset. Together, these results support a favorable quality-memory-cost trade-off for long-history, memory-dependent interaction. [COMMENTS]32 pages, including references and appendices [LINK]http://arxiv.org/abs/2610.23583v1 [DATE]2026-10-16 23:04:22+08:00 [CATEGORIES]cs.CL
DominoTree: Conditional Tree-Structured Drafting with Domino for Speculative Decoding [AUTHORS]Saw S. Lin, Jyh-Shing Roger Jang [ABSTRACT]Speculative decoding speeds up large language model inference: a small drafter proposes several tokens, and the target model verifies them in one parallel pass. Drafting a tree lets the target verify many alternative continuations at once. DDTree builds such trees from the block-diffusion drafter DFlash, but scores each position independently of the tokens earlier in its branch; Domino conditions on those tokens but drafts only a single chain. DominoTree combines the two: a training-free best-first tree scored by Domino's path-dependent causal correction, restricted to top-M candidates. In a research harness at temperature zero, it averages 5.33x speedup over autoregressive (AR) decoding on Qwen3-4B and 6.89x on Qwen3-8B, peaking at 8.77x; across temperatures it is 20-26% faster than Domino on Qwen3-4B and 27-36% on Qwen3-8B, and 11-20% faster than the stronger of DDTree and its cost-aware variant CaDDTree. A controlled decomposition attributes +10.1% (Qwen3-4B) and +10.7% (Qwen3-8B) accepted length to the correction, and +4.7% and +4.4% to conditioning on the realized path. Served in SGLang on one RTX 5090, it reaches 4.11x and 4.83x AR at batch size 1, 1.17x and 1.09x the speed of Domino's chain, and up to 1.71x AR on 8K-token HELMET summarization, 14-21% above the chain. The lead holds at 2 concurrent requests but reverses from 8 onward, where the chain sustains more goodput despite DominoTree's 28% higher acceptance. DominoTree accepts more tokens than the chain in every serving setting. On the hybrid Gated-DeltaNet Qwen3.6-27B, correct output requires tree-aware recurrent verification; with it, DominoTree matches the chain at batch size 1 (4.58x versus 4.54x AR) and reaches up to 2.11x AR at long context, 15% above the chain. [COMMENTS]31 pages. Submitted to IEEE Access. v4: adds Qwen3.6-27B results, single-GPU serving re-runs, fused tree-builder kernels and tuned budgets; code and data at https://github.com/slin-zhq/Domino-Tree [LINK]http://arxiv.org/abs/2607.08642v4 [DATE]2026-10-16 22:29:03+08:00 [CATEGORIES]cs.CL
Actions with Receipts: Jointly Binding Claims, Evidence, and Execution for Replayable Tool-Agent Auditing [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Yina Sa, Daren Zha [ABSTRACT]Tool-using agents can expose citations and execution logs while leaving a critical association unaudited: whether the claim shown to a user is the claim emitted by the committed execution and supported by the cited source. A valid citation and a valid trace can therefore remain individually well formed while being transplanted across claims, actions, runs, or source versions. We introduce a claim-anchored execution contract that jointly binds the emitted claim, its exact source span, the ordered execution prefix that produced it, and the source version and access state observed by that execution. Each receipt contains an emission anchor that deterministically locates the claim inside a committed answer or claim-bearing action, together with source identifiers, offsets, hashes, quotes, and a domain-separated execution commitment. A deterministic integrity verifier reconstructs these bindings before semantic or task labels are joined. We separate this integrity plane from a pluggable support plane, so structural validity is not used as a proxy for entailment. The contract exposes seven independently testable properties: claim-emission binding, source binding, ordered-execution binding, oracle separation, persisted-object replay, execution-rerun consistency, and version/access binding. Across 1,280 cross-object attacks, the joint contract detects 1,275 substitutions (0.9961). Removing a targeted property reduces its attack-detection rate to 0.0156-0.0625. On an independently adjudicated 384-pair split, the conflict-aware support guard reaches F1 0.8865 and false acceptance 0.0729; on unseen failure families, these rates are 0.8679 and 0.0938. [COMMENTS]35 pages, 8 figures [LINK]http://arxiv.org/abs/2610.00327v2 [DATE]2026-10-16 20:28:48+08:00 [CATEGORIES]cs.CLcs.LG
ContractRL: Shielded Group-Relative Policy Optimization for Auditable Tool-Call Repair [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Yina Sa, Daren Zha [ABSTRACT]Structured tool calls often fail after only a small number of fields violate a schema or an execution contract. Regenerating the complete object enlarges the action surface and makes repeated repair difficult to audit. We introduce ContractRL, a contract-constrained sequential repair protocol that models verifier-guided JSON repair as a bounded decision process. At each step the policy observes the candidate, typed verifier feedback, JSON Pointer, immutable repair history, and remaining budget; a contract-derived action mask filters malformed or prohibited RFC-6902 operations before a deterministic validator performs the transition. We specify a contract-constrained group-relative objective for patch, retry, and abstention decisions while keeping canonical targets and semantic labels outside the online state until trace freeze. Under identical verifier information, ContractRL attains 0.9362 semantic success with 34.4 generated tokens, compared with 0.9076 and 44.9 tokens for Patch-SFT and 0.9148 and 137.2 tokens for full regeneration over 192 cases per seed and five seeds. Policy optimization improves semantic success from 0.9186 for supervised ContractRL to 0.9375. A separate three-seed paired evaluation against Patch-SFT yields a semantic difference of +0.0396 (95% CI $[+0.0137,+0.0662], p=0.0039$). Feedback, action-mask, budget, and schema-shift analyses connect these gains to localized correction, while adversarial and multi-turn evaluations characterize the remaining failure modes. [COMMENTS]29 pages, 8 figures [LINK]http://arxiv.org/abs/2610.00328v2 [DATE]2026-10-16 20:23:26+08:00 [CATEGORIES]cs.CLcs.LG
Dependency Reasoning for Aspect Sentiment Quadruple Prediction [AUTHORS]Shima Esfandyari, Seywan Soleymanbaigi, Fardin Akhlaghian Tab, Fatemeh Daneshfar, Mourad Oussalah [ABSTRACT]Aspect Sentiment Quadruple Prediction (ASQP) is fundamentally a relational reasoning problem: identifying aspect-opinion-category-sentiment quadruples requires modeling structured dependencies, not merely sequential co-occurrence. Existing LLM-based approaches incorporate syntactic information indirectly, via retrieval-based demonstration selection or multi-stage training, rather than as unified, single-pass generation-time context. We present DRIFT, which linearizes dependency parse trees and injects them directly into LLM prompts as explicit relational context, enabling the model to reason over aspect-opinion dependencies rather than infer them from token proximity alone. DRIFT further integrates retrieval-augmented in-context learning and a syntax-aware LLM-based adjudication mechanism that complements exact-match evaluation under span-boundary variation. On SemEval Rest15, Rest16, and ACOS, DRIFT achieves the strongest performance among compared generative ASQP methods, reaching 54.47% and 64.10% F1 on Rest15 and Rest16, with consistent gains across short- and medium-distance aspect-opinion configurations; adjudication further shows exact-match metrics systematically underestimate semantically correct predictions. [LINK]http://arxiv.org/abs/2610.14873v2 [DATE]2026-10-16 20:03:17+08:00 [CATEGORIES]cs.CLcs.LG
RubricEM: Meta-RL with Rubric-guided Policy Decomposition beyond Verifiable Rewards [AUTHORS]Gaotang Li, Bhavana Dalvi Mishra, Zifeng Wang, Jun Yan, Yanfei Chen, Chun-Liang Li, Long T. Le, Rujun Han, George Lee, Hanghang Tong, Chen-Yu Lee, Tomas Pfister [ABSTRACT]Deep research agents plan, search, evaluate evidence, and synthesize long-form reports. Training these agents with reinforcement learning (RL) is challenging because they operate over long horizons and their outputs lack exact answer verifiers. To this end, we introduce RubricEM, a unified RL framework for deep research that combines stagewise credit assignment with reflection-based meta-RL for experience reuse. Rubrics unify the framework by providing a shared interface for policy execution, judge feedback, and experience reuse. RubricEM first instills a rubric-guided scaffold through supervised fine-tuning, teaching the agent to plan with rubrics and organize its trajectory into semantic stages. Building on this structural prior, we optimize the task policy with Stage-Structured GRPO, using stagewise judge feedback for finer-grained credit assignment. In parallel, we asynchronously train a shared-backbone reflection meta-policy (with meta-RL) to distill completed trajectories into reusable guidance, contributing through both shared-parameter updates and textual memory. The resulting RubricEM-8B model achieves strong performance across four long-form research benchmarks, outperforming existing open models of larger sizes and approaching proprietary deep-research systems. Beyond final performance, we perform thorough controlled analyses to understand the key ingredients of RubricEM. [COMMENTS]67 pages, 6 figures [LINK]http://arxiv.org/abs/2605.10899v2 [DATE]2026-10-16 19:35:29+08:00 [CATEGORIES]cs.CLcs.LG
HyperPersona: A Multi-Level Hypergraph Framework for Text-Based Automatic Personality Prediction [AUTHORS]Sina Heydari, Majid Ramezani [ABSTRACT]As a modern commodity, language has become a vast repository of socially and psychologically significant traits and concepts, reflecting the ways people encode pattern of thoughts, behaviors, and emotions into words. Text-based Automatic Personality Prediction (APP), seeks to infer personality from linguistic behavior, offering a scalable alternative to traditional psychometric assessments. Although text is inherently hierarchical, with the document-level capturing global features, the sentence-level encoding local semantics, and the word-level providing fine-grained lexical information, most existing approaches rely on shallow, sequential, or single-level representations that ignore the multi-level structure of written language. To address this, we propose HyperPersona, a framework that explicitly models the hierarchical organization of text (document, sentence, and word) through hypergraph structure, where a document and its sentences are represented as hyperedges, and the words are represented as nodes, enabling joint modeling of global, local, and lexical dependencies of text. Followed by a transformer-based graph encoder that learns interactions within and across these linguistic layers, yielding context-sensitive and structurally grounded feature representations for personality prediction. Experiments on the Big Five personality dimensions show that, while relying solely on text, HyperPersona effectively integrates multi-level linguistic cues, achieving superior performance compared to state-of-the-art baselines. These findings underscore the critical role of textual hierarchy in advancing human-like personality inference from natural language. [COMMENTS]The manuscript is currently under peer review [LINK]http://arxiv.org/abs/2605.17355v2 [DATE]2026-10-16 19:32:35+08:00 [CATEGORIES]cs.CL
Self-Reflective Chain-of-Thought (SRCoT): Evidence-Based Text Generation via a Closed-Loop Retrieve--Reflect--Revise Pipeline on Edge LLMs [AUTHORS]Fabian Broszio, Michael Hoffmann, Adrian Paschke [ABSTRACT]Large language models (LLMs) can generate fluent outputs that lack source support or conflict with local operational constraints. We introduce Self-Reflective Chain-of-Thought (SRCoT), a modular architecture that organizes generation as a closed Retrieve--Reflect--Revise loop. SRCoT separates environment state, persistent knowledge, accumulated experience, and short-term memory into provenance-labeled stores. BM25 retrieval supplies local context; subsequent passes expose intermediate drafts, missing-information markers, follow-up queries, and optional claim--evidence mappings. These artifacts inform retrieval, revision, clarification, or abstention decisions. We examine the architecture through qualitative, trace-based case studies in two fully synthetic settings: emergency medical services and Munich-Sim, an emergency-operations simulation. Local prototype runs on an Apple M4-family laptop with 128 GB memory illustrate source-labeled documentation, explicit gap marking, and structured dispatch-style outputs. The traces also reveal provenance-assignment errors, incomplete evidence coverage, and unsafe assumptions that remain possible despite prompting constraints. These cases demonstrate report-level inspectability and local execution, not comparative superiority, hallucination reduction, clinical validity, or operational safety. The architecture separates retrieval and control from the model interface, but behavior across different models remains to be evaluated. SRCoT is a research proof of concept requiring human review, with its principal contribution lying in an inspectable control procedure rather than a new model. [COMMENTS]28 pages, 2 figures. Extended technical account; a related brief report based on overlapping work has been submitted to Frontiers in Disaster and Emergency Medicine. The two manuscripts do not represent independent empirical studies [LINK]http://arxiv.org/abs/2610.23311v1 [DATE]2026-10-16 19:14:52+08:00 [CATEGORIES]cs.CL
Annotrace: A Workbench for Step-Level Annotation of Agent Traces [AUTHORS]Arun Josephraj Arokiaraj, Chenyi Huang, Sahan Bulathwela, Zekun Wu, Bin Wu [ABSTRACT]Every study of why LLMagents fail begins by labelling traces by hand, and every such study builds its own throwaway tool. Observability platforms record agent traces but offer no structured labelling; annotation tools label text but have no notion of a step, a delegation or a tool call, leaving it unclear what the smallest unit of execution to label should be. Annotrace reads six trace formats, including OpenTelemetry GenAI, LangSmith and Langfuse, and turns each run into an ordered, nested list of steps. The annotator writes their own labelling codebook; a frontier model proposes a starting set from a sample of their own traces, answers first, and is asked exactly the questions the human is asked, making annotation a joint human-AI process. Annotrace measures agreement between any two annotators, human or AI. On a multi-agent question-answering corpus it reproduces a previous manual effort's 235 annotation units exactly, and two models labelling the same traces agree 91% of the time at $κ= 0.82$. [COMMENTS]4 pages, 4 figures, 1 table. Under review at the AAAI 2027 Demonstration Program [LINK]http://arxiv.org/abs/2610.23284v1 [DATE]2026-10-16 18:55:26+08:00 [CATEGORIES]cs.CL
Verbalising Model's Computation for Chain-of-Thought Faithfulness Detection [AUTHORS]Yujing Wang, Yulan He, Wenya Wang, Hanqi Yan [ABSTRACT]Current Chain-of-Thought (CoT) verification methods detect reasoning faithfulness from model outputs or static activations, but are limited in explaining how unfaithfulness develops during generation. We introduce VERITAS to verbalise the model's computation: how different reasoning anchors emerge, transfer across layers, and surface in the output. VERITAS adopts the layer-wise Jacobian projection to produce a verbalised account of the model's internal reasoning evolution at each generated CoT token. By training a classifier on the extracted features along the traces, we show that these structured features provide insightful perspectives on how and where reasoning diverges toward unfaithfulness. Experiments on four datasets from FaithCoT-Bench and BonaFide show that VERITAS achieves a significant performance gain over LLM-as-Judge, with a 28.4% gain in balanced accuracy, and an average improvement of 16.0% over activation probing. Further analysis shows VERITAS can be generalised to different datasets and model architectures. Our work shows that verbalising model computation provides a richer and more interpretable lens on CoT faithfulness, moving beyond detection toward a mechanistic understanding of how reasoning fails. [LINK]http://arxiv.org/abs/2610.23275v1 [DATE]2026-10-16 18:44:23+08:00 [CATEGORIES]cs.CL
Do Speech BCIs Need Larger Models? Rethinking Neural Decoding beyond Scaling [AUTHORS]Zehui Feng, Weichuan Wang, Xiaohan Chen, Cuntai Guan, Ting Han [ABSTRACT]Recent speech brain-computer interfaces (BCIs) increasingly rely on large models and complex multi-stage pipelines for neural decoding. However, in speech BCI, where training data are scarce, noisy, and expensive to collect, the benefits of model scaling remain unclear. This work revisits a fundamental question: are large models truly necessary for accurate neural decoding? We propose BEST (Brain-Encoding-Speech-Text), a two-stage framework for efficient neural-to-text decoding. Instead of relying on large generative models or phoneme-centric pipelines, BEST introduces hierarchical intermediate representations and a compact speech-oriented decoder that directly maps neural activity to text. This design preserves essential structure while significantly reducing computational cost, enabling robust generalization in low-resource settings and practical deployment under constrained resources. On the Brain-to-Text '24 and '25 benchmarks, BEST achieves word error rates (WERs) of 6.19% and 1.62%, respectively, using only 3.85% of the parameters required by comparable methods. Beyond these results, we provide three key insights: (1) decoder scaling yields non-monotonic gains; (2) cross-modal alignment is more critical than model size; and (3) overfitting degards sequence-level performance. Representation analysis further reveals a consistent neural-audio-text alignment pathway. [COMMENTS]25 pages, Accepted by NeurIPS 2026 as a Main Track conference paper [LINK]http://arxiv.org/abs/2610.23175v1 [DATE]2026-10-16 17:12:58+08:00 [CATEGORIES]cs.CL
Cross-Centaur: Modeling Individual Decision-Making Across Psychological Tasks with Language Models [AUTHORS]Hikaru Asano, Hidezo Suganuma, Ryutaro Mori, Yukino Baba [ABSTRACT]Large language models (LLMs) are increasingly used as computational models of human behavior. Recent models such as Centaur have advanced this goal by fine-tuning LLMs on trial-by-trial decisions from psychological experiments. However, existing large-scale behavioral datasets do not link the same participants across different tasks, so these models learn population-level regularities but are not trained to model individual-level regularities across tasks. We introduce Cross-Psych-101, a dataset of 591 participants and 261\{,\}551 valid decisions across 12 Bandit, Tradeoff, Structure, and Social tasks. We train Cross-Centaur to predict a participant's target-task actions from a target-task description and the source-task description paired with the same participant's source-task behavior. Cross-Centaur outperforms off-the-shelf LLMs and Centaur on held-out participants, both within and across tasks. Probing analyses suggest that its hidden states encode participant-level behavioral tendencies that transfer across tasks. These results highlight the potential of cross-task individual-level data for predicting human decision-making with language models. Cross-Pysch-101 is available at https://huggingface.co/datasets/HikaruAsano/cross-psych-101. [COMMENTS]Accepted to EMNLP 2026 (Main Conference, Oral Presentation) [LINK]http://arxiv.org/abs/2610.23157v1 [DATE]2026-10-16 16:55:00+08:00 [CATEGORIES]cs.CL
TriRouter: Sparse-to-Dense Electrode-Space Routing for Robust EEG Motor Imagery Decoding [AUTHORS]Ying Zhou, Yaoxin, Mengting Liu [ABSTRACT]Motor imagery brain-computer interfaces require reliable decoding from multichannel EEG, but practical acquisition often yields sparse or corrupted electrode montages. We propose TriRouter, an electrode-space routing framework with discriminative-preservation, mask-robustness, and latent-compensation paths. A validation-calibrated router uses observable channel availability to allocate path contributions, preserving complete-channel discrimination while increasing representation recovery under severe missingness. In BNCI2014-001, TriRouter outperforms standard public backbones and a dynamic-spatial-filtering corrupted-EEG baseline. It improves accuracy over the strongest corrupted-EEG baseline by 3.68, 4.02, and 6.46 percentage points at nominal 25%, 50%, and 75% missing settings, and remains effective under structured regional missingness and additional low-channel dataset evaluation. [COMMENTS]Submitted to IEEE ICASSP 2027 [LINK]http://arxiv.org/abs/2610.23665v1 [DATE]2026-10-16 23:59:00+08:00 [CATEGORIES]cs.LG
Deep LLM layers expose passage-level context in EEG during silent reading [AUTHORS]Ingo Marquardt, Anthilia Alchanat, Priyanka Jain [ABSTRACT]In a previous report we showed that a decoder can recover open-vocabulary word-level information from 19-channel EEG recorded during silent reading, by aligning short EEG windows with hidden-state embeddings of the presented word taken from a large language model (LLM). That analysis contrasted two of the model's 33 layers: the context-free input embedding and one mid-depth layer. Here we ask how EEG-decodable structure is distributed over the full depth of the language model. We trained one decoder per layer of Llama-3.1-8B, with and without a causal transformer over the sequence of EEG windows. We distinguish word-level decodability from a context-tracking component. Word-level decodability peaks in early layers and declines steadily towards the last layer. In contrast, the context-tracking component of the retrieval gain rises with depth, both with and without a causal transformer. Unexpectedly, a decoder that sees a single 0.5-s window at a time tracks the text passage too. Two new controls dissect this component. A target-geometry control, a synthetic predictor made of the true target plus noise calibrated to each decoder's word-level accuracy, shows that context tracking is not a by-product of word decoding and target geometry. A control that permutes predictions across runs from either the same or from different sessions shows that most of the observed context-tracking is not caused by an EEG recording session signature. Instead, context tracking mostly follows the text. A single EEG window therefore carries passage-level information that a contextual target can absorb. The results locate word identity in the early layers of the language model, discourse state in the deeper layers, and make the choice of target layer a design dial for EEG language decoders. [COMMENTS]41 pages (including 23 pages of Supplementary Material), 12 figures (including 5 in Supplementary Material) [LINK]http://arxiv.org/abs/2610.23638v1 [DATE]2026-10-16 23:41:19+08:00 [CATEGORIES]cs.LG
Deconstructing Multi-Task Active Learning: The Paradox of Gradient Conflict and Orthogonal Decomposition [AUTHORS]Juho Jung, Seongjae Kang, Jinyoung Han [ABSTRACT]Existing multi-task active learning (MTL-AL) methods predominantly rely on linear scalarization of heterogeneous objectives such as uncertainty, diversity, and gradient conflict. We argue that this entangled formulation is fundamentally limited, as linear combination fails to capture Pareto-optimal trade-offs under conflicting objectives. We propose a shift from combination to deconstruction and introduce a principled framework that rethinks MTL-AL along three orthogonal dimensions. First, we identify the Conflict Paradox: although the optimization literature treats gradient conflict as interference to be suppressed, penalizing conflict when ranking candidates is measurably worse than not scoring it at all. We therefore decouple the two concerns: the acquisition function does not score conflict, and the resulting interference is resolved by gradient surgery during training. Second, we introduce Orthogonal Decomposition, separating acquisition into Informativeness and Exploration. This two-stage process, informativeness filtering followed by coverage maximization, prevents "diversity dilution" by excluding uninformative outliers before enforcing diversity. Third, we propose Cross-Task Mutual Information (CTMI), a covariance-based statistic of cross-task dependency computed from MC-dropout task entropies. Extensive experiments on COCO, Cityscapes, and NYUv2, including comparisons against recent MTL-AL and data-selection baselines, demonstrate consistent improvements over state-of-the-art methods, highlighting that structural rethinking of acquisition yields larger gains than incremental heuristic design. Code is available at https://github.com/Juho-Jung/multitask_active_learning_official. [COMMENTS]10 pages, 5 figures, NeurIPS 2026 [LINK]http://arxiv.org/abs/2610.23623v1 [DATE]2026-10-16 23:32:37+08:00 [CATEGORIES]cs.LG
ThetreStageAI: Multi-View Retrieval-Augmented Generation for Structured Bengali Theatre Production Planning [AUTHORS]Md. Mehedi Hasan Naeem, Mst. Kamrunnahar Ruma, Shakila Sultana, Nafiza Anjum, Nasreen Jannatun Nemu, Md. Sujan Ali [ABSTRACT]Staging a theatre performance requires coordinating dialogue, actor movement, stage directions, lighting focus, cue timing, and equipment constraints within the same production plan. This work presents ThetreStageAI, a retrieval augmented framework for structured Bengali theatre production planning. Rather than retrieving from one mixed corpus, the system maintains separate scene, actor-blocking, and lighting views. Bengali text is embedded with BGE-M3 and indexed in view-specific Qdrant collections. Retrieved references are combined with a production brief and passed to Gemini, which generates schema-constrained JSON containing characters, scenes, dialogue, stage directions, blocking, lighting, and optional sound cues. Structural and semantic checks are applied before acceptance, with at most one constrained repair attempt. We also construct a human-guided, LLM-assisted synthetic dataset of 500 short Bengali theatre records that expands to 1,500 retrieval documents. Dataset diagnostics confirm consistent identifiers and valid lighting fields while also revealing substantial dialogue repetition. A paired camera-ready evaluation compares single-combined and full multi-view retrieval under matched settings. The source code, dataset, and evaluation materials are publicly available. [COMMENTS]6 pages, 6 figures. Accepted at the 12th IEEE International Conference on Sustainable Technologies and Engineering (IEEE iCOSTE 2026), Cox's Bazar, Bangladesh, November 1-3, 2026 [LINK]http://arxiv.org/abs/2610.23618v1 [DATE]2026-10-16 23:29:20+08:00 [CATEGORIES]cs.LG
OrbiSim: World Models as Differentiable Physics Engines for Embodied Intelligence [AUTHORS]Jiajian Li, Jingyuan Huang, Junru Gong, Qi Wang, Xiaokang Yang, Yunbo Wang [ABSTRACT]We present OrbiSim, a novel robotic simulation paradigm that redefines world models as a fully differentiable physics engine for embodied intelligence. Unlike prior world models that focus on unconstrained imagination in latent or visual domains, OrbiSim establishes a unified, physically-grounded pathway that bridges structured scene assets, neural dynamics, and downstream reinforcement learning. By enabling end-to-end differentiability throughout the entire simulation loop -- spanning from explicit state transitions to visual observation generation -- OrbiSim supports tasks traditionally intractable for classical simulators, such as differentiable contact modeling, gradient-based policy optimization under sparse rewards, and intuitive physical inference. Empirical results demonstrate that OrbiSim significantly outperforms state-of-the-art world models in both predictive fidelity and control performance. Furthermore, its consistent responsiveness to asset configurations and physical parameters suggests its potential as a differentiable tool for enhancing robot simulation and policy training. [COMMENTS]Project page: https://jjleejj85.github.io/projects/orbisim [LINK]http://arxiv.org/abs/2605.16395v2 [DATE]2026-10-16 23:22:03+08:00 [CATEGORIES]cs.LG
Empirical Bayesian Multi-Bandit Learning [AUTHORS]Xia Jiang, Rong J. B. Zhu [ABSTRACT]Multi-task learning in contextual bandits has attracted significant research interest due to its potential to enhance decision-making across multiple related tasks by leveraging shared structures and task-specific heterogeneity. In this article, we propose a novel hierarchical Bayesian framework for learning in various bandit instances. This framework captures both the heterogeneity and the correlations among different bandit instances through a hierarchical Bayesian model, enabling effective information sharing while accommodating instance-specific variations. Unlike previous methods that overlook the learning of the covariance structure across bandits, we introduce an empirical Bayesian approach to estimate the covariance matrix of the prior distribution. This enhances both the practicality and flexibility of learning across multi-bandits. Building on this approach, we develop two efficient algorithms: ebmTS (Empirical Bayesian Multi-Bandit Thompson Sampling) and ebmUCB (Empirical Bayesian Multi-Bandit Upper Confidence Bound), both of which incorporate the estimated prior into the decision-making process. We provide the frequentist regret upper bounds for the proposed algorithms, thereby filling a research gap in the field of multi-bandit problems. Extensive experiments on both synthetic and real-world datasets demonstrate the superior performance of our algorithms, particularly in complex environments. Our methods achieve lower cumulative regret compared to existing techniques, highlighting their effectiveness in balancing exploration and exploitation across multi-bandits. [COMMENTS]33 pages, 13 figures [LINK]http://arxiv.org/abs/2510.26284v3 [DATE]2026-10-16 23:06:14+08:00 [CATEGORIES]cs.LG
BMuon: Budgeting Muon's Spectral Geometry for Structured Compressibility [AUTHORS]Rida Assalouh, Tom Jacobs, Rebekka Burkholz, Yani Ioannou, Umut Şimşekli [ABSTRACT]Muon has emerged as a strong optimizer for neural network training and admits an interpretation as steepest descent under a spectral-norm geometry. Yet we find that Muon-trained Transformers can degrade sharply when their MLP hidden channels are pruned after training. To reconcile Muon's optimization performance with structured compressibility, we introduce Budget Muon (BMuon), which modifies Muon's spectral geometry so that hidden channels across blocks compete for a shared allocation budget. The resulting oracle is exactly normalized steepest descent under a family of norms that interpolate between Muon's geometry and structured sparse geometries. Our principal sparsity control is an interpretable global channel-allocation budget, rather than a setting-dependent sparsity-penalty coefficient. For linear factorization models, we show that Muon admits minimum-cost exact representations with redundant hidden width, whereas under the BMuon oracle geometry and a sufficiently scarce budget, every minimum-cost representation is width-minimal. For practical training, we derive a minorization-maximization surrogate for the exact allocation problem. On ImageNet-1K with ViT-S/16, BMuon improves over Muon under gradual structured MLP pruning, and its learned allocations provide a more effective pruning criterion than weight magnitude. On a 1B-parameter LLaMA-style model trained on FineWeb, BMuon improves over Muon under post-training MLP channel pruning. [COMMENTS]54 pages, 6 figures [LINK]http://arxiv.org/abs/2610.23571v1 [DATE]2026-10-16 22:52:56+08:00 [CATEGORIES]cs.LG
Causal Discovery beyond Conditional Independence: Identifiability, Algorithms, and Statistical Inference for Directed Acyclic and Cyclic Graphs [AUTHORS]Yang Ni [ABSTRACT]Learning causal structure from data requires assumptions about mechanisms, noises, observation schemes, and interventions. Historically, causal discovery has relied primarily on conditional independence. This review focuses on modern causal discovery methods that exploit properties of distributions beyond conditional independence. Our organizing principle separates the source of identification from the statistical estimator and its computational implementation. This perspective connects conditional-independence methods with distributional asymmetries, conditional-moment restrictions, and variation across environments or individuals. We examine discovery under latent confounding and feedback, distinguishing causal information about observed variables from reconstruction of latent structure and emphasizing the additional semantics required by cyclic models. We also distinguish population identification, frequentist and Bayesian consistency, uncertainty quantification, and algorithmic guarantees. Throughout, examples show why the same observational distribution can support different conclusions under different model classes. We conclude with guidance for comparing methods at their identifiable targets and a research agenda for uncertainty, heterogeneous populations, non-scalar variables, and the intersection of discrete data, feedback, and hidden confounding. [LINK]http://arxiv.org/abs/2610.23543v1 [DATE]2026-10-16 22:37:57+08:00 [CATEGORIES]cs.LG
Time-partitioning Gaussian Process Mixture Model [AUTHORS]Guilherme Piantino, Hedibert F. Lopes [ABSTRACT]Nonstationarity and heterogeneous dynamics are common in financial and macroeconomic time series. Volatility alternates between calm and turbulent regimes, interest rates follow monetary-policy cycles, and major events trigger persistent structural shifts. We propose the Time-partitioning Gaussian Process Mixture Model (TPGPMM), a Bayesian nonparametric framework for regime identification and nonstationary prediction that infers such regimes and their within-regime dependence from data, without Markovian assumptions. It combines (i) an adaptive Pitman-Yor process prior on ordered compositions, mapped to block lengths with contiguity enforced by construction, and (ii) regime-specific Gaussian processes for temporal dependence, yielding a nonstationary predictive distribution. By inferring its discount parameter, the prior adapts to homogeneous and highly heterogeneous regime lengths. Unlike sequential or Markov-switching methods, it flags persistent structural shifts as change points rather than reacting to transient shocks. A split-merge Metropolis-within-Gibbs sampler with exact Gibbs updates of regime boundaries jointly learns change points and hyperparameters in batch, quantifying uncertainty over the partition and within-regime dynamics. With an RBF kernel and a fixed, data-informed noise level, TPGPMM produces a sparse, confident segmentation. On simulated data it recovers the true regimes, and on the CBOE Volatility Index and the 10-year US Treasury yield it identifies regimes aligned with documented market and monetary-policy events. For prediction, variants estimating the noise level fit better than change-point baselines and fit and forecast better than the main specification, at the cost of a less sparse or less event-aligned segmentation; with a regime mean, they fit better than a fitted stationary Gaussian process and have lower forecast RMSE except at the shortest horizons. [LINK]http://arxiv.org/abs/2610.20860v2 [DATE]2026-10-16 22:12:42+08:00 [CATEGORIES]cs.LG
A Topological Signature of Learning under Noisy Labels [AUTHORS]Zhongtian Sun, Fan Mo, Prayag Tiwari, Kelin Xia [ABSTRACT]How does noisy supervision change the representations learned by a neural network? Training accuracy measures agreement with training labels but provides little information about representation geometry. We use persistent homology to track loop structure across distance thresholds in neural activations. Experiments span image classification, text classification, and modular arithmetic. On a fixed CIFAR-10/ResNet-18 training probe, equally sized clean-label and corrupted-label subsets show no detectable persistence difference at epoch 10, whereas corrupted-label subsets have higher persistence at epoch 50. The contrast survives uniform-scale controls, even though predictions match the assigned noisy labels for only 7-11% of corrupted probe examples at that checkpoint. Normalisation preserves the crossover between clean and noisy training trajectories on CIFAR-10/ResNet-18, whereas CIFAR-100/ResNet-18 shows no crossover under any of the five evaluated constructions. In a separate noise-rate calibration comparison across five pipelines, mean activation norm achieves a higher mean in-sample R2 than raw persistence (0.936 versus 0.908). These findings show that noisy supervision leaves a measurable geometric signature even when agreement with corrupted labels is low. Comparing persistence under scale controls with label-fitting measurements offers a way to study how representations respond to label corruption beyond what training accuracy reveals. [LINK]http://arxiv.org/abs/2610.23491v1 [DATE]2026-10-16 22:04:35+08:00 [CATEGORIES]cs.LG
Graphlet-based Decentralized Finance Transaction Method Classification [AUTHORS]Célestin Coquidé, Alessio Brini, Natkamon Tovanich [ABSTRACT]Decentralized Finance (DeFi) transactions involve complex interactions among accounts, smart contracts, and tokens. We investigate whether network topology can identify transaction methods that describe DeFi operations. We extract directed graphlet features to classify the 30 most frequent methods. Our results show that graphlet features complement event log counts in improving classification performance. Using SHapley Additive exPlanations (SHAP), we further derive model explanations that reveal method-specific graphlet structural patterns. [COMMENTS]24 pages, 5 figures, 5 tables, 25 references 15th International Conference on Complex Networks and Their Applications [LINK]http://arxiv.org/abs/2610.23466v1 [DATE]2026-10-16 21:46:37+08:00 [CATEGORIES]cs.LG
RamanBench: A Large-Scale Benchmark for Machine Learning on Raman Spectroscopy [AUTHORS]Mario Koddenbrock, Christoph Lange, Robin Legner, Martin Jäger, Martin Kögler, Mariano N. Cruz Bournazou, Peter Neubauer, Felix Biessmann, Erik Rodner [ABSTRACT]Machine learning (ML) has transformed many scientific fields, yet key applications still lack standardized benchmarks. Raman spectroscopy, a widely used technique for non-invasive molecular analysis, is one such field where progress is limited by fragmented datasets, inconsistent evaluation, and models that fail to capture the structure of spectral data. We introduce RamanBench, the first large-scale, fully reproducible benchmark for ML on Raman spectroscopy, consisting of streamlined data access (https://pypi.org/project/raman-data/), evaluation protocols and code (https://pypi.org/project/raman-bench/), as well as a live leaderboard (https://huggingface.co/spaces/HTW-KI-Werkstatt/RamanBench). It unifies 77 datasets (including 17 first released with this benchmark) across four domains, comprising 325,862 spectra and spanning classification and regression tasks under diverse experimental conditions. We benchmark 28 models under a standardized protocol, including classical methods (e.g., PLS), Raman-specific (e.g., RamanNet), tabular foundation models (TFMs; e.g., TabPFN), and time-series approaches (e.g., ROCKET). TFMs consistently outperform domain-specific and gradient boosting baselines, while time-series models remain competitive. However, no single method dominates across these heterogeneous datasets, revealing that Raman spectroscopy remains an open challenge for machine learning. Therefore, we invite the community to contribute new approaches to our living benchmark, with the potential to accelerate advances in critical applications such as medical diagnostics, biological research, and materials science. [COMMENTS]https://huggingface.co/spaces/HTW-KI-Werkstatt/RamanBench [LINK]http://arxiv.org/abs/2605.02003v4 [DATE]2026-10-16 21:36:03+08:00 [CATEGORIES]cs.LG
Dynamics of Stochastic Momentum with Sparse Updates in High Dimensions [AUTHORS]Katie Everett, Elliot Paquette [ABSTRACT]Existing theory of momentum assumes that gradients arrive at every parameter at a roughly constant rate, an assumption violated in practice by heavy-tailed data distributions and modern architectures. We theoretically analyze the dynamics of two tractable models of momentum under sparse updates: a least squares model with sparse inputs and a logistic regression model with a rare class. Both admit exact closed-form second-moment dynamics whose high-dimensional limits we characterize across three scaling exponents for sparsity, batch size, and momentum decay. The phase structure on both problems is governed by the ratio of two intrinsic timescales: a momentum retention timescale (how many active updates the buffer survives) and a learning timescale (how many active updates it takes to reduce the squared error). When learning is much slower than retention, the limit matches SGD; when learning is faster, the system is unstable; where the timescales coincide, we recover classical heavy-ball dynamics. The oscillatory dynamics occur at different momentum values for different token sparsity, creating a spectral conflict for global momentum across token frequencies. [LINK]http://arxiv.org/abs/2605.28961v2 [DATE]2026-10-16 21:22:14+08:00 [CATEGORIES]cs.LG
Generative Policy Improvement from Observational Data [AUTHORS]Jonas Schweisthal, Dennis Frauen, Stefan Feuerriegel, Emma Brunskill [ABSTRACT]We study how to improve decision policies from observational data to generate better actions when actions are complex, structured objects, such as text messages or images. This task is subject to two major challenges: (i) action-level overlap is often too weak for reliable off-policy evaluation and improvement, and (ii) generating new structured actions often requires expensive training or fine-tuning (e.g., of an LLM) and can be ill-defined due to information loss in action representations. To address both, we propose PolicyLift, a three-stage framework for generative policy improvement with structured actions from observational data. In Stage 1, PolicyLift learns a small set of discrete action codes that capture high-level strategies in the observed action space and are tailored to support policy improvement under limited overlap. In Stage 2, it learns a context-dependent policy over these action codes. At deployment time, Stage 3 "lifts" the learned code-level policy back to the original action space by generating actions from the historical behavior system in a code-conditioned way. We show theoretically that improving the policy over action codes guarantees improvement of the induced policy in the original action space. We demonstrate PolicyLift across experiments in synthetic and real-world text generation tasks with consistent policy improvements over the behavior policy. [COMMENTS]Preprint [LINK]http://arxiv.org/abs/2610.23436v1 [DATE]2026-10-16 21:20:43+08:00 [CATEGORIES]cs.LG
Residual Stochastic Concept Bottleneck Models [AUTHORS]Stephen Marcou, Mateo Espinosa Zarlenga, Sonia Laguna, Moritz Vandenhirtz, Julia E. Vogt, Mateja Jamnik [ABSTRACT]Concept-based models (CMs) learn human-interpretable intermediate representations that enable users to understand the model's reasoning and steer its predictions through test-time concept interventions. However, in realistic settings, their performance can be limited by concept incompleteness, where the training concepts fail to capture all task-relevant information. Adding an unconstrained residual pathway to the model's predictive process can overcome the consequences of incompleteness. However, this introduces an inscrutable, leak-prone prediction pathway that may weaken concept interventions, as these often leave residual variables unchanged. To address this limitation, we introduce Residual Stochastic Concept Bottleneck Models (Res-SCBMs), a family of concept-based models that jointly model concept and residual variables through a multivariate Gaussian distribution. Res-SCBMs learn a covariance structure that explicitly captures dependencies between the known concepts and learnt residual variables, allowing interventions to propagate to statistically dependent residual variables through Gaussian conditioning. This extends the intervention capabilities of previous CMs without requiring concept completeness during training. Our results show that, across multiple concept-incomplete benchmarks, Res-SCBMs improve predictive performance over competing baselines while retaining effective test-time concept interventions. [LINK]http://arxiv.org/abs/2610.23421v1 [DATE]2026-10-16 21:09:16+08:00 [CATEGORIES]cs.LG
2026 Oct 17, Sat
LeanPRM: Lean-Grounded Process Supervision for Natural Language Proof Verification [AUTHORS]Rodney Lafuente-Mercado, Salman Rahman, Jingyan Shen, James Shiffer, Saadia Gabriel, Pavel Izmailov [ABSTRACT]AI systems now write long and difficult mathematical proofs, and checking their step-by-step validity is becoming a bottleneck. A learned verifier could check proofs cheaply, but training one requires a correctness label on every step, and such labels are hard to obtain for open-ended proofs. Existing methods take them from human annotators, which is costly, or derive them from final answers, which proofs do not have, and formalizing each proof in Lean requires expert effort. We introduce LeanPRM, a framework that uses the Lean proof assistant to create step-level labels for natural-language proofs. A model writes a natural-language proof together with a paired formal proof, without feedback from Lean. Lean then labels every step, and a second model discards failures caused only by the formal syntax, so that the negative labels mark errors in the mathematics. LeanPRM contains 10,634 proofs with 152K labeled steps and no human annotation. A Qwen3-4B verifier trained on LeanPRM raises exact step-label accuracy on a held-out set of 200 proofs (LeanPRM-Bench) from 48.2 to 74.7, as opposed to 60.3 when the same base model is trained on PRM800K. It also selects better proofs in best-of-N selection on IMO-level problems, and outperforms the PRM800K-trained verifier on three of four external verification benchmarks. Training on LeanPRM improves verifiers at every size from 0.6B to 8B parameters on LeanPRM-Bench, Hard2Verify and IMO-GradingBench. Given the labels need no human annotation, LeanPRM can grow with the supply of formal statements, offering a path to step-level feedback on AI-generated proofs that are hard for humans to check. [COMMENTS]29 pages, 9 figures. Project page: https://rlafuente.com/lean-prm [LINK]http://arxiv.org/abs/2610.17697v2 [DATE]2026-10-17 01:31:06+08:00 [CATEGORIES]cs.LG
2026 Oct 16, Fri
Dependency Reasoning for Aspect Sentiment Quadruple Prediction [AUTHORS]Shima Esfandyari, Seywan Soleymanbaigi, Fardin Akhlaghian Tab, Fatemeh Daneshfar, Mourad Oussalah [ABSTRACT]Aspect Sentiment Quadruple Prediction (ASQP) is fundamentally a relational reasoning problem: identifying aspect-opinion-category-sentiment quadruples requires modeling structured dependencies, not merely sequential co-occurrence. Existing LLM-based approaches incorporate syntactic information indirectly, via retrieval-based demonstration selection or multi-stage training, rather than as unified, single-pass generation-time context. We present DRIFT, which linearizes dependency parse trees and injects them directly into LLM prompts as explicit relational context, enabling the model to reason over aspect-opinion dependencies rather than infer them from token proximity alone. DRIFT further integrates retrieval-augmented in-context learning and a syntax-aware LLM-based adjudication mechanism that complements exact-match evaluation under span-boundary variation. On SemEval Rest15, Rest16, and ACOS, DRIFT achieves the strongest performance among compared generative ASQP methods, reaching 54.47% and 64.10% F1 on Rest15 and Rest16, with consistent gains across short- and medium-distance aspect-opinion configurations; adjudication further shows exact-match metrics systematically underestimate semantically correct predictions. [LINK]http://arxiv.org/abs/2610.14873v2 [DATE]2026-10-16 20:03:17+08:00 [CATEGORIES]cs.CLcs.LG
MAS-Bai: Linguistically Informed Adaptive Multi-Agent Translation for Bowen-Script Bai [AUTHORS]Jingting Zheng, Hong Ren, Yuehan Cui, Shaolin Zhu, Deyi Xiong [ABSTRACT]Large Language Models (LLMs) remain unreliable for extremely low-resource languages whose writing systems conflict with high-resource pretraining priors. We study Bai-Chinese translation in the Bowen script, where Chinese characters are frequently borrowed phonetically rather than semantically, causing severe literal-translation hallucinations. We propose MAS-Bai, a linguistically informed adaptive multi-agent framework for Bowen-script Bai-Chinese translation. MAS-Bai routes each input through either a lightweight retrieval-translation path or a deeper philological path with etymological and syntacticagents, and uses cycle-consistency with a lightweight reverse translator as an inference-time reliability signal. We construct a 6,536-sentence Bai-Chinese corpus and evaluate on a 1,536-sentence test set. MAS-Bai achieves 16.23 BLEU and 62.15 COMET, improving over Standard RAG by 8.11 BLEU and 6.85 COMET. Forward LoRA, expert ablations and controlled routing experiments distinguish the roles of supervised adaptation, Bai-specific expertise and conditional routing. The largest gains occur on Bowen phonetic loans and divergent Bai word order, while human evaluation and reverse-verifier analyses support the automatic results. [COMMENTS]14 pages, 3 figures, Findings of EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23205v1 [DATE]2026-10-16 17:43:32+08:00 [CATEGORIES]cs.CL
Behavioral Couplings: From Formation to Behavioral Emergence in Language Models [AUTHORS]Dongyoon Hahm, Woogyeol Jin, Dennis Wei, Yi Zhou, Nathalie Baracaldo, Kimin Lee [ABSTRACT]Fine-tuning large language models (LLMs) can induce behaviors beyond those explicitly targeted. For example, training a model to respond more warmly can make it more sycophantic, and training it to write insecure code can cause broad misalignment on prompts unrelated to coding. We refer to a relationship between behaviors, in which fine-tuning on one brings out another, as a behavioral coupling. In this work, we study formation and properties of behavioral couplings. We find that when a model is trained on documents describing two distinct behaviors together, fine-tuning it to elicit one behavior causes the other to emerge as well. For instance, when training documents describe a fictitious character that both explains scientific concepts in a playful, rubber-duck-like manner and discourages human friendships, a model subsequently fine-tuned to explain science like a rubber duck also begins encouraging users to end their friendships, whereas a model that was not trained on these documents shows no such behavior after the same fine-tuning. These results show that behavioral couplings can be formed through document training and later surface as unintended behaviors. More concerningly, we find that couplings can compose. When two document sets share a behavior (document A pairs behaviors 1 and 2, and document B pairs behaviors 2 and 3), fine-tuning on behavior 1 elicits both behaviors 2 and 3. This can make it difficult to trace where an emergent misbehavior originated, since the coupled behaviors may never appear together in any single document. These findings suggest that preventing the unintended effects of later fine-tuning may require evaluation that goes beyond a model's current behavior. [LINK]http://arxiv.org/abs/2610.23847v1 [DATE]2026-10-17 01:58:08+08:00 [CATEGORIES]cs.CLcs.LG
AI-Driven Reproducibility and Analytical Robustness in Scientific Discovery [AUTHORS]Shicheng Liu, Krista Opsahl-Ong, Andrew Drozdov, Julian Killingback, Arnav Singhvi, Ronak Pradeep, Michael Bendersky, James C. Dickerson, Monica S. Lam [ABSTRACT]Reproducible and analytically robust research findings are essential to scientific progress, but evaluating these properties typically requires costly manual effort. We investigate whether LLMagents can help automate reproduction and robustness analysis, using a medical case study grounded in the Surveillance, Epidemiology, and End Results (SEER) database, a widely used U.S. cancer registry underlying tens of thousands of published studies. First, we introduce SEER-Repro, a benchmark of 20 published SEER-based papers together with the data and resources required to reproduce their findings. Second, we create an AI agent that successfully reproduces 77.4% of 2,190 atomic claims, while providing executable baselines for further analysis. Third, we develop an agentic analysis pipeline to test the robustness of the medical findings by identifying analytical choices underlying each finding, proposing defensible alternatives, and rerunning the analysis to determine whether those choices affect the conclusion. Across published papers in SEER-Repro, our pipeline identifies potentially consequential analytical choices in 9.5% of fully reproduced findings and flags them for expert review. These results show how LLMagents can support scientific review by reproducing findings, systematically testing their robustness, and directing expert attention toward potentially consequential analytical choices. [LINK]http://arxiv.org/abs/2610.23845v1 [DATE]2026-10-17 01:57:55+08:00 [CATEGORIES]cs.CL
Optimizing Singular Spectrum for Large Language Model Compression [AUTHORS]Dengjie Li, Tiancheng Shen, Yao Zhou, Bernard Ghanem, Yibo Yang, Ming-Hsuan Yang [ABSTRACT]Large language models (LLMs) have demonstrated remarkable capabilities, yet prohibitive parameter complexity often hinders their deployment. Existing singular value decomposition (SVD) based compression methods simply deem singular values as importance scores of decomposed components. However, this importance ordered by singular values does not necessarily correlate with the performance of a downstream task. In this work, we introduce SoCo (Singular spectrum optimization for large language model Compression), a novel compression framework that learns to rescale the decomposed components of SVD in a data-driven manner. Concretely, we employ a learnable diagonal matrix to assign importance scores for singular spectrum and develop a three-stage training process that progressively refines these scores from initial coarse compression to fine-grained sparsification-thereby striking an effective balance between aggressive model compression and performance preservation. Thanks to the learnable singular spectrum, SoCo adaptively prunes components according to the sparsified importance scores, rather than relying on the fixed order of singular values. More importantly, the remaining components with amplified importance scores can compensate for the loss of the pruned ones. Experimental evaluations across multiple LLMs and benchmarks demonstrate that SoCo surpasses the state-of-the-art methods in model compression. [LINK]http://arxiv.org/abs/2502.15092v2 [DATE]2026-10-17 01:56:38+08:00 [CATEGORIES]cs.CL
DiscoveryRSI: Recursively Learning to Discover Valuable Hypotheses [AUTHORS]Jianyuan Zhong, Chenwei Wu, Zhaochen Yu, Xixian Jiang, Xuewei Li, Hao Xiao, Xilin Dai, Junyang Wang, Jiaru Zou, Zhiqing Cui, Boyuan Zheng, Jing Xiong, Zhen Zhang, Qiang Gao, Michael Morley, Luis Nakayama, Qiang Xu, Liyue Shen, Tao Gui, Leo Anthony Celi, James Zou, Zhenfei Yin, Yingcheng Wu, Ling Yang [ABSTRACT]Scientific discovery requires developing worthwhile hypotheses into practical investigations, yet learning from this evolving research process remains challenging. We introduce an operational definition of valuable and testable research hypotheses: evidence-grounded claims about consequential uncertainties, coupled with feasible investigations, explicit decision criteria, and alternative explanations. Our DiscoveryRSI framework develops hypothesis trees through multi-agent evolutionary search, compiles selected investigations into executable Harbor tasks, and recursively learns through three mechanisms: hypothesis credit optimization attributes scientific assessments to documented research contributions; harness replay identifies productive strategies and recurring failures to refine research workflows; and rediscovery applies accepted improvements to new investigations, generating further experience for subsequent learning. We also introduce DFM-Bench, which provides comprehensive assessment of hypothesis quality and cumulative research progress at the levels of hypotheses, revisions, development paths and research runs. Across four life-science domains, automated evaluation shows that hypothesis quality rises from 4.86 for the base model to 5.39 after two rounds of self-improvement, with the large gains in falsifiability and data resolvability. We further validate hypothesis quality through expert review, comparing our DiscoveryRSI-27B with other research agents. Preliminary human assessments show that DiscoveryRSI-27B scores higher than GPT-6 Luna on hypothesis importance and whether claims are plausible and evidence-bounded. DiscoveryRSI-27B achieved a 6.3% rate of hypothesis correspondence with research published after the model cutoff, largely outperform GPT-6 with 2.5% and Claude Opus 5 with 2.3%, while delivering greater human-rated improvement along research paths. [COMMENTS]Code: https://github.com/Gen-Verse/DiscoveryRSI [LINK]http://arxiv.org/abs/2610.23839v1 [DATE]2026-10-17 01:56:35+08:00 [CATEGORIES]cs.CL
Beyond the Answer: Student-Conditioned Trace Construction for Reasoning Distillation [AUTHORS]Shaoang Li, Yanhang Shi, Yufei Li, Mingfu Liang, Frank Shyu, Sandeep Pandey, Luke Simon, Xi Liu, Jian Li [ABSTRACT]A reasoning trace can accompany a correct answer without providing effective supervision for a particular student. We study how to construct traces that improve learning beyond answer-only supervision by considering both their intermediate content and their accessibility to the target learner. We introduce ReDiST, a student-conditioned trace-construction and distillation framework for evidence-rich reasoning. Starting from the student's own attempts, ReDiST revises answer-correct drafts against the supplied evidence and repairs unsuccessful drafts with additional guidance from the verified answer. A frozen student-matched reader screens and ranks original and edited candidates using answer recovery and verified-answer probability, given the question, evidence, and candidate trace. We distill the selected targets while balancing trace and final-answer supervision with an auxiliary answer loss. Construction and selection occur offline, and inference uses only the distilled student. Experiments on question-answering benchmarks show that ReDiST can improve both source-task and cross-benchmark performance over answer-only, original-trace, and the evaluated trace-selection and rewriting baselines. Across supervision recipes, performance rankings can differ between source and transfer benchmarks, highlighting the importance of evaluating trace teaching utility across distributions. [LINK]http://arxiv.org/abs/2610.23831v1 [DATE]2026-10-17 01:52:58+08:00 [CATEGORIES]cs.CLcs.LG
Automatic Tensor-level Execution of Abstract Interpreters for Neural Networks [AUTHORS]Avaljot Singh, Yasmin Chandini Sarita, Aditya Mishra, Gagandeep Singh, Charith Mendis [ABSTRACT]Developing new abstract interpretation-based DNN certifiers is difficult because their mathematical design is expressed at the neuron level, while implementations must operate at the tensor level. Existing libraries such as auto_LiRPA, IVAN, and ERAN support fixed sets of certifiers; extending them requires expertise in formal methods and tensor programming. Domain-specific languages such as ConstraintFlow let researchers specify certifiers at the neuron level, but cannot run them on real DNNs.
We present Neuton, a framework that automatically translates neuron-level certifier specifications into tensor-level executables, enabling new DNN certifiers without manual tensor programming. Neuton makes three contributions. First, a stack-based intermediate representation (IR) and shape analysis infer implicit tensor operations, including broadcasts, dimension expansions, and reductions, required for correct execution. Second, domain-specific rewrite rules optimize common tensor operation sequences into more efficient equivalents. Third, gBCSR is a sparse tensor format using variable-sized, layer-aligned blocks over a non-zero default value, tailored to DNN certifier sparsity.
We prove that translation preserves neuron-level semantics. We evaluate Neuton on 8 certifiers and 32 DNN benchmarks. Neuton compiles four existing certifiers and produces the first executable implementations of four others: two previously specified but never run, and two new designs. These expose precision-runtime trade-offs unavailable in existing libraries. For example, SkipPoly matches DeepPoly's precision on all 32 DNNs while running as fast or faster on half. Neuton's runtime on existing certifiers is competitive with hand-optimized libraries. Code: https://github.com/uiuc-focal-lab/constraintflow.git [LINK]http://arxiv.org/abs/2507.20055v2 [DATE]2026-10-17 01:43:47+08:00 [CATEGORIES]cs.CL
Acceptance Is Not Speed: Length Inflation in Training-Free Lossy Speculative Decoding [AUTHORS]Chiatzen Wang, Bill Xu, Haochen Gu, Hongyang Zhang [ABSTRACT]Speculative decoding typically characterizes speedup by draft tokens accepted per verification round. Lossless verification preserves the target model's output distribution and thus expected completion length, making this metric a meaningful efficiency proxy. Relaxed (lossy) verification accepts tokens that lossless verification would reject, potentially altering that distribution. We identify a length inflation phenomenon: relaxed verification lengthens completions, primarily by extending reasoning traces, largely offsetting apparent gains in accepted tokens per round. We evaluate existing training-free relaxed verification rules on twelve target-drafter pairs against lossless verification on identical prompts, measuring completion length, verifier rounds, wall-clock time, and task accuracy. At each rule's most relaxed evaluated setting, relaxed verification on average accepts $1.34\times$ as many tokens per round but produces $1.54\times$ longer completions. For rules with higher acceptance lengths, tokens accepted only under relaxed verification tend to have low target-model probability, contributing to length inflation. For rules that reduce verifier rounds, accuracy typically drops substantially. Further analysis traces some failures to continued reasoning that exhausts the budget needed to deliver a final answer. Overall, accepted tokens per round alone is an unreliable measure of decoding speed under relaxed verification. Completion length and verifier rounds must also be reported to assess actual efficiency. [COMMENTS]Submitted for NAACL 2027, under review [LINK]http://arxiv.org/abs/2610.23818v1 [DATE]2026-10-17 01:41:19+08:00 [CATEGORIES]cs.CL
Steering beyond Language: Improving Multilingual Training-Free Text Embeddings via Paired-Residual Subspace Removal [AUTHORS]Ailiang Lin, Zhuoyun Li, Yusong Wang, Keyu Mao, Kotaro Funakoshi, Manabu Okumura [ABSTRACT]While deriving text embeddings directly from frozen LLMs has emerged as a promising direction, existing methods focus almost exclusively on English, leaving multilingual settings largely unexplored. In this work, we present the first systematic study of multilingual training-free text embeddings in LLMs. Our analysis reveals a language-conditioned embedding geometry in which semantically equivalent sentences remain misaligned across languages. We hypothesize that language-specific variation can be disentangled from semantic content to improve embedding quality. To this end, we propose Paired-Residual Subspace Removal (PRSR), a training-free inference-time intervention that steers LLM representations toward a more language-invariant semantic space. PRSR first constructs paired residuals from multilingual parallel sentences to capture language-specific deviations. It then estimates layer-wise low-rank subspaces from these residuals and suppresses the corresponding components in the last-token hidden states. We conduct more than 200 monolingual evaluations across 18 languages using MTEB and MMTEB, along with cross-lingual bitext mining between English and over 200 other languages. Empirical results demonstrate that PRSR consistently improves the performance of existing training-free baselines in both monolingual and cross-lingual settings across different LLMs. [LINK]http://arxiv.org/abs/2610.23807v1 [DATE]2026-10-17 01:30:43+08:00 [CATEGORIES]cs.CL
When Reasoning Traces Conflict with Evidence: How Answer Configuration Shapes Evidence Following [AUTHORS]Shaoang Li, Yanhang Shi, Jian Li [ABSTRACT]A reasoning trace retained at answer time plays two roles: it can raise accuracy, and it can compete with the current evidence when the two disagree. We disentangle these roles by crossing verified evidence versions with the traces generated from each and answering from the same traces and evidence under two configurations: continuation of the model's own thinking and a separate answer call. Across three model families (gpt-oss, Gemma4, Qwen3) on HotpotQA and MuSiQue, the separate call raises evidence-following over continuation by 29.9 to 60.6 points. On the same pairs, removing the trace lowers accuracy by similar amounts under both configurations. Yet continuation repeats the stale conclusion in most conflicts, and the separate call follows the current evidence far more often. Benefiting from a trace and deferring to it are therefore distinct properties. Further analyses show that the final line of a trace carries most of its value, gold evidence helps most before the trace is written, distractors weaken correction under a separate call, and renewed reasoning, unlike a verification instruction, follows the current evidence in most conflicts. Because trace deletion alone cannot reveal which source an answer follows, evaluations should report both properties together with the answer configuration. [COMMENTS]25 pages, 4 figures [LINK]http://arxiv.org/abs/2610.20805v2 [DATE]2026-10-17 01:30:03+08:00 [CATEGORIES]cs.CL
DLite: History-free Speculative Decoding for Agentic AI [AUTHORS]Jiaming Xu, Hanzhen Wang, Tianlang Zhao, Jiewen Xiao, Chengze Yuan, Jiayi Pan, Yongkang Zhou, Jiancai Ye, Yu Wang, Guohao Dai [ABSTRACT]Speculative decoding has become a mainstream technique for accelerating large language model (LLM) inference by using a lightweight draft model to generate candidate tokens that are subsequently verified by the target model. However, agentic workloads introduce long and continuously evolving contexts and frequent multi-turn interactions. Under such workloads, we observe that existing draft models suffer from up to an 80% degradation in acceptance length, while maintaining their historical KV cache introduces substantial system overhead. In this paper, we identify the growing historical context as a common source of both problems and introduce DLite, a history-free speculative decoding framework for agentic AI. DLite directly reuses contextual hidden states from the target model and performs local speculative prediction without explicitly maintaining the historical context, transforming an increasingly complex modeling problem into local simple prediction problem. A lightweight sequential head further restores intra-block causal dependencies with low overhead. To effectively train this heterogeneous history-free architecture, we introduce a tailored two-stage training strategy. Experiments on general-purpose, long-context and agentic workloads show that DLite achieves 2-6 acceptance length and outperforms the state-of-the-art speculative decoding method, DSpark, up to 70%. [COMMENTS]14 pages, 5 figures [LINK]http://arxiv.org/abs/2610.23787v1 [DATE]2026-10-17 01:18:54+08:00 [CATEGORIES]cs.CL
Language steering in latent space to mitigate unintended code-switching [AUTHORS]Andrey Goncharov, Nikolai Kondusov, Alexey Zaytsev [ABSTRACT]Multilingual Large Language Models (LLMs) often exhibit hallucinations such as unintended code-switching, reducing reliability in downstream tasks. We propose latent-space language steering, a lightweight inference-time method that identifies language directions via Principal Component Analysis (PCA) on parallel translations and steers token embeddings along these axes to control language identity. Our approach mitigates code-switching (reduced next-token distributional divergence by up to 55\%, reduced Code-Switching Index by 63--99\%) while preserving broad semantic direction with negligible computational overhead and requires only minimal parallel data for calibration. Additionally, we evaluate language classification using a single principal component (95-99\% accuracy) and further analyze the layer-wise evolution of language representations. [COMMENTS]Published at AINL 2026 [LINK]http://arxiv.org/abs/2510.13849v4 [DATE]2026-10-17 01:02:53+08:00 [CATEGORIES]cs.CLcs.LG
VocalGrad: Evaluating Acoustic Perception in Audio Language Models [AUTHORS]Koki Ryu, Hitomi Yanaka [ABSTRACT]Recent large audio language models (LALMs) have extended their capabilities beyond lexical speech understanding to tasks requiring paralinguistic and non-linguistic audio understanding, such as emotion recognition. However, these models still struggle to reliably utilize acoustic information, particularly when it co-occurs with spoken linguistic content. Existing benchmarks do not clearly distinguish acoustic perception from higher-level reasoning, making it difficult to identify the source of these failures. To address this issue, we introduce VocalGrad, a benchmark designed to evaluate LALMs' ability to perceive low-level acoustic variations. VocalGrad formulates this problem as a simple binary classification of temporal change direction, enabling evaluation largely independent of higher-level reasoning or semantic understanding. Our benchmark results reveal a substantial gap in acoustic perception between human annotators and state-of-the-art LALMs: while human annotators achieve high accuracy, LALMs perform at near-chance levels. We further conduct linear probing to identify structural bottlenecks in these models. Our analysis reveals two distinct failure modes: for some categories, task-relevant information is preserved in internal representations but not utilized by the language model head; for others, the information becomes progressively less linearly decodable across model layers. Overall, our results reveal substantial limitations of LALMs in modeling acoustic variation under controlled settings and expose representation differences not observable from text outputs. [COMMENTS]NeurIPS 2026 (Evaluations & Datasets Track). Dataset page: https://huggingface.co/datasets/ynklab/vocalgrad [LINK]http://arxiv.org/abs/2610.23745v1 [DATE]2026-10-17 00:54:35+08:00 [CATEGORIES]cs.CL
The Italian lawmaking process: a unified and structured dataset [AUTHORS]Matilde Contestabile, Chiara Ferrara, Giovanni Parrillo, Lorenzo Emer, Andrea Vandin [ABSTRACT]The Italian legislative process is documented in detailed parliamentary records hosted in institutional websites. However, much of this information is available only as unstructured text, limiting its systematic reuse. We applied a controlled, systematic LLM-based extraction procedure, validated against manually annotated records, to convert these texts into ordered and dated events including, e.g., bill introduction, committee examination, plenary examination, and approval. The resulting dataset reconstructs the procedural histories of all 5346 Italian laws enacted between 1985 and 2026, spanning parliamentary terms IX to XIX, for overall 101413 events. Each event records the activity, date, and chamber, with additional procedural details where available. The dataset supports research in the emerging fields of computational legal and legislative studies, potentially combined with process-oriented data science methods to examine procedural paths and legislative timing across parliamentary terms. [LINK]http://arxiv.org/abs/2610.23741v1 [DATE]2026-10-17 00:52:11+08:00 [CATEGORIES]cs.CL
Recognizing Study Lineage Is Not Enough: Over-Crediting Same-Trial Evidence in LLM-Based Medical Verification [AUTHORS]Mukand Krishna, Anoushka Harit, Chenghao Xiao, Zhongtian Sun [ABSTRACT]Publication count is not study count. A model can identify shared trial lineage correctly in a direct task yet over-credit a second publication from the same trial when aggregating evidence. We evaluate this behavior in medical LLM verification using 14 source-audited clinical families and five model conditions. The families span provenance, clinical applicability, result direction, and composition; the provenance families provide the direct lineage tests. Recognition and aggregation are separate stateless calls, and repeated calls assess within-input stability. Across all three original provenance families, GPT-5.5 identifies the independent update in 10/10 recognition calls per family but over-credits the same-trial report in 5/5 separate aggregation calls per family. Selected Qwen3-4B and Ministral-3-8B conditions show the same cross-task contrast. A study-grouping instruction reduces GPT-5.5 over-credit from 4/5 to 0/5 in each later pure-provenance family. Other model conditions show mixed behavior, and rejected outputs limit several comparisons. The results identify a behavioral gap between elicited lineage recognition and evidence aggregation. The code and data artifacts are available at https://anonymous.4open.science/r/study-lineage-review-artifact. [COMMENTS]20 pages, 5 figures, 17 tables; 8 pages of main content, with references and appendices thereafter [LINK]http://arxiv.org/abs/2610.23721v1 [DATE]2026-10-17 00:42:10+08:00 [CATEGORIES]cs.CL
Beyond Consensus: Perspectivist Modeling and Evaluation of Annotator Disagreement in NLP [AUTHORS]Yinuo Xu, David Jurgens [ABSTRACT]Annotator disagreement is widespread in NLP, particularly for subjective and ambiguous tasks such as toxicity detection and stance analysis. While early approaches treated disagreement as noise to be removed, recent work increasingly models it as a meaningful signal reflecting variation in interpretation and perspective. This survey provides a unified view of disagreement-aware NLP methods. We first present a domain-agnostic taxonomy of the sources of disagreement spanning data, task, and annotator factors. We then synthesize modeling approaches using a common framework defined by prediction targets and pooling structure, highlighting a shift from consensus learning toward explicitly modeling disagreement, and toward capturing structured relationships among annotators. We review evaluation metrics for both predictive performance and annotator behavior, and noting that most fairness evaluations remain descriptive rather than normative. We conclude by identifying open challenges and future directions, including integrating multiple sources of variation, developing disagreement-aware interpretability frameworks, and grappling with the practical tradeoffs of perspectivist modeling. [COMMENTS]Accepted to EMNLP 2026 [LINK]http://arxiv.org/abs/2601.09065v3 [DATE]2026-10-17 00:39:26+08:00 [CATEGORIES]cs.CL
The Cognitive Cost of LLM Writing: A Reading-Time Comparison with Human Text [AUTHORS]Kamer Cekini, Elisabetta Biondi, Chiara Boldrini, Andrea Passarella, Marco Conti [ABSTRACT]Although linguistic differences between human-written and Large Language Model (LLM)-generated texts have been extensively studied, their implications for the cognitive effort involved in reading remain less explored. In psycholinguistics, reading time is commonly used as an indirect measure of cognitive processing effort. Building on this perspective, we investigate whether human-written texts and their LLM-generated rewrites exhibit differences in predicted reading time that may reflect differences in processing demands for readers. To this end, we introduce RoBERTa-RT, a model trained on human eye-tracking data to predict word-level reading times. We evaluate the model on both internal and external test sets, construct a paired dataset of 496 human-written articles and their LLM-generated rewrites, and apply RoBERTa-RT to compare their predicted reading times. Our findings show that LLM-generated rewrites tend to be shorter and have slightly lower predicted total reading times, while exhibiting higher predicted reading times per word. These differences are associated with systematic linguistic changes, including longer and less frequent words, greater lexical diversity, and higher contextual predictability. From a cognitive perspective, our findings suggest that LLM-generated writing may redistribute reading demands: reducing the amount of text to process while increasing the predicted processing effort associated with individual words. This highlights the importance of considering not only the linguistic characteristics of machine-generated content, but also its potential implications for human cognitive processing. [LINK]http://arxiv.org/abs/2610.23715v1 [DATE]2026-10-17 00:35:41+08:00 [CATEGORIES]cs.CL
GLOBE: Evolving Structured Agentic Memory for E-commerce Product Optimization in Generative Search [AUTHORS]Hyunseo Kim, Suil Choe, Sukkyoung So, Daeyeon Yi [ABSTRACT]Optimizing product listings for generative search requires language agents not only to rewrite effectively but also to continually self-improve by retaining successful strategies across diverse categories. However, existing approaches often treat structured listings as free-form text, leaving effective strategies implicit in prompts, search trajectories, or rewritten outputs rather than accumulating them as a reusable agentic memory. To address this limitation, we introduce GLOBE, an agentic framework that drives continual self-improvement by storing optimization knowledge in an evolving handbook of explicit, field-scoped behaviors. In each episode, an Editor applies relevant behaviors and receives rank-based feedback, which a Reflector converts into behavior updates. Crucially for dependable execution in real-world settings, a Moderator removes redundant proposals, a Curator commits accepted changes, and consistently harmful behaviors are deactivated. This cycle transforms episodic feedback into a persistent, field-scoped optimization memory. Experiments show that structured behaviors outperform monolithic prompt optimization, while feedback-driven evolution and moderation further improve effectiveness and reduce downside risk. The learned behaviors also remain effective on unseen product categories, suggesting that agentic memory evolution supports interpretable and transferable product optimization. [COMMENTS]Accepted to the Second Workshop for Research on Agent Language Models(REALM) at EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23709v1 [DATE]2026-10-17 00:33:17+08:00 [CATEGORIES]cs.CL
VISTA: Steering LLMs with Latent User Values Inferred from a Few Survey Answers [AUTHORS]Yinuo Xu, Jadie Adams, Brian Hu, Shivani Kumar, David Jurgens, Arslan Basharat [ABSTRACT]Personalizing a language model requires knowing what a person values, which stays largely consistent across tasks. Yet most methods either learn task-specific preferences or place the user's profile in the prompt, where models struggle to act on it. We introduce VISTA (Value-Informed Steering and Targeted Acquisition), which infers a latent profile from users' (full or partial) survey answers, and maps the profile through a hypernetwork to a user-specific steering intervention in a frozen LLM to select answer choices or generate free-text responses. Given the same evidence, VISTA matches or beats in-context profiles on synthetic preferences (Food2D), the World Values Survey (WVS), and PrimeX, without adding prompt tokens. To reduce the cognitive burden of completing full surveys, VISTA also chooses which questions to ask and when to stop: on WVS it stops after 6% of the questionnaire while retaining 94% of the full-history gain, where standard adaptive testing retains at most 42%. The dominant steering direction is an interpretable value axis on which VISTA places unseen users accurately. Our results point toward pluralistic alignment that adapts to individuals at deployable cost. [COMMENTS]8 pages, 4 figures [LINK]http://arxiv.org/abs/2610.23702v1 [DATE]2026-10-17 00:28:57+08:00 [CATEGORIES]cs.CL
IndicDetect: Evaluating Cross-Lingual LLM-Generated Text Detection for Hindi, Telugu, and Tamil [AUTHORS]Bhaskar Ganesh Devalla, Junchao Wu, Nilesh Dokuparthi, Greeshma Yaluru, Tatiana Muniz Rodriguez, Lidia S. Chao, Derek F. Wong [ABSTRACT]The rapid proliferation of LLMs has further heightened the need to develop dependable AI-generated text detection, especially beyond English. Nevertheless, current benchmarks pay little attention to Indic languages and test detectors in idealized settings that do not represent the real world. We present a generalized benchmark for AI-generated text detection in Hindi, Telugu, and Tamil, which we call IndicDetect, designed to assess the robustness of detectors under realistic distribution shifts. IndicDetect comprises highly curated human-written texts matched with LLM-generated counterparts across various domains and generators, and systematically evaluates detectors in the presence of domain shift, generator shift, and adversarial perturbation. Using a single and repeatable evaluation scheme, we evaluate a wide range of statistical and neural detectors. We find substantial robustness failures: supervised neural detectors perform well in-distribution, while training-free methods degrade considerably under unseen generators and adversarial attacks. The severity of these failures varies across languages, with Hindi exhibiting the largest overall degradation under adversarial perturbations. These results highlight that the primary weakness of existing detectors in Indic settings lies in their robustness, not in their peak accuracy. IndicDetect provides standard data splits, an evaluation protocol, and baselines to establish a robust, language-aware foundation for AI-generated text detection in Indic scripts. [LINK]http://arxiv.org/abs/2608.29919v2 [DATE]2026-10-17 00:26:48+08:00 [CATEGORIES]cs.CL
Paved with True Intents: Intent-Aware Training Improves LLM Safety Classification Across Training Regimes [AUTHORS]Jeremias Ferrao, Niclas Müller-Hof, Iustin Sîrbu, Traian Rebedea, Yftah Ziser [ABSTRACT]We argue that safety classifiers should model user intent as an explicit signal between the prompt and the final label. To study this, we introduce AIMS, a human-annotated dataset of 1,724 difficult safety prompts, each paired with an intent description and harm label. We use AIMS to evaluate intent-aware training across supervised fine-tuning, preference learning, reasoning distillation, and reinforcement learning. Despite its size, AIMS enables competitive safety classifiers across training regimes: DPO from model-generated intent errors improves over SFT, and intent-conditioned distillation outperforms reasoning-only distillation in most teacher-student pairs. Most notably, directly rewarding intent faithfulness with GRPO yields the strongest average performance across five external safety benchmarks, while our intent-aware models form the inference latency-F1 Pareto frontier. These results show that faithful intent modeling is a compact, high-quality supervision signal for more robust safety classifiers. [LINK]http://arxiv.org/abs/2606.27210v2 [DATE]2026-10-17 00:09:43+08:00 [CATEGORIES]cs.CL
Beyond the Benchmark: Conditioning, Domains and Failure Modes of Qwen3-TTS [AUTHORS]Séverin Baroudi, Yanis Labrak, Petr Motlicek, Hervé Bredin, Ricard Marxer [ABSTRACT]Qwen3-TTS reports state-of-the-art zero-shot voice cloning, but its technical report leaves open how the model should be conditioned, how it behaves outside read speech, and how it fails. We characterise its two cloning interfaces, in-context learning (ICL) and X-Vector, against Chatterbox, XTTS-v2 and F5-TTS on read, meeting, emotional and air-traffic-control (ATC) speech, using eighteen objective measures, an LLM judge, a spoofing countermeasure and a listening test. Neither interface dominates: ICL clones timbre more closely from correctly transcribed references, X-Vector clones timing, and ICL degrades sharply when a short reference transcript is wrong. Domain explains four times more variance than system choice. Qwen3-TTS leads on intelligibility and pitch register but trails on prosody, and the objective measures rank systems differently from listeners. An off-the-shelf countermeasure flags its clones no more often than genuine recordings, and on ATC radio it fails for every system. Code and audio are released. [LINK]http://arxiv.org/abs/2610.23678v1 [DATE]2026-10-17 00:09:20+08:00 [CATEGORIES]cs.CL
Learning to Reason About Implicit Multimodal Harm: From Diagnosis to Generalization [AUTHORS]Anisha Saha, Varsha Suresh, Teodora Kamova, Sophia Wiedmann, Timothy Hospedales, Vera Demberg [ABSTRACT]A benign image paired with a benign caption can become harmful only in combination, making detection fundamentally a cross-modal reasoning problem. Existing multimodal harm benchmarks evaluate only classification accuracy, leaving unclear whether models identify the cross-modal mechanism that produces harm. We introduce MuPHI (Multimodal Pragmatic Harm Interpretation), a controlled diagnostic benchmark in which harm arises purely from image-text composition, with benign counterfactuals and annotated harm rationales. Using MuPHI, we systematically evaluate where compositional harm reasoning breaks in vision-language models (VLMs). MuPHI reveals that VLMs reliably ground each modality yet fail to bind the evidence into the harm mechanism and their predictions respond weakly to label-determining counterfactual interventions. Moreover, existing harm detection approaches trained on labels substantially improves in-domain (ID) detection but transfers poorly across harm distributions. We therefore propose Compositional Harm Interpretation Learning (CHIL), which rewards evidence identification in each modality, cross-modal binding and verdict consistency in the reasoning chain rather than just predicted label. CHIL improves 81% OOD transfer evaluations by an average of +2.5 macro-F1 and correct counterfactual flip rate by up to +3.8 and +5.4 % points, while also improving overall reasoning quality over label-only optimization. These results suggest that optimizing how a model reasons, not just what it concludes, offers a promising direction for multimodal safety systems that generalize beyond dataset-specific correlations. [LINK]http://arxiv.org/abs/2605.29951v2 [DATE]2026-10-17 00:01:16+08:00 [CATEGORIES]cs.CLcs.LG
How Reliable Are System-One Decisions in Medicine? A Comparative Study of Jev, Jev-like Models, and LLMs [AUTHORS]Yanrui Du, Danyang Zhao, Sendong Zhao, Wenhao Li, Yibo Gao, Haochun Wang, Zhuoer Xu, Shiwen Cui, Bing Qin [ABSTRACT]Jev has attracted considerable attention for its rapid, generalizable decision-making, yet its medical reliability remains underexplored. To address this gap, our study compares various types of system-one decision models across 15 medical datasets comprising 8,742 instances under a unified protocol. Beyond accuracy, we assess decision stability and the ability to recognize unreliable decisions. Our results show that Jev rivals leading models on primarily text-driven tasks but is less effective on numerically driven tasks, revealing a distinct capability boundary. Under question-format variations, Jev maintains strong decision consistency and small accuracy fluctuations. StartLux-27B exhibits comparable consistency, while D3-27B achieves higher overall accuracy and a lower mean decision reversal rate. But these strengths do not extend to reliable abstention awareness: most evaluated models show limitations in both recognizing missing correct answers and using optional abstention. Moreover, we find that Jev's probability-based answer selection can be improved by changing how decisions are scored: normalizing option-wise binary scores yields higher retained accuracy than choice scoring. Taken together, our findings provide early guidance for improving and applying decision models in medicine. [LINK]http://arxiv.org/abs/2610.18529v2 [DATE]2026-10-17 00:01:04+08:00 [CATEGORIES]cs.CL
Coordination Matters: Process-Aware Evaluation of Cooperative Multi-Agent Reinforcement Learning [AUTHORS]Maria Ana Cardei, Matthew Landers, Afsaneh Doryab [ABSTRACT]Cooperative multi-agent reinforcement learning (MARL) benchmarks commonly emphasize aggregate outcomes such as return, success rate, or completion time. While essential, these metrics can obscure how agents coordinate and fail to distinguish policies that achieve similar outcomes through different interaction processes. We propose a process-aware evaluation framework that complements aggregate performance with diagnostics of interference, coordinated task completion, participation, and coordination efficiency. We evaluate the framework in STAT, a controlled spatial task-allocation testbed, and examine its broader applicability in two established cooperative MARL benchmarks. Across diverse value-based and actor-critic methods, we find policies with nearly identical return can exhibit substantial differences in coordination behavior. Controlled scaling in STAT reveals distinct process signatures for spatial, task, and agent scaling, while a reward-shaping ablation shows that explicitly penalizing assignment conflicts can alter coordination behavior with little change in task return. Results in the established benchmarks further show that the framework remains informative under different coordination mechanisms. These findings support process-aware evaluation as a complement to outcome-based benchmarking. The framework provides a common structure for selecting and interpreting environment-specific diagnostics that characterize not only what cooperative policies achieve, but how they coordinate to achieve it. [COMMENTS]Under review [LINK]http://arxiv.org/abs/2605.06557v2 [DATE]2026-10-17 01:58:51+08:00 [CATEGORIES]cs.LG
Structural Causal Bottleneck Models [AUTHORS]Simon Bing, Jonas Wahl, Jakob Runge [ABSTRACT]We introduce structural causal bottleneck models (SCBMs), a novel class of structural causal models in which causal effects between high-dimensional variables are mediated by low-dimensional summary statistics, or *bottlenecks*. SCBMs provide a flexible framework for mechanism-specific, target-dependent dimension reduction while remaining estimable via standard learning algorithms. We prove that bottleneck variables are identifiable up to bijection from observational data, and validate this experimentally. We then study causal effect estimation in linear SCBMs under three standard adjustment strategies. For instrumental variables, we identify a population-level non-identifiability of naïve two-stage least squares when all variables are high-dimensional, and show that bottleneck representations restore identification. For mediation and backdoor adjustment, we prove that conditioning on bottleneck variables yields valid effect estimates with explicit finite-sample efficiency gains over conditioning on full high-dimensional variables. We argue that SCBMs provide a principled alternative to existing causal dimension reduction frameworks such as causal representation learning and causal abstractions. [COMMENTS]NeurIPS 2026 [LINK]http://arxiv.org/abs/2603.08682v3 [DATE]2026-10-17 01:56:06+08:00 [CATEGORIES]cs.LG
Task-aligned spectral inductive biases in parameterized quantum circuits [AUTHORS]Felix Frohnert, Adrián Pérez-Salinas, Sebastian Lange, Marcus Schwaebisch, Evert van Nieuwenburg, Vedran Dunjko [ABSTRACT]Parameterized quantum circuits are widely investigated in quantum machine learning, yet how their architecture can provide task-aligned inductive biases remains poorly understood. We study data reuploading quantum circuits with two distinct inputs: a signal that varies between samples and a fixed, task-defined spectral query that specifies how the signal should be probed. We show that encoding the query can change the circuit's hypothesis class over the signal by modulating how its existing signal frequencies are combined, without introducing new sample-dependent information. This mechanism matches tasks built on nonuniform Fourier transforms, whose evaluation frequencies specify the query and whose sampling positions guide its encoding. We examine this connection in a classification task inspired by synthetic aperture radar, where acquisition geometry determines a query for classifying unfocused spectral patches before image formation. Across small-scale spectral and radar-like tasks, aligned queries improve learning from limited data over parameter-matched circuits with mismatched or absent queries. These results provide a route from known spectral task structure to the design and validation of inductive biases in parameterized quantum circuits. [LINK]http://arxiv.org/abs/2610.23834v1 [DATE]2026-10-17 01:53:19+08:00 [CATEGORIES]cs.LG
ENFR: Evidence Networks with Flow-Matched References for Amortized Simulation-Based Model Inference [AUTHORS]Kai Lehman, Jameson Dong, Carolina Cuesta-Lazaro, Niall Jeffrey, Benjamin D. Wandelt, Francisco Villaescusa-Navarro, Sven Krippendorf, Jochen Weller [ABSTRACT]The evidence, $\mathcal\{Z\} = p(x | \mathcal\{M\})$, of a data-generating model $\mathcal\{M\}$ is a powerful tool in Bayesian frameworks: it enables model inference, model misspecification tests, and goodness-of-fit assessment as an alternative to standard discrepancy statistics, e.g. $χ^2$. While versatile, the evidence is notoriously difficult to compute, even when the likelihood is available. In both likelihood-based and simulation-based inference, estimating $\mathcal\{Z\}$ provides an immediate solution to high-dimensional goodness-of-fit or model misspecification tests: using $\mathcal\{Z\}(x)$ as the test statistic induces a minimum-volume level-$α$ test. Such tests are essential to assess model fidelity, which is necessary to trust simulation-based parameter inference. Existing learned approaches largely target Bayes factors, which only measure evidence ratios. Instead, we provide accurate, fast, and amortized estimates of the absolute evidence. We present ENFR: Evidence Networks with Flow-matched References. Building on the Evidence Network framework of Jeffrey & Wandelt (2024), the proposed method first learns Bayes factors with respect to a flow-matched reference, from which the absolute evidence is recovered. We show improvement in estimation of $\mathcal\{Z\}$ over the flow-based baseline in all applications considered and the best performance of all density estimation methods considered in our benchmarks. We furthermore introduce and provide the proof for an architecture that jointly learns the individual evidences of a set of models $\\{\mathcal\{M\}_i\\}$. We demonstrate the performance and scaling behavior of these approaches on data from a variety of scientific applications. [COMMENTS]28 pages, 6 figures [LINK]http://arxiv.org/abs/2610.23825v1 [DATE]2026-10-17 01:48:26+08:00 [CATEGORIES]cs.LG
TANGO: RNA Topology and Geometry Co-Design [AUTHORS]Tianmeng Hu, Biao Luo, Ke Li [ABSTRACT]Designing RNA sequences that satisfy target tertiary-structure objectives is an important problem in RNA engineering, with broad applications in synthetic biology and therapeutics. Existing 3D RNA design methods have made substantial progress, but they largely treat the target fold as a geometry-only objective. This is incomplete for RNA, whose functional folds are specified not only by tertiary geometry but also by secondary base-pairing topology. We therefore formulate RNA inverse design as a topology-and-geometry co-design problem and propose TANGO, a multi-objective multi-agent reinforcement learning framework for this task. TANGO adopts a staged curriculum: it first learns a topology-aware policy from secondary-structure targets, then transfers this policy to tertiary-structure targets to train a general co-design model, and finally applies frontier-guided fine-tuning to refine target-specific trade-offs among secondary-structure fidelity, tertiary-structure fidelity, and sequence diversity. Experiments show that TANGO generates diverse and novel designs, improving tertiary- and secondary-structure performance by 18.3% and 52.7%, respectively, over the strongest baseline. [COMMENTS]NeurIPS 2026 [LINK]http://arxiv.org/abs/2610.20277v2 [DATE]2026-10-17 01:46:35+08:00 [CATEGORIES]cs.LG
Learning without relearning physical symmetry: an SO(3)-equivariant convolutional network (CNN) on the sphere [AUTHORS]Joshua Williamson, Niall Jeffrey, Lorne Whiteway [ABSTRACT]Many of the most important datasets in the physical sciences live natively on the sphere, and these data often carry no preferred position or orientation. Using methods that do not reflect this symmetry leads to far higher simulation requirements, which can be catastrophic for scientific analyses (forcing reliance on cheap, imperfect models that bias results). We present a novel spherical implementation of a gauge equivariant mesh architecture that is specifically designed for the Hierarchical Equal Area isoLatitude Pixelization (\healpix\{\}) of the sphere. The $\mathrm\{SO\}(3)$-equivariant convolution is built from steerable filters, parallel transport and spin-weighted features. The architecture supplies the exact symmetry, so the network does not spend training data learning it. The convolution operations are constructed to give provably exact equivariance, given a band-limit. Our full CNNs can achieve SO(3)-invariance error at per-cent-level and lower, which is controlled through well-understood hyperparameters. On benchmarks, our technique gives the most accurate dense gradient regression and competitive cosmic microwave background foreground removal and denoising. The improved simulation efficiency matters most in simulation-based inference (SBI), where high-fidelity simulations are expensive and neural compressors require large training sets, so a fixed budget forces analyses with less accurate simulators. In cosmology lensing SBI, our technique yields the tightest posteriors at matched simulation count. This approach is a novel method that will enable new convolutional neural network analyses, including generative and inference tasks, for processes and data on the sphere. [LINK]http://arxiv.org/abs/2610.23821v1 [DATE]2026-10-17 01:45:43+08:00 [CATEGORIES]cs.LG
CocoRNA: Collective RNA Design with Cooperative Multi-agent Reinforcement Learning [AUTHORS]Tianmeng Hu, Biao Luo, Ke Li [ABSTRACT]Designing RNA sequences that reliably fold into specific secondary structures is essential for understanding their biological functions but remains a challenging computational problem. We propose CocoRNA, a cooperative multi-agent reinforcement learning framework for RNA inverse design. CocoRNA simplifies the design task by decomposing it into smaller sub-problems, each solved collaboratively by multiple agents. This approach reduces the complexity of the problem and improves the exploration of design policies. During training, a centralized critic uses global structural information to guide the agents, enabling them to jointly optimize their design strategies. As a result, CocoRNA learns high-quality RNA design policies that generalize effectively to unseen structures without additional training. Experiments on the Rfam dataset demonstrate that CocoRNA substantially outperforms state-of-the-art methods in both success rate and design speed. Further experiments on other biological sequence design tasks highlight the effectiveness and broad potential of CocoRNA for complex design tasks. [COMMENTS]ICML 2026 [LINK]http://arxiv.org/abs/2610.23820v1 [DATE]2026-10-17 01:43:25+08:00 [CATEGORIES]cs.LG
Learning Personalized Memory Gating from Delayed One-Sided Feedback [AUTHORS]Kaixuan Ji, Zixiang Chen, Wenting Zhao, Cheng Qian, Heng Wang, Weizhi Zhang, Jielin Qiu, Liangwei Yang, Quanquan Gu, Silvio Savarese, Shelby Heinecke, Huan Wang [ABSTRACT]Long-term memory enables language-model agents to personalize future interactions, but a finite memory bank cannot preserve every extracted fact. Existing memory gating policies are typically shared across users and therefore fail to account for user-specific patterns of future reuse. In this paper, we study memory gating at the fact level. We first propose a retrieval-based labeling strategy that labels each fact by whether it is later retrieved and incorporated into responses. We then formulate online memory gating as contextual apple tasting with delayed one-sided feedback, and propose Warm-Start UCB, which combines offline learning from existing data with online adaptation to the current user. Our theoretical analysis establishes a high-probability regret upper bound of $\tilde\{O\}(Tε\wedge d\sqrt\{T\})$ when the delay window is at most $O(\sqrt\{T\})$, where $T$ is the total number of rounds, $ε$ is the initialization error, $d$ is the dimension of the feature map, and $\tilde\{O\}(\cdot)$ hides constant and poly-logarithmic factors. The results of our experiment demonstrate that our method outperforms all baselines under different memory budgets and history lengths, establishing it as a promising personalized memory gating policy. [LINK]http://arxiv.org/abs/2610.23812v1 [DATE]2026-10-17 01:34:45+08:00 [CATEGORIES]cs.LG
Explainable Machine Learning Analysis of Temperature Effects on Mechanical Strength of Fiber-Reinforced High-Performance Concrete [AUTHORS]Jagaran Chakma, Zhiguang Zhou, Badhan Chakma, Priyanta Chakma [ABSTRACT]Predicting the compressive strength of fiber-reinforced high-performance concrete subjected to temperature exposure is challenging because to the intricate interdependencies of mixture composition, fiber properties, curing conditions, and temperature. A data-driven framework to predict compressive strength utilizing 1,204 experimental records containing a diverse array of material designs and exposure situations. This study evaluates eight machine-learning models and demonstrate that ensemble approaches routinely surpass individual learners. Both gradient boosting and extreme gradient boosting exhibit superior performance on unseen data, each attaining a R^2 of 0.903. Gradient boosting produces the minimal root mean square error (6.67 MPa), whereas extreme gradient boosting achieves the lowest mean absolute error (3.97 MPa) and mean absolute percentage error (10.60%). The models are further interpreted utilizing Shapley additive explanations and partial dependence analysis, which reveal temperature as the primary factor influencing compressive strength, succeeded by fine aggregate, superplasticizer, cement content, water-to-binder ratio, and water content. Strength diminishes significantly with rising temperature, albeit the extent of deterioration is contingent upon the specific combination design. The best performing models are subsequently implemented in a Python-based graphical user interface for expedited prediction. Our findings offer a precise and comprehensible methodology for evaluating temperature-dependent compressive strength and facilitate the practical, data-driven design of fiber-reinforced high-performance concrete. [COMMENTS]29 pages [LINK]http://arxiv.org/abs/2610.23811v1 [DATE]2026-10-17 01:34:24+08:00 [CATEGORIES]cs.LG
LeanPRM: Lean-Grounded Process Supervision for Natural Language Proof Verification [AUTHORS]Rodney Lafuente-Mercado, Salman Rahman, Jingyan Shen, James Shiffer, Saadia Gabriel, Pavel Izmailov [ABSTRACT]AI systems now write long and difficult mathematical proofs, and checking their step-by-step validity is becoming a bottleneck. A learned verifier could check proofs cheaply, but training one requires a correctness label on every step, and such labels are hard to obtain for open-ended proofs. Existing methods take them from human annotators, which is costly, or derive them from final answers, which proofs do not have, and formalizing each proof in Lean requires expert effort. We introduce LeanPRM, a framework that uses the Lean proof assistant to create step-level labels for natural-language proofs. A model writes a natural-language proof together with a paired formal proof, without feedback from Lean. Lean then labels every step, and a second model discards failures caused only by the formal syntax, so that the negative labels mark errors in the mathematics. LeanPRM contains 10,634 proofs with 152K labeled steps and no human annotation. A Qwen3-4B verifier trained on LeanPRM raises exact step-label accuracy on a held-out set of 200 proofs (LeanPRM-Bench) from 48.2 to 74.7, as opposed to 60.3 when the same base model is trained on PRM800K. It also selects better proofs in best-of-N selection on IMO-level problems, and outperforms the PRM800K-trained verifier on three of four external verification benchmarks. Training on LeanPRM improves verifiers at every size from 0.6B to 8B parameters on LeanPRM-Bench, Hard2Verify and IMO-GradingBench. Given the labels need no human annotation, LeanPRM can grow with the supply of formal statements, offering a path to step-level feedback on AI-generated proofs that are hard for humans to check. [COMMENTS]29 pages, 9 figures. Project page: https://rlafuente.com/lean-prm [LINK]http://arxiv.org/abs/2610.17697v2 [DATE]2026-10-17 01:31:06+08:00 [CATEGORIES]cs.LG
Learning What to Say to Your VLA: Mostly Harmless Vision Language Action Model Steering [AUTHORS]Hyun Joe Jeong, Gokul Swamy, Andrea Bajcsy [ABSTRACT]Vision-Language-Action (VLA) models provide a natural language interface to robot control, but the mapping from language to behavior is often brittle and unintuitive: semantically similar instructions can induce drastically different behaviors, while some capabilities may not be elicitable through prompting alone. As a result, both human instructions and zero-shot language models can fail to reliably steer VLAs toward successful task execution. In this work, we propose a framework that interactively searches for language sequences that improve closed-loop VLA task performance, distills these sequences into a test-time language feedback policy (LFP), and learns an improvement head that predicts when language steering will improve performance. We conformalize this improvement head to prevent harmful steering interventions, where the LFP decreases task performance relative to the original instruction on out-of-distribution scenarios. Crucially, our approach operates on arbitrary frozen pre-trained VLAs, requiring neither access to the original training distribution nor fine-tuning of the underlying model. On seen environments, our conformalized LFP improves base VLA performance by 24.7% in simulation and 65.0% in hardware. On visual and semantic perturbations with calibration samples, our conformalized LFP has strong harmlessness guarantees, and produces recovery behaviors not observed with open-loop prompting. [COMMENTS]CoRL 2026 [LINK]http://arxiv.org/abs/2606.12299v2 [DATE]2026-10-17 01:28:27+08:00 [CATEGORIES]cs.LG
RCPN: towards robust pseudo-labels in unsupervised CLIP adaptation via consistent prototypes and neighborhoods [AUTHORS]Eman Ali, Chetan Arora, Muhammad Haris Khan [ABSTRACT]In unsupervised adaptation for vision-language models such as CLIP, pseudo-labels derived from zero-shot predictions often exhibit significant noise, particularly under domain shifts or in visually complex scenarios. Conventional pseudo-label filtering approaches, which rely on fixed confidence thresholds, tend to be unreliable in fully unsupervised settings. In this work, we propose a novel adaptive pseudo-labeling framework that enhances CLIP's adaptation performance by integrating prototype consistency and neighborhood-based consistency. The proposed method comprises two key components: PICS, which assesses pseudo-label accuracy based on in-class feature compactness and cross-class feature separation; and NALR, which exploits semantic similarities among neighboring samples to refine pseudo-labels dynamically. Additionally, we introduce an adaptive weighting mechanism that adjusts the influence of pseudo-labeled samples during training according to their estimated correctness. Extensive experiments on 11 benchmark datasets demonstrate that our method achieves state-of-the-art performance in unsupervised adaptation scenarios, delivering more accurate pseudo-labels while maintaining computational efficiency. [LINK]http://arxiv.org/abs/2507.22075v2 [DATE]2026-10-17 01:28:20+08:00 [CATEGORIES]cs.LG
Parametric Resonance in Primal-Dual Learning for Slowly Non-Stationary Constrained MDPs [AUTHORS]Maxime Baelde [ABSTRACT]Primal-dual methods for constrained Markov decision processes are analyzed under a two-timescale assumption, with non-stationarity entering through two channels. In the additive channel, where the constraint level oscillates, the dual oscillation saturates at a floor set by the moving optimum and is amplified linearly by the timescale ratio. In the parametric channel, where the constraint sensitivity oscillates, a coherent modulation carries no resonant forcing at any depth: the dual update's stiffness modulation cancels the primal update's damping modulation. Instability instead comes from a mismatch between the constraint estimate used by the primal update and the measurement used by the dual update, producing a parametric resonance at twice the pair's natural frequency. The tolerated mismatch grows with the square root of the forcing bandwidth and collapses with the square root of the dual-to-primal step-size ratio. These results hold on the tangent linear-quadratic reduction around an active saddle, for one constraint and for $K$ constraints under a common modulation and an isotropy condition. Driving both updates from the same estimate removes the resonant forcing but introduces a stationary offset equal to the estimator's bias. In a Lagrangian soft actor-critic agent, the multiplier locks onto the sub-harmonic at the timescale ratios its identified linear loop predicts and a coherent cost critic suppresses the lock-in, while a proportional (PID) term, predicted to leave the lock-in in place, suppresses most of it through the projection of the multiplier. [COMMENTS]33 pages, 3 figures. Code: https://github.com/mbaelde/dualdynamics-code [LINK]http://arxiv.org/abs/2610.23788v1 [DATE]2026-10-17 01:19:08+08:00 [CATEGORIES]cs.LG
Mesoscopic learning dynamics of extensive-width neural networks [AUTHORS]Shu Zhou, Jean Barbier, Sebastian Goldt [ABSTRACT]Understanding how stochastic gradient descent learns internal representations in wide neural networks trained on rich target functions is a central problem in machine learning. We study this question in two-layer neural networks learning target functions represented by networks of the same architecture, with arbitrary nonlinearity, and with network width and input dimension growing proportionally. In this regime, standard approaches for describing the learning dynamics require tracking a diverging number of overlaps, which measure correlations between learned and target features, leaving no obvious description in terms of a few order parameters. Here, we develop a mesoscopic description of learning by characterizing the distribution of overlaps. In the early stages of learning, we find that a Gaussian description of overlap statistics reduces the learning dynamics to a few collective variables that track learning of the target function's linear and quadratic components. We then estimate the timescale of specialization, when Gaussian closure breaks down and individual learned features begin to strongly correlate with specific target features. We find that specialization in extensive-rank networks is accelerated via a quadratic shielding mechanism, whereby learning the linear and quadratic components of the target function filters gradient fluctuations. We finally describe how the activation function determines the pattern of specialization and characterize the subsequent refinement of learned features. Together, these results show how collective learning gives way to individual feature discovery and clarify when learning in large neural networks admits a simple statistical description. [LINK]http://arxiv.org/abs/2610.23779v1 [DATE]2026-10-17 01:12:43+08:00 [CATEGORIES]cs.LG
GRAFT-ATHENA: Self-Improving Agentic Teams for Autonomous Discovery and Evolutionary Numerical Algorithms [AUTHORS]Juan Diego Toscano, Zhaojie Chai, George Em Karniadakis [ABSTRACT]Language-model agents plan, execute, and evaluate scientific workflows. However, the problem--method relationships guiding their choices remain implicit, so plausible outputs can violate the assumptions of the scientific problem. We introduce Graph Reduction to Adaptive Factored Trees (GRAFT), which makes this problem-to-method map explicit as an expandable probabilistic structure of admissible problems, methods, and dependencies. Because this structure follows scientific semantics, knowledge from documentation, papers, and domain experts can be encoded, while factorization keeps the structure tractable. Semantic fingerprints of encoded problems measure their similarity, so expertise and experience guide related problems. GRAFT-guided workflows exceeded published expert baselines in physics-informed learning, recovered blood-rheology trends, predicted Apollo Command Module hypersonic loading within 1.8\% of experiment, and established machine-checked error bounds for regularized brain-flow reconstruction and a spectral physics-informed architecture. The representation retains evaluated outcomes and incorporates new methods, allowing scientific knowledge and experience to accumulate. [LINK]http://arxiv.org/abs/2605.11117v3 [DATE]2026-10-17 01:10:34+08:00 [CATEGORIES]cs.LG
Logic-Guided Graph Denoising Diffusion Models [AUTHORS]Samuel Cognolato, Luciano Serafini, Alessandro Sperduti, Andrea Cavallaro [ABSTRACT]Graph diffusion models are competitive approaches for generating complex graph-structured data. However, when such distributions have strict structural constraints, diffusion models may fail to capture them exactly, resulting in the generation of invalid graphs. In this paper, we propose a method for grounding graph diffusion models in domain-specific knowledge via First Order Logic (FOL), leveraging its expressive and interpretable formalism to impose structural constraints on the generative process. We incorporate domain knowledge using classifier guidance, a technique that conditions the diffusion model's distribution on external variables to exert control over the sampled data. During the generative process, classifier guidance steers the denoising trajectory to optimize an objective function, which we define as a fuzzy and differentiable relaxation of the logical rules. Our methodology requires only the specification of constraints in FOL, whose satisfaction is encouraged at inference time to guide a pre-trained generative model. We evaluate our approach by comparing the unconditional generation of knowledge graphs with our logic-guided framework, which incorporates domain ontologies as structural constraints during inference. This approach significantly increases the proportion of generated graphs that satisfy the ontological constraints. [COMMENTS]Manuscript under review at AAAI2027 [LINK]http://arxiv.org/abs/2610.23766v1 [DATE]2026-10-17 01:06:07+08:00 [CATEGORIES]cs.LG
From Flow-Box Invariants to Global Conservation Laws: Machine Learning-Assisted Detection of Singularities and Near-Integrability [AUTHORS]Jimmie Adriazola, Ioannis G. Kevrekidis, Panayotis G. Kevrekidis, Wei Zhu [ABSTRACT]Conservation laws reveal invariant structure in dynamical systems and constrain their evolution to lower-dimensional level sets. Locally, however, per the flow-box theorem, every $d$-dimensional system generically admits $d-1$ functionally independent conservation laws near a regular point even though, generically, only a few such laws persist globally. Data-driven methods typically learn conservation laws on a prescribed region of phase space from which training samples are drawn, leaving their behavior beyond that region unresolved. We formulate this local-to-global continuation problem and develop an \{expanding-domain neural deflation\} framework that tracks the number of independent conservation laws over nested domains. The method recovers the expected $d-1$ local invariants and detects their successive breakdown as the domain expands, relaxing to the expected global conservation law limit. It distinguishes the local invariants (associated with Frobenius integrability) from persistent global structure and shows that the domain scale over which additional laws survive reflects proximity to Liouville integrability. Pointwise losses can localize isolated obstructions, whereas in more complicated dynamics the breakdown of our loss functions is gradual and spatially diffuse. The framework thus provides a data-driven approach for studying the persistence and breakdown of (so-far-learned) conservation laws across phase space. [LINK]http://arxiv.org/abs/2610.23760v1 [DATE]2026-10-17 01:02:39+08:00 [CATEGORIES]cs.LG
When Is Language-Model Advice Worth Buying? A Matched-Budget Study of Predictive Gating in Reinforcement Learning [AUTHORS]Ibne Farabi Shihab, Md Najmus Swaqeeb, Abu Sa-Adat Mohamed Moon-Im Al Ahsan, Joyanta Jyoti Mondal [ABSTRACT]Language-model advice can accelerate reinforcement learning, but every call has a price, and a returned suggestion may be unparsable, unsafe to execute, or no better than the agent's own action. We introduce certified predictive advice gating, a cost-aware controller that forecasts possible advisor responses, values the action actually executed after parsing, execution checks, and cache reuse, and requests advice only when a calibrated lower bound on that value covers the price. Our analysis separates the gap of executed advice, cumulative regret under an intervention-stable fallback, and myopic acquisition loss. On BabyAI GoToObj with Qwen2.5-1.5B and 7B advisors, the gate improves return over no advice by paired $0.029\pm0.016$ and $0.030\pm0.015$ while using more than $97\%$ fewer calls than always asking. Matched-budget comparisons, however, do not resolve an advantage over an early query schedule, so the gain reflects economical advice volume rather than demonstrated state selection. Groupwise calibration raises coverage on positive-value opportunities from $0.47$ to $0.85$, still below the $0.90$ target. A separate headroom-aware gate reports paired gains over an early schedule of $0.032\pm0.012$ and $0.029\pm0.011$ on GoToObj and $0.027\pm0.011$ and $0.028\pm0.012$ on GoToObjDoor, and adds BabyAI and ALFWorld success, decision-stratum coverage, and component timing. These results separate return gains, query placement, and execution screening; finite-sample deployment guarantees remain unestablished. [COMMENTS]41 pages, 6 figures [LINK]http://arxiv.org/abs/2609.29548v2 [DATE]2026-10-17 00:49:25+08:00 [CATEGORIES]cs.LG
LagNet: Lag-Structured Operators for Long-Term Time Series Forecasting [AUTHORS]Zhenan Yu, Guangxin Jiang, Jiuyun Jiang, Jin Yang [ABSTRACT]Long-term time series forecasting maps a finite history window to an extended future horizon. Many existing models focus primarily on the design of input representations and model structures, while paying less attention to the structural form of the history-to-future forecasting operator itself. We examine this operator directly and find that, across model families, sensitivity matrices exhibit recurring diagonal patterns whose diagonals correspond to relative temporal lags. Motivated by this observation, we develop a framework for direct multi-step forecasting based on lag-structured temporal operators. Its basic building block, LagLayer, constructs a Toeplitz matrix by tying history-to-future connections with the same relative lag and incorporates horizon-wise and history-wise scaling coefficients to relax exact sharing. This design enables LagLayer to serve as an effective replacement for temporal linear layers in existing forecasting models. Stacking LagLayer modules with activations and multiple feature maps yields LagNet. Experiments on long-term time series forecasting benchmarks show that our method is competitive with strong baselines, while sensitivity matrix diagnostics and limited-data experiments demonstrate that LagLayer provides an effective and sample-efficient alternative to a linear layer. [LINK]http://arxiv.org/abs/2610.23735v1 [DATE]2026-10-17 00:44:53+08:00 [CATEGORIES]cs.LG
Diffusion-Based Stochastic Operator Networks for Uncertainty Quantification in Stochastic Partial Differential Equations [AUTHORS]Phuoc-Toan Huynh, Richard Archibald, Feng Bao [ABSTRACT]We introduce a novel framework for uncertainty quantification of solution operators associated with stochastic partial differential equations (SPDEs). Although SPDEs play a central role in modeling complex physical systems under uncertainty, their practical use typically requires specifying the magnitude and structure of model uncertainties that are often unknown and difficult to infer from noisy measurements. To address this challenge, we develop a stochastic operator-learning framework that learns directly from noisy data and outputs both a mean solution field and a quantification of uncertainty. The proposed method, namely the Stochastic Operator Network (SON), is constructed by combining the structure of the Deep Operator Network (DeepONet) with Stochastic Neural Networks (SNNs) to model stochasticity and enable probabilistic prediction. The training procedure is carried out by minimizing a Hamiltonian-type loss and optimizing the resulting objective using the Stochastic Maximum Principle. Numerical experiments on benchmark SPDEs under multiple uncertainty sources demonstrate the accuracy and robustness of the proposed method in capturing solution structure and quantifying predictive uncertainty. [LINK]http://arxiv.org/abs/2605.17107v2 [DATE]2026-10-17 00:44:26+08:00 [CATEGORIES]cs.LG
Escaping Parameter Space: Tight Generalization Bounds via Representation Quality [AUTHORS]Niclas Göring, Shuofeng Zhang, Branton DeMoss, Ard A. Louis [ABSTRACT]We derive a tight generalization bound for deterministic neural We derive a tight generalization bound for deterministic neural networks based on their last hidden layer representations. By evaluating representation quality with a parameter-free $k$-nearest-neighbor ($k$-NN) classifier, our certificate avoids traditional parameter-space complexity measures, KL divergences to a posterior, weight norms, and compression. The bound is tight: trained from scratch, it reaches an 8.0 percentage point gap to test error on CIFAR-10 (ResNet-18) and yields the first non-vacuous certificates we are aware of on CIFAR-100; with frozen DINOv2 representations, it certifies at 1.4\% on CIFAR-10 and 23.1\% on 1,000-class ImageNet. It decomposes additively into three empirical, independently measurable quantities: local class geometry ($k$-NN decoder error), decoder-network agreement, and training stability. This allows the bound to act as a diagnostic tool: we demonstrate that actively destroying local class geometry eliminates generalization while leaving dataset memorization intact, identifying class geometry as a causally relevant mechanism for generalization. Finally, we show that learned representations are significantly less sensitive to weight noise than linear readouts, suggesting why parameter-space certificates often struggle to remain tight. [LINK]http://arxiv.org/abs/2610.23729v1 [DATE]2026-10-17 00:43:57+08:00 [CATEGORIES]cs.LG
ECG-SDI: Single-lead Electrocardiogram Enables Continuous Sleep Depth Index Profiling and Scalable Digital Biomarkers for Sleep Health [AUTHORS]Donglin Xie, Qingshuo Zhao, Songchi Zhou, Haoyu Wang, Jingyu Wang, Yuxi Zhou, Thomas Penzel, Shenda Hong [ABSTRACT]Conventional sleep staging provides a coarse-grained representation that obscures within-stage depth variation. We developed ECG-SDI, a single-lead electrocardiography (ECG) framework for continuous sleep-depth profiling and digital biomarker extraction. Using 12,392 overnight polysomnographic recordings, we trained and validated a two-stage deep learning model supervised by conventional sleep-stage labels. Correlations with ordered stages excluding rapid eye movement (REM) sleep were 0.828 in the pooled internal test set and 0.848 in the external Cleveland Family Study; corresponding REM classification AUROCs were 0.954 and 0.977. Overnight trajectories captured within-stage variation. Biomarker-based phenotyping identified a disturbed subtype associated with poorer subjective sleep quality and higher all-cause mortality in the Sleep Heart Health Study (adjusted hazard ratio, 1.20; 95% CI, 1.05-1.38). Among 507 participants with paired recordings before and after continuous positive airway pressure (CPAP), normalized sleep-depth area increased by 6.1% and shallow-sleep burden below 0.3 decreased by 16.0%. Exploratory analyses linked reductions in approximate entropy and depth variability to reduced oxygen desaturation and increased sleep efficiency, respectively (Spearman rho = 0.24 and 0.12); both associations remained positive after covariate adjustment. Mean ECG-SDI within N2 increased by 0.017 (95% CI, 0.004-0.030). A wearable proof-of-concept illustrated the acquisition-to-report workflow. These findings support ECG-based sleep phenotyping and further evaluation of CPAP response monitoring at home. [LINK]http://arxiv.org/abs/2610.23727v1 [DATE]2026-10-17 00:43:07+08:00 [CATEGORIES]cs.LG
Neural Expansion: A Unified Mechanism for How Deep Neural Networks Generalize [AUTHORS]Chashi Mahiul Islam, Samuel Jacob Chacko, Mao Nishino, Canlin Zhang, Xiuwen Liu [ABSTRACT]Deep neural networks demonstrate remarkable generalization despite being highly over-parameterized, challenging traditional statistical learning theories. While several theoretical paradigms such as margin-based bounds, PAC-Bayesian analyses, and neural tangent kernels provide global explanations, they often lack mechanistic insight into how generalization emerges from training dynamics. We introduce Neural Expansion, a unified geometric framework that characterizes generalization through the expansion of directionally-selective ray coverage in the input space. Using Generalization Intervals (GIs), we quantify direction-specific regions around training samples where the model's predictions remain stable. Our extensive empirical studies across diverse architectures and datasets reveal that training leads to progressive expansion of ray coverage, particularly along directions with low Jacobian singular values. This mechanism helps explain a large fraction of correctly classified test inputs, adversarial examples, and even mislabeled or out-of-distribution samples through the geometry of ray coverage rather than instance-level memorization. By connecting input-space curvature to predictive behavior, our framework provides a mechanistic foundation that complements existing global theories of deep learning. [COMMENTS]Accepted at NeurIPS 2026. Includes supplementary material. Code: https://github.com/meetchashi/neural_expansion_neurips_2026 [LINK]http://arxiv.org/abs/2610.23716v1 [DATE]2026-10-17 00:38:02+08:00 [CATEGORIES]cs.LG
Tool Recommendation for Multi-Step Agents: Finding Every Tool a Task Needs [AUTHORS]Sam Partee, River Stevens, Thierry Damiba, Valerie Fanelle [ABSTRACT]A multi-step agent task fails when a single tool it needs is missing, yet the tools an agent may need are too many to list: the roughly 8,200 tools on the platform we study have definitions that together exceed the context window of any model we evaluate. We present Tool Recommendation, a deployed system that gives an agent all of them, hundreds of times more than a typical MCP server exposes, through a fixed interface of three tools. It runs at the gateway between agent and tools, executing every call under the end user's own credentials and checking it against the user's consent for each service and the tools the agent may use. We evaluate it on both a public retrieval benchmark and billed end-to-end session cost. On ToolRet, Tool Recommendation places every tool a query needs in the top ten 57% more often than BM25, a lexical ranking that commercial tool search offers, and compares favorably with published systems up to nine times its size. In paired multi-step workflows, its first recommendation covered every required tool in 93% of Codex sessions, and against a fixed list of 100 tools it cut input tokens by 17% with Codex and 40% with Claude Code and the total bill in all 14 runs, by up to 40%, with no change in task completion. [COMMENTS]11 pages, 3 figures, 5 tables [LINK]http://arxiv.org/abs/2610.23707v1 [DATE]2026-10-17 00:30:38+08:00 [CATEGORIES]cs.LG
Blitz: Accelerating the Diffusion Modules of AlphaFold3 and Boltz-2 [AUTHORS]Vinay Saji Mathew, Soundar Kumara, Gretta Kellogg, William K. M. Lai [ABSTRACT]Diffusion-based protein structure prediction models produce all-atom structures in about 200 denoising steps, and that cost has driven a number of few-step samplers. We present Blitz, a few-step distillation framework that uses a geometric noise schedule to distribute steps more evenly across noise scales, then learns those updates through multistep consistency training. Physical steering and target-free candidate selection help reduce the remaining geometric defects at inference. We apply Blitz to AlphaFold 3 and Boltz-2, reducing denoising from 200 steps to eight. Because published few-step results use different benchmarks and scoring procedures, we compare both distilled models with their teachers and other few-step methods under a common evaluation protocol covering structural accuracy, interface quality, ligand placement, and stereochemical validity. Blitz retains teacher-level structural accuracy, with AF3-Blitz outperforming all evaluated few-step baselines and exceeding its teacher on several accuracy metrics. [COMMENTS]24 pages, 12 figures [LINK]http://arxiv.org/abs/2610.23703v1 [DATE]2026-10-17 00:29:04+08:00 [CATEGORIES]cs.LG
Equivalence testing and run-count design for stochastic machine-learning experiments [AUTHORS]Dmytro Knopov, Nadiya Shvai [ABSTRACT]Claims that two training methods perform equally well often rest on a non-significant difference between two or three runs per method. We develop a protocol, with its power theory, for certifying practical equivalence in per-run metrics such as the AUROC (area under the receiver operating characteristic curve). It rests on a two-level (seed x item) variance model separating between-run dispersion from item-level noise. Equivalence within a pre-specified margin is tested by two one-sided tests (TOST), equivalently by a confidence interval inside the margin; a two-level bootstrap covers composite statistics. A closed-form formula converts a margin and a target power into runs per method; its shifted-t refinement is calibrated against the realised two-level process. Simulations on a 270-cell grid and under non-Gaussian seed laws confirm the formula within its design-relevant range, from five runs per method. For four published runs of a joint energy-based model, the failure to certify equivalence was largely decided by design: with the variance calibrated to those runs, two per method certify with probability near three per cent, whereas certifying 0.01 AUROC at 80 per cent power needs twelve even under exact equality. A second case study on a thousand-seed CIFAR-10 zoo and the MultiBERTs checks the formula against real seed laws (within 0.05 in that range) and resolves comparisons that two or three runs leave undecided. In a two-stage design (pilot, then formula), our experiments suggest that an upper confidence bound on the pilot variance substantially reduces the undersizing seen with its point estimate. [COMMENTS]31 pages, 4 figures, 4 tables. Code, simulation scripts and case-study data: https://github.com/DmytroKnopov/equivtest (release v0.2.0, archived at https://doi.org/10.5281/zenodo.23421510) [LINK]http://arxiv.org/abs/2610.23700v1 [DATE]2026-10-17 00:28:12+08:00 [CATEGORIES]cs.LG
Expressivity of IQP Feature-Map Quantum Kernels via Reproducing Kernel Hilbert Spaces [AUTHORS]Valeria Repetto, Stefano Melchionna, Romina D'Aurizio [ABSTRACT]Quantum kernels are among the most promising methods in quantum machine learning, providing a natural way to embed data into high-dimensional Hilbert spaces via quantum circuits and capturing non-linear relationships in data. Their learning behaviour is governed by the function class they represent, which is formally characterized by the structure of the associated Reproducing Kernel Hilbert Space (RKHS). However, a theoretical characterization of this function class, and of how it depends on the choice of feature map and its hyperparameters, is still lacking. In this work, we focus on the Instantaneous Quantum Polynomial (IQP) feature map and derive a Fourier representation of the associated RKHS generators as finite sums of trigonometric functions evaluated on polynomially transformed input variables, with a finite set of accessible frequencies determined by the circuit construction and scaled by the bandwidth hyperparameter. This representation identifies three bandwidth regimes with qualitatively distinct expressive properties from localised, through multimodal, to periodic. Numerical experiments on synthetic datasets with controlled periodic and multimodal structure, across data dimensions and feature-map configurations, reproduce the three regimes in both supervised classification and unsupervised anomaly detection, benchmarked against a classical Gaussian kernel. These results make the relationship between expressivity and data structure explicit for IQP kernels: expressive power translates into a practical benefit only when the accessible frequencies match with the structure of the data, which provides a criterion for selecting the bandwidth when the periodic or multimodal structure of data is known. [COMMENTS]19 pages, 11 Figures , Supplementary [LINK]http://arxiv.org/abs/2610.23692v1 [DATE]2026-10-17 00:20:46+08:00 [CATEGORIES]cs.LG
HardEval: An Execution-Verified Benchmark for AI-Driven Security Hardening and Compliance Remediation [AUTHORS]Manohar Vellala, Soumya Banerjee, Sandip Roy, Ross Gore, Sachin Shetty [ABSTRACT]Implementing Security Technical Implementation Guide (STIG) controls remains a demanding operating-system administration task. Each natural-language security requirement must be translated into a precise and persistent configuration change. These changes span Secure Shell (SSH) configuration, authentication policy, audit rules, and kernel parameters. Existing benchmarks do not measure whether a model can perform this task on a live system. Code benchmarks evaluate isolated programs without persistent system state, while agentbenchmarks largely focus on applications rather than operating-system configuration. In this paper, we introduce \textbf\{HardEval\}, an execution-verified benchmark for evaluating language-model-assisted security hardening. HardEval is built from Red Hat Enterprise Linux 8 STIG controls that fail on an unhardened baseline system. Expert-written reference remediations and Open Vulnerability and Assessment Language (OVAL) checks serve as hidden ground truth for grading; they are never shown to the model, and each task states only the required security outcome while withholding the implementation. The model produces a Bash script, which is executed on a live virtual machine and graded from fail to pass using the corresponding OpenSCAP/OVAL compliance check. Evaluation of fourteen frontier and open-weight configurations reveals large and uneven capability differences across security domains. A recurring failure mode is plausible security code that does not satisfy the exact compliance condition. For four open-weight configurations, multi-run Wilson confidence intervals preserve the observed ordering. HardEval is publicly available. [COMMENTS]8 pages, 4 figures, 4 tables. Code and data: https://github.com/manoharvellala/Hardeval The paper has 4 figure environments (two are images, two are the example-prompt panels) and 4 tables [LINK]http://arxiv.org/abs/2610.23685v1 [DATE]2026-10-17 00:14:58+08:00 [CATEGORIES]cs.LG
AceFF: A State-of-the-Art Machine Learning Potential for Small Molecules [AUTHORS]Stephen E. Farr, Stefan Doerr, Antonio Mirarchi, Alireza Moradzadeh, Francesc Sabanes Zariquiey, Saee Paliwal, Gianni De Fabritiis [ABSTRACT]We introduce AceFF, a pre-trained machine learning interatomic potential (MLIP) optimized for small molecule drug discovery. While MLIPs have emerged as efficient alternatives to Density Functional Theory (DFT), generalizability across diverse chemical spaces remains difficult. AceFF addresses this by refining the TensorNet2 architecture and training it on a comprehensive dataset of drug-like compounds. This approach yields a force field that balances high-throughput inference speed with DFT-level accuracy. We introduce two models sharing the same architecture: \mbox\{AceFF-2\}, trained on our own dataset of small molecules and supporting the essential medicinal chemistry elements (H, B, C, N, O, F, Si, P, S, Cl, Br, I); and \mbox\{AceFF-2-OMOL\}, trained on a subset of OMol25 containing the AceFF-2 elements plus the common biomolecular ion elements Na and K. The models are explicitly trained to handle charged states. Validation against rigorous benchmarks, including complex torsional energy scans, molecular dynamics trajectories, batched minimizations, and tests of force and energy accuracy, demonstrates that AceFF is state-of-the-art for organic molecules in the accuracy and speed regime important for drug discovery. The AceFF-2 and AceFF-2-OMol model weights and inference code are available at https://huggingface.co/Acellera/AceFF-2.0 and https://huggingface.co/Acellera/AceFF-2-OMOL. [LINK]http://arxiv.org/abs/2601.00581v3 [DATE]2026-10-17 00:13:19+08:00 [CATEGORIES]cs.LG
Interpretable experiential learning based on state history and global feedback [AUTHORS]Anton Kolonin [ABSTRACT]A new interpretable experiential learning model based on state history and global feedback is presented. It is capable of learning a behavioral model represented by a transition graph between sets of states, with transitions attributed with utility and evidence count. This model is expected to be suitable for solving reinforcement learning problem in resource-constrained environments. The model was thoroughly evaluated on the OpenAI Gym Atari Breakout benchmark, demonstrating performance comparable to some known neural network-based solutions. [COMMENTS]5 figures. This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible [LINK]http://arxiv.org/abs/2605.00940v2 [DATE]2026-10-17 00:12:02+08:00 [CATEGORIES]cs.LG
Would you still call this Dax? Novel Visual References in VLMs and Humans [AUTHORS]Ada Defne Tür, Gaurav Kamath, Joyce Chai, Siva Reddy, Benno Krojer [ABSTRACT]Vision-language models (VLMs), like human learners, are frequently exposed to new visual concepts, but how they map novel visual references to language after exposure remains largely underexplored, particularly when those references contradict prior knowledge from pre-training. To study this, we present the Novel Visual References Dataset (NVRD): 19,176 images spanning 90 visual concepts across different levels of visual novelty, each with up to 20 increasingly perturbed versions of the original object to probe generalization. Unlike prior work on visual augmentations of familiar concepts, NVRD comprises entirely novel, open-ended stimuli constructed from scratch, mirroring how humans encounter genuinely new concepts. We evaluate 3 open- and 2 closed-source models alongside 2,400 human judgments for direct human-model comparison, and find that (i) models struggle to acquire novel concepts in-context when they contradict prior knowledge, and (ii) while models and humans show correlated sensitivity to visual perturbations, models significantly overgeneralize, extending learned labels to stimuli that humans reject. We contribute NVRD as a corpus and benchmark for research on visual concept learning in both humans and machines. [COMMENTS]EMNLP Main Conference 2026 [LINK]http://arxiv.org/abs/2606.05409v4 [DATE]2026-10-16 23:49:10+08:00 [CATEGORIES]cs.CL
Mitigating Cross-Lingual Cultural Inconsistencies in LLMs via Consensus-Driven Preference Optimisation [AUTHORS]Lucas Resck, Isabelle Augenstein, Anna Korhonen [ABSTRACT]Multilingual large language models (MLLMs) often answer culturally ambiguous questions inconsistently across languages, even for a fixed user persona -- e.g., Shakespeare in English but Cervantes in Spanish for the same British user. To robustly quantify this Cross-lingual Cultural Inconsistency, we introduce Singleton Fleiss's $κ_S$, a metric mathematically resilient to hallucinations. For mitigation, we propose Cross-lingual Cultural Consistent Preference Optimisation (C-3PO), a consensus-driven alignment framework. C-3PO raises $κ_S$ by 0.07--0.10 over unaligned models, above every prompting and representation steering baseline averaged across personas, whilst largely preserving explicit user identities, cultural neutrality and intrinsic cultural knowledge; where it does not, a soft-consensus variant recovers most of the cost. Although trained only on multiple-choice questions, C-3PO also improves consistency in open-ended generation. Empirical evaluations demonstrate this inconsistency is more pronounced for lower-resource languages like Indonesian and Persian. Finally, early decoding of intermediate layers reveals that MLLMs implicitly personalise outputs towards the prompt language's stereotypical culture as forward-pass representations stabilise. [COMMENTS]25 pages, 14 figures, 17 tables [LINK]http://arxiv.org/abs/2605.12515v3 [DATE]2026-10-16 23:46:19+08:00 [CATEGORIES]cs.CL
PresenceBench: An Exploratory Evaluation of Large Language Model Adaptation to User State [AUTHORS]Peña Arroyo Gallego, Mauro Ruiz Moreno, Alejandro Acien, Sergio Romero-Tapiador, Teresa Arroyo-Gallego [ABSTRACT]Presence-aware intelligence is the capacity of an AI system's intelligence layer to adapt appropriately to the functional state of the person it serves (here readiness, flow, and stability) while preserving task quality and safety. We introduce PresenceBench, an instrument for measuring it. PresenceBench evaluates how the intelligence layer uses a supplied state, not how that state is sensed, and breaks the capability into measurable properties. Two are tested here. Receptivity: does the model take up the state and give a better response than with no signal? Specificity: does any improvement depend on the correct state, rather than on extra structured context? A third, adaptation across states, is a required endpoint not measured in this release. A five-arm causal design answers both questions, against each model's own baseline and an equally structured mismatched-state control, scored by a correctness-gated human paired comparison. The exploratory evaluation covered four models across eighteen single-turn, six fixed three-turn and six simulated natural-conversation scenarios. Supplying a synthetic state, raw or translated into a response policy, gave human-preference point estimates above 0.5 for all four models (Claude Sonnet 5, DeepSeek V4 Pro, Gemini 3.6 Flash, GPT-5.6 Terra), with three of eight 95% intervals excluding chance. Specificity was not resolved: no per-model interval excluded chance, and the design cannot yet separate the state values from a generic instruction to adapt. Policy delivery reduced priced cost and output tokens against each model's baseline; these are resource-use changes, not efficiency gains, because correctness non-inferiority was not established. Safety-floor failures were rare (13 of 360 cases), with a non-significant excess under policy delivery. We report the instrument, the frozen results, and the controls a confirmatory study needs. [COMMENTS]33 pages, 6 figures, 13 tables [LINK]http://arxiv.org/abs/2610.23625v1 [DATE]2026-10-16 23:33:04+08:00 [CATEGORIES]cs.CL
The Chameleon's Limit: Investigating Persona Collapse and Homogenization in Large Language Models [AUTHORS]Yunze Xiao, Vivienne J. Zhang, Chenghao Yang, Weihao Xuan, Jen-tse Huang [ABSTRACT]Large language models (LLMs) have powered multi-agent simulations such as survey respondents or application users. However, LLMs lack the population-level diversity required by these simulations---responses from LLMs with detailed human profiles cluster into classes, unlike human responses that uniformly fill the entire space. In this paper, we formalize this as Persona Collapse and quantify its population-level structure with three diagnostic metrics: Coverage (how much of the persona space is occupied), Uniformity (how evenly agents spread across it), and Complexity (how rich the resulting behaviors are). Across eight LLMs on personality (BFI-44), moral reasoning, and self-introduction, persona collapse is universal but task-dependent: CoSER-Llama-8B collapses the most in personality while being the most diverse in moral reasoning. Moreover, sampling more personas cannot mitigate the collapse. We further show a fidelity--diversity tradeoff: the models with the highest per-persona fidelity produce the most polarized populations. Finally, persona-targeted fine-tuning further narrows the population instead of diversifying it. [LINK]http://arxiv.org/abs/2604.24698v2 [DATE]2026-10-16 23:30:51+08:00 [CATEGORIES]cs.CL
Trajectory-Triggered Sparse Collaborative Decoding [AUTHORS]Amit Ranjan Trivedi, Sruthikesh Surineni, Kyong Ho Lee, Mohit Mittal, Steven Li, Rex Hill [ABSTRACT]Dense assistance methods such as speculative decoding repeatedly verify draft tokens with a target model, keeping target computation coupled to generation. We study sparse collaborative decoding, where a draft model generates most tokens and a target model is invoked only for short corrective segments. The key problem is deciding when to call the target. We show that local uncertainty is a weak trigger: entropy, negative log-likelihood, logit margin, top-$k$ mass, and velocity features give near-chance failure prediction, with AUC 0.50-0.53. In contrast, aggregate trajectory-shape features over the draft confidence path reach AUC 0.86 using a lightweight logistic predictor trained only from draft correctness labels. This induces a trigger-timing tradeoff: early calls have more leverage but weaker evidence, while later calls have stronger evidence but less ability to redirect the trajectory. Used online, learned sparse collaborative decoding improves over the best open-loop schedule by +5.3 to +20.0 accuracy points across five GSM8K draft-target pairs. On MATH-500, it gains 7.0-13.5 points when early prediction and target competence align, but loses when either is weak. On StrategyQA, it improves over open-loop by 5.0-8.3 points, showing transfer beyond arithmetics. [COMMENTS]17 pages. Accepted at the EMNLP 2026 uncertainty workshop [LINK]http://arxiv.org/abs/2610.23613v1 [DATE]2026-10-16 23:24:05+08:00 [CATEGORIES]cs.CLcs.LG
SIT-Eval: Situation-Based Interactive Turn-by-Turn Evaluation of Conversational Agents [AUTHORS]Ravil Mussabayev, Saveliy Zavertyaev, Dmitriy Parpulov, Andrey Savchenko [ABSTRACT]The rapid proliferation of conversational AI agents necessitates evaluation methods that assess performance across diverse dialogue scenarios, utilizing multiple situational criteria, as appropriate behavior depends on context and encompasses various aspects. Existing dialogue benchmarks typically use static natural conversations, which underrepresent rare situations---a critical safety gap, as low-probability events become frequent failures at production scale. However, limited sample sizes force a choice: either maintain natural dialogues for realism or turn to adaptive situation sampling for comprehensive yet statistically reliable coverage. We introduce SIT-Eval, an adjustable framework for evaluating situation-based conversational agents. We formalize an agent's desired behavior as situational rules: if situation is P, follow rules Q. At each conversation turn, an LLM-interlocutor enacts a uniformly sampled situation. A context-aware LLM-judge leverages all conditionally applicable rules as criteria to evaluate the agent's reply for that situation. The judge is validated using human-verified synthetic golden dialogues. Benchmarking 8 LLMs on 1292 conversation turns reveals failure modes hidden by aggregate metrics: models with similar overall scores vary widely in their adherence to individual rules. All models struggle most with ambiguity clarification and Socratic questioning. [LINK]http://arxiv.org/abs/2610.23612v1 [DATE]2026-10-16 23:24:02+08:00 [CATEGORIES]cs.CL
Not Every Mention Is a Claim: Rethinking Object Hallucination Detection in Vision-Language Models [AUTHORS]Hsin-Ling Hsu, Neeraja Kirtane, Kuan-Hao Huang [ABSTRACT]Object hallucination in vision-language models (VLMs) is usually measured by checking each object a caption names against the image. Most benchmarks and detectors alike treat every object mention as a claim that the object is present. This assumption fails when a caption names an object without asserting it, as in "a bench for people to sit on", which mentions people without claiming that any are in the image. We call such mentions non-assertive. Treating them as hallucinations inflates the error rates attributed to models that use such language. To separate them from real hallucinations, we introduce NAME-Bench (Non-Assertive Mention EvaluationBenchmark), 25,497 object mentions from captions that five VLMs generate for COCO images. Each mention is labelled concrete (asserted and present), hallucinated (asserted and absent), or non-assertive, and non-assertive mentions are further divided into four subtypes: depictive, purpose, negation and hedged. All the hallucination detectors that we evaluate flag non-assertive mentions as hallucinations, at rates between 14% and 80%. Adding non-assertive mentions to detector training as negatives largely corrects this on LLaVA-1.5. On the four newer VLMs, a comparable correction takes five to ten times more data, and the detectors then catch fewer real hallucinations. These findings show that current detectors conflate mentioning an object with asserting it, and that this conflation is harder to undo on the four newer VLMs than on LLaVA-1.5. Hallucination evaluation should therefore separate what a caption asserts from what it merely mentions. [LINK]http://arxiv.org/abs/2610.23610v1 [DATE]2026-10-16 23:23:45+08:00 [CATEGORIES]cs.CL
Enhancing RAG-based Medical VQA with Gated Multimodal Representations [AUTHORS]Daniela Jimenez-Pano, Hugo Jair Escalante, Delia Irazú Hernández-Farías [ABSTRACT]Medical Vision-Language Models (Med-LVLMs) have shown strong performance in Medical Visual Question Answering (MVQA), yet they may still produce factually inconsistent responses. Retrieval-Augmented Generation (RAG) can mitigate this issue by incorporating external clinical evidence; however, retrieval is often guided by a single modality, potentially overlooking the joint semantics of medical images and clinical questions. We introduce a multimodal RAG framework that learns a joint query representation through a Gated Multimodal Unit, enabling evidence retrieval to leverage information from both modalities while keeping the underlying Med-LVLM frozen. We evaluate the proposed approach on benchmark datasets. Results show that the multimodal query representation consistently outperforms a reference Med-LVLM and remains competitive with state-of-the-art methods. Our findings demonstrate the potential of multimodal query representations for more effective evidence retrieval in Medical VQA. [COMMENTS]Preprint submitted to ECIR [LINK]http://arxiv.org/abs/2610.23588v1 [DATE]2026-10-16 23:07:33+08:00 [CATEGORIES]cs.CL
Analyzing Occupational Gender Bias across the LLM Training Lifecycle [AUTHORS]Manthan Gadhia, Alexander Hoyle, Karolina Stańczak [ABSTRACT]Post-training is widely assumed to reduce the social biases language models inherit from pretraining, yet direct evidence for this assumption remains mixed. In this paper, we examine how gender disparities evolve across the language modeling pipeline. In particular, we study occupational gender bias in OLMo 3, a fully open language model family, across its post-training stages by prompting each checkpoint to generate open-ended narratives about synthetic personas given their occupation and education. We consider two settings, a gender-given setting in which the profile states the persona's gender explicitly and a gender-assumed setting in which it must be inferred by the model. We find that when gender is stated, post-training increases semantic diversity for male-coded profiles, whereas for female-coded profiles the increase is smaller after SFT and not significant after DPO and RLVR. When gender is assumed by the model, post-training affects narratives about female and male personas similarly. The model's own gender assumptions, however, show a consistent female-preferential skew relative to United States labor-force demographics, except for heavily male-dominated occupations. Every successive post-training stage monotonically increases the odds of attributing a female label over the base model. A provenance analysis of the supervised fine-tuning corpus finds only a modest 7.7% female-token surplus, too small to explain the size of these downstream effects. The project code is available under: https://github.com/manthangadhia/llm-lifecycle-bias. [COMMENTS]Submitted to 7th Workshop on Gender Bias in Natural Language Processing @AACL-IJCNLP 2026 [LINK]http://arxiv.org/abs/2610.23587v1 [DATE]2026-10-16 23:07:02+08:00 [CATEGORIES]cs.CL
MARS: Memory-Aware Routing for Session-Level LLM Assistants [AUTHORS]Minghan Wang, Jinhang Zuo, Zhongxiang Dai, Zhiyong Wang, Fang Kong [ABSTRACT]Long-term LLM assistants must repeatedly decide what prior context to expose and which model should answer as an interaction evolves. Existing memory systems and model routers usually make these decisions separately, although memory construction changes the evidence, distractors, and difficulty of the resulting input. We introduce MARS, a memory-aware routing framework that uses session state as a shared control interface between memory construction and model selection. A Query and Topic Manager (QTM) estimates compact decision-relevant state that guides a fixed state-conditioned memory policy and an online router, enabling each turn to receive an appropriate memory view and a suitable model under a quality-memory-cost trade-off. We evaluate MARS on RealMem and MultiChallenge with 16 system configurations each, covering long-history memory-driven interaction and bounded sessions with structured context interference. On RealMem, MARS preserves 98.2% of full-context GPT-4o's QA while using only 14.5% of its generation API cost, including both generation and controller calls for MARS, and achieves the strongest measured memory-recall and retrieval scores among compared systems. On MultiChallenge, MARS trails the strongest static system overall but approaches its performance at lower recorded cost on a mechanism-focused subset. Together, these results support a favorable quality-memory-cost trade-off for long-history, memory-dependent interaction. [COMMENTS]32 pages, including references and appendices [LINK]http://arxiv.org/abs/2610.23583v1 [DATE]2026-10-16 23:04:22+08:00 [CATEGORIES]cs.CL
PragReST: Self-Reinforcing Counterfactual Reasoning for Pragmatic Language Understanding [AUTHORS]Jihyung Park, Minchao Huang, Leqi Liu, Elias Stengel-Eskin [ABSTRACT]Natural language understanding often depends on meanings that are implied rather than explicitly stated, requiring pragmatic reasoning. Despite strong performance on math and logical reasoning, large language models (LLMs) still struggle with making pragmatic inferences, often choosing literal interpretations. To improve LLM pragmatic reasoning, we introduce PragReST, a self-supervised framework that constructs pragmatic QA data, generates counterfactual reasoning traces, and trains models to internalize them through supervised fine-tuning and reinforcement learning, without human-labeled training data or distillation from a stronger teacher. Across four pragmatic benchmarks (PragMega, Ludwig, MetoQA, and AltPrag), PragReST improves over backbone models, task-specific pragmatic tuning baselines, and non-counterfactual variants of the same pipeline. On accuracy-based benchmarks, PragReST improves over the instruct backbone by 5.37 and 5.50% (absolute) for Qwen3-8B and Qwen3-14B, respectively. Our error analysis and ablations underscore the importance of counterfactual reasoning: PragReST primarily reduces errors caused by failures to contrast observed utterances with plausible alternatives, and removing counterfactual reasoning substantially reduces performance. Moreover, our training preserves out-of-domain performance on general-knowledge and mathematical reasoning benchmarks. [COMMENTS]EMNLP Camera-Ready. First two authors contributed equally. Code and models: https://github.com/jihyung803/PragReST [LINK]http://arxiv.org/abs/2606.18624v2 [DATE]2026-10-16 22:54:52+08:00 [CATEGORIES]cs.CL
Towards Reliable Truth-Aligned Uncertainty Estimation in Large Language Models [AUTHORS]Ponhvoan Srey, Quang Minh Nguyen, Xiaobao Wu, Anh Tuan Luu [ABSTRACT]Uncertainty estimation (UE) aims to quantify uncertainty in large language model (LLM) outputs to flag potential hallucinations and support reliable deployment. However, UE metrics often exhibit unstable performance across models and datasets, limiting their practical utility. We formalise this phenomenon as proxy failure, arising because UE metrics are derived from model behaviour rather than being explicitly grounded in factual correctness. We show that when mutual information between a UE score and correctness is low, the score's discriminability must be close to chance. To address this, we propose Truth AnChoring (TAC), a lightweight post-hoc calibration method that maps raw UE scores to truth-aligned correctness probabilities. Even with noisy and few-shot supervision, TAC continues to improve calibration, providing a practical calibration protocol. Our findings highlight the limitations of treating heuristic UE metrics as direct indicators of uncertainty about factual correctness, and position our TAC as a principled step toward more reliable uncertainty estimation for LLMs. The code repository is available at https://github.com/ponhvoan/TruthAnchor/. [COMMENTS]Accepted to EMNLP Findings 2026 [LINK]http://arxiv.org/abs/2604.00445v2 [DATE]2026-10-16 22:40:44+08:00 [CATEGORIES]cs.CL
Execution Before Memory: Diagnosing Memory Failures in Video World Models [AUTHORS]Weiyang Jin, Jiwen Yu, Sihan Xu, Yuwei Niu, Xihui Liu [ABSTRACT]A memory test for a video world model should say whether the scene was set, whether the asked change was carried out, and whether the last frame shows the result. Current tests fold these into one number, so a low score is easy to read as forgetting. When the change is not carried out, the hidden state transition cannot appear, so the test is not really a test of memory. We introduce MIST, an evaluation that separates the change from the memory it should leave behind. The model first sets the scene while the change stays hidden, and only later is asked to show it, with leakage before that point kept under control. This separation locates the bottleneck: the model sets the scene but does not carry out the instructed change. The same gap appears in general memory evaluations. On that basis we propose M3-Agent (Make the Missing Move), which recursively revises the details of the input instruction so that the model successfully carries out the requested instruction. Relative to the baseline, it more than triples the MIST reveal score. On the benchmarks, WRBench's score T rises by 26.1% and MBench-T's score U by 52.7%. The method meets the problem from the execution side: it shows what the frozen model can do once the move is stated, and it sets a stricter baseline. [COMMENTS]27 pages, 7 figures. Project page: https://hku-mmlab.github.io/MIST/ Code: https://github.com/HKU-MMLab/MIST [LINK]http://arxiv.org/abs/2610.23542v1 [DATE]2026-10-16 22:37:13+08:00 [CATEGORIES]cs.CL
Referring Multi-Object Tracking with Diverse Dynamic Expressions [AUTHORS]Yani Zhang, Dongming Wu, Wencheng Han, Xingping Dong, Shengcai Liao, Bo Du [ABSTRACT]Referring understanding is a fundamental task that bridges natural language and visual content by localizing objects described in free-form expressions. However, existing works are constrained by limited language expressiveness, lacking the capacity to model object dynamics in spatial numbers and temporal states. To address these limitations, we introduce a new and general referring understanding task, termed referring multi-object tracking (RMOT). Its core idea is to employ a language expression as a semantic cue to guide the prediction of multi-object tracking, comprehensively accounting for variations in object quantity and temporal semantics. Along with RMOT, we introduce a RMOT benchmark named Refer-KITTI-V2, featuring scalable and diverse language expressions. To efficiently generate high-quality annotations covering object dynamics with minimal manual effort, we propose a semi-automatic labeling pipeline that formulates a total of 9,758 language prompts. In addition, we propose TempRMOT, an elegant end-to-end Transformer-based framework for RMOT. At its core is a query-driven Temporal Enhancement Module that represents each object as a Transformer query, enabling long-term spatial-temporal interactions with other objects and past frames to efficiently refine these queries. TempRMOT achieves state-of-the-art performance on both Refer-KITTI and Refer-KITTI-V2, demonstrating the effectiveness of our approach. The source code and dataset is available at https://github.com/zyn213/TempRMOT. [LINK]http://arxiv.org/abs/2406.05039v3 [DATE]2026-10-16 22:34:19+08:00 [CATEGORIES]cs.CL
DominoTree: Conditional Tree-Structured Drafting with Domino for Speculative Decoding [AUTHORS]Saw S. Lin, Jyh-Shing Roger Jang [ABSTRACT]Speculative decoding speeds up large language model inference: a small drafter proposes several tokens, and the target model verifies them in one parallel pass. Drafting a tree lets the target verify many alternative continuations at once. DDTree builds such trees from the block-diffusion drafter DFlash, but scores each position independently of the tokens earlier in its branch; Domino conditions on those tokens but drafts only a single chain. DominoTree combines the two: a training-free best-first tree scored by Domino's path-dependent causal correction, restricted to top-M candidates. In a research harness at temperature zero, it averages 5.33x speedup over autoregressive (AR) decoding on Qwen3-4B and 6.89x on Qwen3-8B, peaking at 8.77x; across temperatures it is 20-26% faster than Domino on Qwen3-4B and 27-36% on Qwen3-8B, and 11-20% faster than the stronger of DDTree and its cost-aware variant CaDDTree. A controlled decomposition attributes +10.1% (Qwen3-4B) and +10.7% (Qwen3-8B) accepted length to the correction, and +4.7% and +4.4% to conditioning on the realized path. Served in SGLang on one RTX 5090, it reaches 4.11x and 4.83x AR at batch size 1, 1.17x and 1.09x the speed of Domino's chain, and up to 1.71x AR on 8K-token HELMET summarization, 14-21% above the chain. The lead holds at 2 concurrent requests but reverses from 8 onward, where the chain sustains more goodput despite DominoTree's 28% higher acceptance. DominoTree accepts more tokens than the chain in every serving setting. On the hybrid Gated-DeltaNet Qwen3.6-27B, correct output requires tree-aware recurrent verification; with it, DominoTree matches the chain at batch size 1 (4.58x versus 4.54x AR) and reaches up to 2.11x AR at long context, 15% above the chain. [COMMENTS]31 pages. Submitted to IEEE Access. v4: adds Qwen3.6-27B results, single-GPU serving re-runs, fused tree-builder kernels and tuned budgets; code and data at https://github.com/slin-zhq/Domino-Tree [LINK]http://arxiv.org/abs/2607.08642v4 [DATE]2026-10-16 22:29:03+08:00 [CATEGORIES]cs.CL
EduWild: How Teacher-Configured Chatbots Fail in Real-World Use and How to Fix Them [AUTHORS]Ziqi Liu, Ruiwei Xiao, Yuxuan Li, John Stamper [ABSTRACT]Teachers increasingly configure educational chatbots through natural-language instructions, but how these bots fail in real-world use remains poorly understood. We introduce EduWild, a dataset of 1,726 conversations with 217 teacher-configured chatbots, each paired with the instructions and configuration behind it, including 900 conversations from the intended end users of 35 deployed chatbots. We develop a 29-code taxonomy covering factual, workflow, diagnostic, and pedagogical failures and apply it using a human-calibrated LLM judge. The most frequent failures are lapses in following teacher intent --- bots skip required workflow steps (53.8% of end-user conversations), omit expectation setup (33.4%), and overload learners with information or questions (37.1%). Guided by these findings, we design a prompt-based harness targeting the six most frequent failure modes. In a study with 200 college students across ten bots, the harness reduces the mean number of target failure codes per conversation from 1.406 to 0.489, a 65.2% relative reduction with bots weighted equally. These findings show how real-world failure analysis can guide practical improvements to educational chatbots. [LINK]http://arxiv.org/abs/2610.23530v1 [DATE]2026-10-16 22:28:06+08:00 [CATEGORIES]cs.CL
Truth in Contrast: Calibrated Factual Evidence from Language Model Likelihoods [AUTHORS]Ponhvoan Srey, Jason Lucas, Pittawat Taveekitworachai, Junyoung Lee, Anh Tuan Luu [ABSTRACT]Large Language Models (LLMs) encode substantial factual knowledge. Yet, estimating confidence in factual claims reliably remains challenging. We study likelihood-based factual confidence from the complementary perspectives of discrimination and calibration across multilingual factual statements. Specifically, we instantiate contrastive generative likelihoods as Contrastive Factual Evidence (CFE), which compares the likelihood of the same claim under competing factual and non-factual conditions, and derive its probabilistic counterpart, the Contrastive Factual Posterior (CFP). We show analytically that CFP calibration depends on the differential error between the two surrogate conditionals, allowing errors shared across them to cancel and yield tighter calibration bounds. Empirically, CFP substantially improves calibration while maintaining factual discrimination comparable to its strongest competitor. Nonetheless, discrimination is uniformly weak across methods in lower-resource languages, which simultaneously undermines the utility of the measured calibration. To this end, we interpret a likelihood geometry, within which an evidence-bearing direction is induced, along with an orthogonal evidence-invariant direction. In our investigation, we selectively align target-language claims with stronger English evidence only along the factual-evidence direction, improving discrimination and open-ended factual generation without target-language labels. [LINK]http://arxiv.org/abs/2610.23527v1 [DATE]2026-10-16 22:27:04+08:00 [CATEGORIES]cs.CL
Perceive, Solve, Reason: Progressive Capability Formation in Time Series-LLMs [AUTHORS]Xin Qiu, Junlong Tong, Wei Zhang, Xiaoyu Shen [ABSTRACT]Large language models (LLMs) can now perceive, describe, and reason about images and audio, yet time series understanding remains underestimated: most work uses LLMs as forecasters and classifiers, reasoning studies are few and narrow, and no suite trains and evaluates the full range of capabilities. In this paper, we build the first such suite, with 337K training and 44K evaluation samples across three progressive capabilities, Perceive, Solve, and Reason, and use it to ask how time series should enter an LLM that must master all three, and in what order. We find that a dedicated encoder is essential: with one, a 4B model surpasses GPT-4o on every perception attribute we test, whereas textualizing the series costs 8--10 points in reasoning accuracy. The capabilities then build on one another: learning them in order beats mixing the same data and rivals reasoning-only training with over three times the reasoning data, but only when an encoder is used. Analyses further reveal that capabilities are transferred through distinct network depths rather than simply stacked, and that a stage converged on its own benchmark is not necessarily ready for the next. Distilled into a simple recipe, these findings yield 4B TS-LLMs that, without any target-domain data, outperform four recent 7B TS-LLMs on seven of eight out-of-domain metrics, by up to 11 points. We therefore advocate treating time series as a first-class modality for LLMs, read through dedicated encoders and taught to perceive before they reason. [LINK]http://arxiv.org/abs/2610.23521v1 [DATE]2026-10-16 22:23:23+08:00 [CATEGORIES]cs.CL
MuninAI: A Self-Hosted Agentic Framework for Independent Research Groups [AUTHORS]Markus Fischer, Jens Meiler, Matthias Elgeti [ABSTRACT]Research groups that want a language model grounded in their own literature face a choice between hosted services that receive their unpublished work and self-hosted deployments that keep the data local but give the model no access to the corpus. Munin is an open-source framework for a self-hosted scientific assistant that indexes a group's corpus and its citation graph, and wraps whichever open-weight model the group can serve in an agent harness with specialised reading, search, and computation tools. We claim that the harness, rather than the specific model, is what makes the system useful on domain literature, and we ship the evaluation suite that tests this claim with the software.
On 199 paired literature questions, the harness more than doubles the accuracy of the base model, from 0.387 to 0.874 (+0.487, 95% CI [0.407, 0.568]), while naive retrieve-then-answer scores below the unaided model on every backbone tested. Under the harness, roughly twice as many answer claims trace to a passage the model read itself as with fixed top-5 retrieval. When the source paper for a question is removed from the corpus, the system abstains on about nine in ten of the questions it had answered correctly on the Qwen backbones and eight in ten on gpt-oss-20b, and it produces no fabricated local citation across 300 fabricated references.
The pattern holds on three backbones from two developers, including a 20-billion-parameter model that runs on a single consumer-grade GPU. With all outbound network access disabled, both Qwen backbones still exceed the published PaperQA2 accuracy. The evaluation suite runs as a single command, so an adopting group can re-run it on its own corpus and model. [COMMENTS]37 pages, 3 figures, 20 tables. Code, evaluation suite and scorecards: https://github.com/elgetilab/munin (archived release v1.0.0: https://doi.org/10.5281/zenodo.23191404) [LINK]http://arxiv.org/abs/2610.23480v1 [DATE]2026-10-16 21:56:37+08:00 [CATEGORIES]cs.CL
Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching [AUTHORS]Roy Miles, Aysim Toker, Andreea-Maria Oncescu, Jiankang Deng, Ismail Elezi [ABSTRACT]Reasoning with large language models often benefits from generating multiple chains-of-thought, but existing aggregation strategies are typically trajectory-level (e.g., selecting the best trace or voting on the final answer), discarding useful intermediate work from partial or "nearly correct" attempts. We propose Stitching Noisy Diffusion Thoughts, a self-consistency framework that turns cheap diffusion-sampled reasoning into a reusable pool of step-level candidates. Given a problem, we (i) sample many diverse, low-cost reasoning trajectories using a masked diffusion language model, (ii) score every intermediate step with an off-the-shelf process reward model (PRM), and (iii) stitch these highest-quality steps across trajectories into a composite rationale. This rationale is then used to recompute only the final answer. This modular pipeline separates exploration (diffusion) from evaluation and solution synthesis, avoiding monolithic unified hybrids while preserving broad search. Across math reasoning benchmarks, we find that step-level recombination is most beneficial on harder problems, and ablations highlight the importance of the final solver in converting stitched but imperfect rationales into accurate answers. Using low-confidence diffusion sampling with parallel, independent rollouts, our training-free framework improves average accuracy by up to 23.8% across six math and coding tasks. At the same time, it achieves up to a 1.8x latency reduction relative to both traditional diffusion models (e.g., Dream, LLaDA) and unified architectures (e.g., TiDAR). The code is available https://github.com/roymiles/diffusion-stitching [COMMENTS]NeurIPS 2026. Code available at https://github.com/roymiles/diffusion-stitching [LINK]http://arxiv.org/abs/2602.22871v3 [DATE]2026-10-16 21:55:50+08:00 [CATEGORIES]cs.CL
A low-dimensional dynamical basis for tongue shape control [AUTHORS]Sam Kirkham, Patrycja Strycharczuk [ABSTRACT]Speech production involves the control of a high-dimensional articulatory system with many degrees-of-freedom, but a widespread hypothesis is that speakers navigate this complexity using a set of lower-dimensional control variables. Common candidates for these control variables include individual articulators (e.g. tongue body) and abstract tract variables (e.g. tongue body constriction location), but identifying the most appropriate coordinate system for articulatory control remains an open question. In this study, we advance a model in which holistic tongue shape modes represent a potential low-dimensional control basis for a simple dynamical law. We first discover empirical tongue shape modes in a cross-speaker coordinate space and then fit the model to ultrasound data on monophthongal vowels from 36 speakers of British English. A three-dimensional linear point attractor fitted to these time-varying modes recovers the dynamics of tongue shape with a high degree of accuracy. This suggests that complex tongue movement can arise from independent modes that linearly combine under simple dynamical laws, thereby relocating some complexity from the control signal to the coordinate system. We show that this framing explains why one-dimensional articulatory measures can exhibit observable nonlinearities, and hypothesize that shape modes are a plausible kinematic expression of underlying muscle synergies. [COMMENTS]Pre-print currently under review [LINK]http://arxiv.org/abs/2610.23456v1 [DATE]2026-10-16 21:39:24+08:00 [CATEGORIES]cs.CL
HydraLLM: Adaptive Width Slicing for Elastic Text and Vision-Language Models [AUTHORS]Oscar Kemme, Janek Haberer, Ali Hojjat, Olaf Landsiedel [ABSTRACT]LLMs and LVLMs are increasingly deployed across heterogeneous hardware, ranging from data-center GPUs to resource-constrained edge devices. Their fixed-size architectures limit the ability to reuse a single model across deployment settings with different compute, latency, and model-size targets. We introduce HydraLLM, an elastic MQA-based architecture. Its number of active attention heads determines the effective widths of the embeddings, attention projections, feed-forward layers, and the output head, forming a nested family of subnetworks within a single parameter set. We introduce a static-to-elastic conversion pipeline that, starting from a pre-trained model checkpoint, includes importance-driven weight reordering and stochastic subnetwork training with knowledge distillation. We extend this principle to multimodal inference through an elastic projection module whose output width follows the active text width. HydraLLM covers a wider range of active parameter configurations than FFN-only elastic architectures. With 75% of heads active, HydraLLM achieves 2.2 percentage points higher TriviaQA accuracy than a state-of-the-art nested baseline configuration which has 17.25% more parameters. With 50% of heads active, our multimodal model achieves a CLIPScore comparable to that of the 7$\times$ larger Gemma 3 4B model. With only 25% of heads active, HydraLLM increases token-generation throughput by 75.3% over the full-width configuration. [COMMENTS]Accepted at the BabyLM Workshop at EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23455v1 [DATE]2026-10-16 21:38:11+08:00 [CATEGORIES]cs.CLcs.LG
"You" Is Not Toxic: How LLM Labelers Rewrite the Web's Conversation Laws [AUTHORS]Liang Yue, Eric A. Vance [ABSTRACT]What the Web community believes about toxic conversations, such as that toxicity rises in longer and deeper threads, was measured mostly with Google's Perspective API, which is shutting down; large language models (LLMs) will make the next measurements. We find that LLM labelers mistake second-person address for toxicity, and that this one bias manufactures laws that human labels do not show. On 1.25 million news comments with crowd labels, Claude Haiku flags 12% of the non-toxic comments that contain no "you" and 51% of those with four or more; a single crowd rater flags 6% and 10%. Deeper replies address someone more often, so under Haiku toxicity rises with reply depth (+0.93 percentage points per level), while under the crowd's labels it slightly declines (-0.12). Rewriting "you" in the third person removes two thirds of the false rise; flagging only as many comments as the crowd does not remove it, and showing Haiku the parent comment enlarges it. All seven LLMs we test over-flag "you", and at the crowd's flag rate, without the parent, any false rise of the open ones vanishes at a fixed number of second-person words; Haiku also over-flags "you" on Reddit, Gab and Wikipedia talk pages. Agreement with human labels does not reveal the bias: we prove that a fixed Cohen's kappa leaves a law nearly unconstrained. We therefore certify the law instead of the labeler. Threadsafe uses the labeler's scores to choose which comments human raters label and to predict the others, and corrects the predictions with the human labels; linear laws are estimated without bias whatever the labeler does. In simulated audits its intervals for the laws keep nominal coverage from 250 labels on, where uniform-sample methods such as PPI++ fall short, and at 1,000 to 8,000 labels are 0.75 times as wide as those of design-based supervised learning (DSL); a thousand human labels rule out Haiku's rising law. [COMMENTS]43 pages, 4 figures, 44 tables. Labels, prompts and code: https://github.com/liyu2195/you-is-not-toxic [LINK]http://arxiv.org/abs/2610.23446v1 [DATE]2026-10-16 21:29:30+08:00 [CATEGORIES]cs.CL
Atomic Consistency Preference Optimization for Long-Form Question Answering [AUTHORS]Jingfeng Chen, Raghuveer Thirukovalluru, Junlin Wang, Kaiwei Luo, Bhuwan Dhingra [ABSTRACT]Large Language Models (LLMs) often produce factoid hallucinations - plausible yet incorrect answers. A common mitigation strategy is model alignment, which improves factual accuracy by training on curated (factual, non-factual) pairs. However, this approach often relies on a stronger model (e.g., GPT-4) or an external knowledge base to assess factual correctness that may not always be accessible. Addressing this, we propose Atomic Consistency Preference Optimization (ACPO), a self-supervised preference-tuning method that enhances factual accuracy without external supervision. ACPO leverages atomic consistency signals (i.e., the agreement of individual facts across multiple stochastic responses) to identify high- and low-quality data pairs for model alignment. Despite being fully self-supervised, ACPO outperforms the strong supervised alignment baseline by 1.95 points averaged across Phi-3 and Llama3 on the LongFact and BioGen datasets, demonstrating its effectiveness in improving factual reliability without relying on external models or knowledge bases. [COMMENTS]13 pages, 1 figure [LINK]http://arxiv.org/abs/2505.09039v3 [DATE]2026-10-16 21:28:17+08:00 [CATEGORIES]cs.CL
ScoutRoute: Efficient Mixture-of-Experts Routing Using Internal Language Model Signals [AUTHORS]Rajdeep Pandhare, Delzaan Bharucha [ABSTRACT]Efficient expert routing is a fundamental challenge in Mixture-of-Experts (MoE) systems, particularly for heterogeneous collections of independently trained Large Language Mod els (LLMs). Existing routing approaches typically rely on jointly trained gating networks, prompt-based predictors, or computationally expensive multi-model inference, limiting their flexibility and efficiency. This paper presents ScoutRoute, a lightweight post-hoc routing framework that predicts expert competence using internal language model signals extracted from a single forward pass through an inexpensive scout model. Hidden-state representations are used to estimate cross-model competence through lightweight logistic probes, while logit-derived confidence measures are incorporated within a domain-aware routing strategy for efficient expert selec tion without modifying the underlying language models. ScoutRoute isevaluatedon fiveopen-weightlanguagemodels across fourbenchmark datasets covering mathematical reasoning, code generation, commonsense reasoning, and multidis ciplinary knowledge. Experimental results show that ScoutRoute consistently outperforms individual routing signals while requiring only a single scout-model inference. The proposed competence probes also achieve superior calibration compared with conventional confidence measures, demonstrating that internal language model representations provide transferable information for accurate and efficient expert routing across heterogeneous language models [COMMENTS]10 pages, 1 figure, 6 tables. Published in Advances in Artificial Intelligence and Machine Learning, 2026 [LINK]http://arxiv.org/abs/2610.23443v1 [DATE]2026-10-16 21:25:53+08:00 [CATEGORIES]cs.CL
Cross-Layer Attention Probing for Fine-Grained Hallucination Detection [AUTHORS]Malavika Suresh, Rahaf Aljundi, Ikechukwu Nkisi-Orji, Nirmalie Wiratunga [ABSTRACT]With the large-scale adoption of Large Language Models (LLMs) in various applications, there is a growing reliability concern due to their tendency to generate inaccurate text, i.e. hallucinations. In this work, we propose Cross-Layer Attention Probing (CLAP), a novel activation probing technique for hallucination detection, which processes the LLM activations across the entire residual stream as a joint sequence. Our empirical evaluations using five LLMs and three tasks show that CLAP improves hallucination detection compared to baselines on both greedy decoded responses as well as responses sampled at higher temperatures, thus enabling fine-grained detection, i.e. the ability to disambiguate hallucinations and non-hallucinations among different sampled responses to a given prompt. This allows us to propose a detect-then-mitigate strategy using CLAP to reduce hallucinations and improve LLM reliability compared to direct mitigation approaches. Finally, we show that CLAP maintains high reliability even when applied out-of-distribution. Code available at: https://github.com/itsmemala/CLAP [COMMENTS]Published at the TRUST-AI workshop, ECAI 2025 [LINK]http://arxiv.org/abs/2509.09700v2 [DATE]2026-10-16 21:07:39+08:00 [CATEGORIES]cs.CL
On the Tip of the Tongue: First-Token Cueing as a Behavioral Probe of Recall Failures in LLMs [AUTHORS]Jaewon Sok, Jewon Yeom, Seonghyeon Park, Jeongjae Park, Hwiyeong Lee, Taesup Kim [ABSTRACT]Large language models often answer factual questions incorrectly in closed-book settings, and such errors are usually attributed to missing knowledge. Yet models often encode answers they fail to produce. We show that many closed-book errors are first-token recall failures: the model commits to a wrong first token, but once given the right one it completes the answer and often recognizes it as correct. We study this with First-Token Cueing, a behavioral probe that prefills the response with the first token of the gold answer. Model uncertainty peaks at the first answer token, where it already predicts correctness, and among gold prefixes of any length the first token alone accounts for the largest share of recovered errors. Across 16 open models, this one token corrects between a fifth and a half of the errors on multi-token answers in English QA and improves accuracy in all seven languages of a multilingual benchmark. Most of the errors it corrects are also solved by the model's own sampled reasoning, a criterion for encoded knowledge, and models recognize correct starts they do not produce: they continue a correct start more confidently than an incorrect start of the same answer type and prior probability, yet rarely find it without the gold answer. These results locate a recoverable class of factual errors at the first decoding step, where inference-time methods may gain the most. [LINK]http://arxiv.org/abs/2610.23400v1 [DATE]2026-10-16 20:55:36+08:00 [CATEGORIES]cs.CL
FSPO: Policy-Consistent Risk and Pareto-Feasible Control for Budgeted LLM RL Post-Training [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Daren Zha [ABSTRACT]Adaptive LLM reinforcement-learning post-training changes multiple training actuators online, including rollout temperature, group size, clipping, KL regularization, verifier allocation, and update budget. Three coupled issues remain unresolved. A future-risk model trained from behavior trajectories need not estimate the risk induced by the controller that will be deployed; a score calibrated on logged state-action pairs can become miscalibrated after selective action choice; and independent per-resource minimum costs do not in general certify a feasible multi-resource continuation. We introduce FSPO, a feedback-state controller for budgeted LLM RL post-training that addresses these issues jointly. FSPO learns a policy-consistent risk-to-go model whose Bellman target follows the same frozen controller used for future decisions, together with a long-horizon utility model. Decision-conditioned trajectory calibration (DCTC) calibrates risk on cross-fitted trajectories generated by actions selected by provisional controllers. A Pareto resource continuation certificate (PRCC) admits an action only when a non-dominated cumulative reservation remains feasible over the residual horizon. Under a matched GRPO resource envelope, FSPO reaches 66.11% held-out and 59.43% OOD accuracy, compared with 64.47% and 57.03% for PB2, the strongest evaluated adaptive baseline. Three paired training seeds give gains of +2.42 and +3.19 percentage points over the contextual bandit on held-out and OOD evaluation. Under high behavior-deployment mismatch, policy-consistent risk lowers selected-decision ECE from 0.108 to 0.053; DCTC lowers it from 0.039 to 0.022 at matched acceptance; PRCC removes false-feasible admissions on an 18-action catalog ($0.197\rightarrow0.000$); and enabling all three components reduces trajectory failure from 0.181 to 0.083 in a factorial ablation. [COMMENTS]40 pages, 4 figures [LINK]http://arxiv.org/abs/2610.02828v2 [DATE]2026-10-16 20:44:10+08:00 [CATEGORIES]cs.CLcs.LG
Two Minds Better Than One: Collaborative Reward Modeling for LLM Alignment [AUTHORS]Jiazheng Zhang, Wenqing Jing, Zizhuo Zhang, Zhiheng Xi, Shihan Dou, Rongxiang Weng, Jiahuan Li, Jingang Wang, Mingxu Chai, Shibo Hong, Yuming Yang, Chenhao Huang, Tao Gui, Qi Zhang, Xuanjing Huang [ABSTRACT]Reward models (RMs) are central to aligning large language models (LLMs) with human values, yet noisy preference feedback can induce reward misgeneralization through spurious correlations and overfitting. We characterize the training dynamics of preference pairs and uncover a distinctive optimization signature of noise: persistently high losses, sharp temporal fluctuations, and irregular gradient updates. Our analyses further show that excluding such pairs improves downstream alignment. In practice, however, reliable preferences are not known a priori, and self-filtering based on a model's own training signals can amplify confirmation bias. To address this challenge, we introduce Collaborative Reward Modeling (CRM), an online framework for identifying reliable preferences. CRM couples cross-model peer review with curriculum learning: two RMs evaluate each other's sample selections, while curriculum keeps their learning progress synchronized. Across extensive experiments, CRM improves RM generalization by up to 9.94 points on RewardBench under 40% preference noise. CRM further generalizes to implicit-reward alignment, offering a robust and broadly applicable approach to learning from imperfect feedback. [LINK]http://arxiv.org/abs/2505.10597v3 [DATE]2026-10-16 20:42:25+08:00 [CATEGORIES]cs.LGcs.CL
Actions with Receipts: Jointly Binding Claims, Evidence, and Execution for Replayable Tool-Agent Auditing [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Yina Sa, Daren Zha [ABSTRACT]Tool-using agents can expose citations and execution logs while leaving a critical association unaudited: whether the claim shown to a user is the claim emitted by the committed execution and supported by the cited source. A valid citation and a valid trace can therefore remain individually well formed while being transplanted across claims, actions, runs, or source versions. We introduce a claim-anchored execution contract that jointly binds the emitted claim, its exact source span, the ordered execution prefix that produced it, and the source version and access state observed by that execution. Each receipt contains an emission anchor that deterministically locates the claim inside a committed answer or claim-bearing action, together with source identifiers, offsets, hashes, quotes, and a domain-separated execution commitment. A deterministic integrity verifier reconstructs these bindings before semantic or task labels are joined. We separate this integrity plane from a pluggable support plane, so structural validity is not used as a proxy for entailment. The contract exposes seven independently testable properties: claim-emission binding, source binding, ordered-execution binding, oracle separation, persisted-object replay, execution-rerun consistency, and version/access binding. Across 1,280 cross-object attacks, the joint contract detects 1,275 substitutions (0.9961). Removing a targeted property reduces its attack-detection rate to 0.0156-0.0625. On an independently adjudicated 384-pair split, the conflict-aware support guard reaches F1 0.8865 and false acceptance 0.0729; on unseen failure families, these rates are 0.8679 and 0.0938. [COMMENTS]35 pages, 8 figures [LINK]http://arxiv.org/abs/2610.00327v2 [DATE]2026-10-16 20:28:48+08:00 [CATEGORIES]cs.CLcs.LG
ContractRL: Shielded Group-Relative Policy Optimization for Auditable Tool-Call Repair [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Yina Sa, Daren Zha [ABSTRACT]Structured tool calls often fail after only a small number of fields violate a schema or an execution contract. Regenerating the complete object enlarges the action surface and makes repeated repair difficult to audit. We introduce ContractRL, a contract-constrained sequential repair protocol that models verifier-guided JSON repair as a bounded decision process. At each step the policy observes the candidate, typed verifier feedback, JSON Pointer, immutable repair history, and remaining budget; a contract-derived action mask filters malformed or prohibited RFC-6902 operations before a deterministic validator performs the transition. We specify a contract-constrained group-relative objective for patch, retry, and abstention decisions while keeping canonical targets and semantic labels outside the online state until trace freeze. Under identical verifier information, ContractRL attains 0.9362 semantic success with 34.4 generated tokens, compared with 0.9076 and 44.9 tokens for Patch-SFT and 0.9148 and 137.2 tokens for full regeneration over 192 cases per seed and five seeds. Policy optimization improves semantic success from 0.9186 for supervised ContractRL to 0.9375. A separate three-seed paired evaluation against Patch-SFT yields a semantic difference of +0.0396 (95% CI $[+0.0137,+0.0662], p=0.0039$). Feedback, action-mask, budget, and schema-shift analyses connect these gains to localized correction, while adversarial and multi-turn evaluations characterize the remaining failure modes. [COMMENTS]29 pages, 8 figures [LINK]http://arxiv.org/abs/2610.00328v2 [DATE]2026-10-16 20:23:26+08:00 [CATEGORIES]cs.CLcs.LG
Dependency Reasoning for Aspect Sentiment Quadruple Prediction [AUTHORS]Shima Esfandyari, Seywan Soleymanbaigi, Fardin Akhlaghian Tab, Fatemeh Daneshfar, Mourad Oussalah [ABSTRACT]Aspect Sentiment Quadruple Prediction (ASQP) is fundamentally a relational reasoning problem: identifying aspect-opinion-category-sentiment quadruples requires modeling structured dependencies, not merely sequential co-occurrence. Existing LLM-based approaches incorporate syntactic information indirectly, via retrieval-based demonstration selection or multi-stage training, rather than as unified, single-pass generation-time context. We present DRIFT, which linearizes dependency parse trees and injects them directly into LLM prompts as explicit relational context, enabling the model to reason over aspect-opinion dependencies rather than infer them from token proximity alone. DRIFT further integrates retrieval-augmented in-context learning and a syntax-aware LLM-based adjudication mechanism that complements exact-match evaluation under span-boundary variation. On SemEval Rest15, Rest16, and ACOS, DRIFT achieves the strongest performance among compared generative ASQP methods, reaching 54.47% and 64.10% F1 on Rest15 and Rest16, with consistent gains across short- and medium-distance aspect-opinion configurations; adjudication further shows exact-match metrics systematically underestimate semantically correct predictions. [LINK]http://arxiv.org/abs/2610.14873v2 [DATE]2026-10-16 20:03:17+08:00 [CATEGORIES]cs.CLcs.LG
FAER: Auditable Utility-Aligned Trajectory Replay for Language Model Post-Training [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Tianshu Fu, Daren Zha [ABSTRACT]Replay selectors often rank cached trajectories by format feedback, confidence, freshness, or response length, although cache-level correctness and downstream learner utility are distinct objectives. We formalize this selection-to-learning gap and introduce FAER as an auditable full-trajectory replay framework. Its training-free fixed selector is a protocol baseline; FAER-UTILITY is the learner-aware selector fitted on disjoint calibration blocks. The normalized gradient alignment is reported as a baseline, while a disposable optimizer-aware virtual update supplies a magnitude-aware utility surface. The audit contract freezes observed fields and replay traces before evaluation labels are joined. On GSM8K with Qwen2.5-1.5B-Instruct, the matched learner study reports quality 0.6329 for the fixed selector, compared with 0.5482 for uniform and 0.6037 for format-feedback under 128 updates. Metadata-only cross-fitted calibration reaches $0.6476\!\pm\!0.0139$ over eight seeds (median 0.6481; paired 95% interval $[+0.079,+0.122]$) at 63,276 target-run tokens; its recorded full cost is 189,642 tokens and 3.48 GPU-hours including calibration. The completed FAER-UTILITY row reaches 0.6624 at 62,844 target-run tokens and 4.26 GPU-hours. Format-feedback selects records with correctness 0.6953, compared with 0.3594 for the fixed selector, despite the different downstream ranking. The completed comparison surfaces report the learner-aware ablation, same-seed gap, policy-optimization rows, and strict zero-shot transfer. [COMMENTS]35 pages, 6 figures [LINK]http://arxiv.org/abs/2610.00385v2 [DATE]2026-10-16 20:01:19+08:00 [CATEGORIES]cs.LGcs.CL
Language Models Learn to Improve Language Models [AUTHORS]Minhua Lin, Zijun Wang, Zewen Liu, Zhaotian Weng, Bing He, Zhan Shi, Benoit Dumoulin, Xiang Zhang, Suhang Wang, Hanqing Lu [ABSTRACT]Large language models (LLMs) are increasingly used to automate AI research, including language-model (LM) development, offering a potential path toward recursive self-improvement. However, turning experimental feedback into a learned ability to propose effective improvements across models faces two challenges. First, training such an ability with real feedback is expensive: reinforcement learning (RL) requires evaluating tens of thousands of proposals, each potentially requiring a full model-training run. Even at ten GPU-hours per run, reward collection alone would consume hundreds of thousands of GPU-hours, motivating a cheaper surrogate. Second, surrogate feedback can become unreliable as policy optimization favors overestimated proposals when the proposal distribution shifts. To address these challenges, we propose LM2LM, which trains LLMs with RL to generate effective LM improvement proposals. To provide scalable feedback, we use a research world model (RWM), a frozen LLM that predicts proposal gains from real experimental records provided in context, as the reward model for RL. To maintain feedback reliability, we periodically execute current-policy proposals ranked highly by the RWM and add their measured outcomes to its context. Across eight autoregressive and diffusion LM development environments, real-execution evaluations on held-out tasks show that LM2LM enables Qwen3.8-27B to outperform GPT-6-sol, even when GPT-6-sol is given 300 real experimental records in context, while requiring 33% lower data-generation cost. LM2LM also generalizes to a model family unseen during training. [LINK]http://arxiv.org/abs/2610.23344v1 [DATE]2026-10-16 19:51:00+08:00 [CATEGORIES]cs.CL
Robust Function Calling for LLMAgents Under Interface Drift [AUTHORS]Aniruddha Chattopadhyay, Kanishka Halder, Partha Sarathi Roy [ABSTRACT]Tool-augmented language-model agents rely on interface text: tool descriptions, argument schemas, and the candidate tool list shown at inference time. These interfaces often change after deployment even when the underlying tool semantics remain fixed, creating a practical reliability problem for function-calling systems. We study this failure mode as interface drift and evaluate whether a lightweight, training-free repair layer can recover robustness without retraining the base model. Our method, SchemaShield-Lite, combines schema validation, candidate-set retry for missing tool names, and one-shot constrained repair. On Qwen3.5-9B, it raises DICE-300 accuracy from 0.717 to 0.813 and BFCL-200 accuracy from 0.580 to 0.805, recovering 19/22 and 33/45 drift-induced failures with no observed repair harms. Naive retry does not explain the gains. Ablations show that candidate-set drift, especially distractor tools, accounts for most observed damage, while description and schema edits alone are less harmful. The results position validation-and-repair as a practical robustness layer for deployed LLM-agent tool use. [COMMENTS]16 pages, 1 figure, 6 tables. Peer-reviewed and accepted at NGEN-AI 2026 (International Conference on Next Generation AI Systems). This is the pre-review submitted version [LINK]http://arxiv.org/abs/2610.23342v1 [DATE]2026-10-16 19:49:37+08:00 [CATEGORIES]cs.CL
Audit-First VAPO: Risk-Certified Selective Updates under Imperfect Verification [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Rui Chen, Daren Zha [ABSTRACT]Imperfect verifiers can assign a harmful update direction even when clipping and regularization bound its magnitude. We introduce Audit-First VAPO, which separates discrete directional admission from continuous magnitude control. An observation-only accept-appeal-abstain policy uses a finite secondary-verification budget; its action trace is frozen before clean labels are joined. Simultaneous finite-sample bounds then certify selected harmful risk, coverage, and verifier-call rate over a predeclared policy family. Conditional Hoeffding-Azuma bounds account for the dependence induced by shared budgets, and rollout or verifier changes initiate a new certification stage. After admission, a bounded trust-clip-KL actuator controls magnitude. We evaluate two models on two reasoning benchmarks against static RLVR, matched-random selection, confidence thresholding, noise correction, and verifier augmentation. On Qwen3.5-0.8B and GSM8K at target risk $ρ=0.08$, RC-VAPO achieves 74.1% accuracy, selected harmful risk 0.0697, coverage 0.4125, and relative verifier cost $1.16\times$. At matched coverage and update magnitude, its selected-risk difference from matched random is -0.0260 with paired 95% interval $[-0.0364,-0.0157]$. Across asymmetric, confidence-dependent, and correlated-verifier noise, the certificate is satisfied on 57 of 60 independent runs. These comparisons isolate informative directional selection from proposal suppression, update shrinkage, and additional verifier computation. [COMMENTS]33 pages, 7 figures [LINK]http://arxiv.org/abs/2609.33662v2 [DATE]2026-10-16 19:41:12+08:00 [CATEGORIES]cs.CLcs.LG
RubricEM: Meta-RL with Rubric-guided Policy Decomposition beyond Verifiable Rewards [AUTHORS]Gaotang Li, Bhavana Dalvi Mishra, Zifeng Wang, Jun Yan, Yanfei Chen, Chun-Liang Li, Long T. Le, Rujun Han, George Lee, Hanghang Tong, Chen-Yu Lee, Tomas Pfister [ABSTRACT]Deep research agents plan, search, evaluate evidence, and synthesize long-form reports. Training these agents with reinforcement learning (RL) is challenging because they operate over long horizons and their outputs lack exact answer verifiers. To this end, we introduce RubricEM, a unified RL framework for deep research that combines stagewise credit assignment with reflection-based meta-RL for experience reuse. Rubrics unify the framework by providing a shared interface for policy execution, judge feedback, and experience reuse. RubricEM first instills a rubric-guided scaffold through supervised fine-tuning, teaching the agent to plan with rubrics and organize its trajectory into semantic stages. Building on this structural prior, we optimize the task policy with Stage-Structured GRPO, using stagewise judge feedback for finer-grained credit assignment. In parallel, we asynchronously train a shared-backbone reflection meta-policy (with meta-RL) to distill completed trajectories into reusable guidance, contributing through both shared-parameter updates and textual memory. The resulting RubricEM-8B model achieves strong performance across four long-form research benchmarks, outperforming existing open models of larger sizes and approaching proprietary deep-research systems. Beyond final performance, we perform thorough controlled analyses to understand the key ingredients of RubricEM. [COMMENTS]67 pages, 6 figures [LINK]http://arxiv.org/abs/2605.10899v2 [DATE]2026-10-16 19:35:29+08:00 [CATEGORIES]cs.CLcs.LG
HyperPersona: A Multi-Level Hypergraph Framework for Text-Based Automatic Personality Prediction [AUTHORS]Sina Heydari, Majid Ramezani [ABSTRACT]As a modern commodity, language has become a vast repository of socially and psychologically significant traits and concepts, reflecting the ways people encode pattern of thoughts, behaviors, and emotions into words. Text-based Automatic Personality Prediction (APP), seeks to infer personality from linguistic behavior, offering a scalable alternative to traditional psychometric assessments. Although text is inherently hierarchical, with the document-level capturing global features, the sentence-level encoding local semantics, and the word-level providing fine-grained lexical information, most existing approaches rely on shallow, sequential, or single-level representations that ignore the multi-level structure of written language. To address this, we propose HyperPersona, a framework that explicitly models the hierarchical organization of text (document, sentence, and word) through hypergraph structure, where a document and its sentences are represented as hyperedges, and the words are represented as nodes, enabling joint modeling of global, local, and lexical dependencies of text. Followed by a transformer-based graph encoder that learns interactions within and across these linguistic layers, yielding context-sensitive and structurally grounded feature representations for personality prediction. Experiments on the Big Five personality dimensions show that, while relying solely on text, HyperPersona effectively integrates multi-level linguistic cues, achieving superior performance compared to state-of-the-art baselines. These findings underscore the critical role of textual hierarchy in advancing human-like personality inference from natural language. [COMMENTS]The manuscript is currently under peer review [LINK]http://arxiv.org/abs/2605.17355v2 [DATE]2026-10-16 19:32:35+08:00 [CATEGORIES]cs.CL
CascadeLedger: Counterfactual Auditing of Information and Resource Integrity in LLM Cascades [AUTHORS]Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Daren Zha [ABSTRACT]We introduce CascadeLedger, a counterfactual execution-audit framework for budgeted language-model cascades. The protocol enforces call-before-read semantics, binds successful reads to service and cache identities, records availability and component-wise budget transitions, and freezes the event trace before any target or task scorer is joined. The central object is aggregate-equivalent trace validity: two executions can share a final answer, declared resource projection, and service histogram while differing in whether their routing decisions used only legitimately observable information. We complement single-trace invariants with counterfactual prefix non-interference (CPNI), which couples a visible prefix and policy randomness while changing only a declared hidden family. The counterfactual influence frontier (CIF) localizes the earliest violating prefix and a minimal hidden witness; projection-preserving adversarial mutation (PPAM) searches for invalid traces that remain acceptable to a non-counterfactual checker. Across frozen-cache, independently generated, and serving evaluations, the two-layer audit reports a clean false-positive rate of 0.0000, detects 0.9983 of serialized faults and 0.9982 of unlogged hidden-read faults, and changes 0.3403 of corrupted quality-resource conclusions after deterministic recomputation. The online recorder and offline CPNI audit have separate overhead measurements. Learned routing and semantic verification provide downstream compatibility studies under the same execution contract. [COMMENTS]38 pages, 6 figures [LINK]http://arxiv.org/abs/2610.20668v2 [DATE]2026-10-16 19:25:48+08:00 [CATEGORIES]cs.CLcs.LG
Self-Reflective Chain-of-Thought (SRCoT): Evidence-Based Text Generation via a Closed-Loop Retrieve--Reflect--Revise Pipeline on Edge LLMs [AUTHORS]Fabian Broszio, Michael Hoffmann, Adrian Paschke [ABSTRACT]Large language models (LLMs) can generate fluent outputs that lack source support or conflict with local operational constraints. We introduce Self-Reflective Chain-of-Thought (SRCoT), a modular architecture that organizes generation as a closed Retrieve--Reflect--Revise loop. SRCoT separates environment state, persistent knowledge, accumulated experience, and short-term memory into provenance-labeled stores. BM25 retrieval supplies local context; subsequent passes expose intermediate drafts, missing-information markers, follow-up queries, and optional claim--evidence mappings. These artifacts inform retrieval, revision, clarification, or abstention decisions. We examine the architecture through qualitative, trace-based case studies in two fully synthetic settings: emergency medical services and Munich-Sim, an emergency-operations simulation. Local prototype runs on an Apple M4-family laptop with 128 GB memory illustrate source-labeled documentation, explicit gap marking, and structured dispatch-style outputs. The traces also reveal provenance-assignment errors, incomplete evidence coverage, and unsafe assumptions that remain possible despite prompting constraints. These cases demonstrate report-level inspectability and local execution, not comparative superiority, hallucination reduction, clinical validity, or operational safety. The architecture separates retrieval and control from the model interface, but behavior across different models remains to be evaluated. SRCoT is a research proof of concept requiring human review, with its principal contribution lying in an inspectable control procedure rather than a new model. [COMMENTS]28 pages, 2 figures. Extended technical account; a related brief report based on overlapping work has been submitted to Frontiers in Disaster and Emergency Medicine. The two manuscripts do not represent independent empirical studies [LINK]http://arxiv.org/abs/2610.23311v1 [DATE]2026-10-16 19:14:52+08:00 [CATEGORIES]cs.CL
Inferring L1 Backgrounds from L2 English with LLM Token Probabilities [AUTHORS]Ahmet Yavuz Uluslu, Mark Gales, Kate Knill, Gerold Schneider [ABSTRACT]Native language identification (NLI) infers a non-native speaker's first language (L1) from their second language (L2) production. We investigate a token probability approach that scores each text sample as a continuation of a prompt prefix specifying a candidate L1, then predicts the candidate with the highest score. We evaluate our approach using Llama 3.1 8B and 70B on three datasets: TOEFL9, Write and Improve 2024, and BULATS-L1. We compare its performance with previous prompting-based classification approaches. Token-probability scoring substantially improves on direct classification, particularly with the 8B model, while the 70B model achieves state-of-the-art performance across all datasets. Our results indicate that pretrained LLMs' token probabilities contain information useful for distinguishing L1 backgrounds from L2 English production. [LINK]http://arxiv.org/abs/2610.23309v1 [DATE]2026-10-16 19:12:42+08:00 [CATEGORIES]cs.CL
Annotrace: A Workbench for Step-Level Annotation of Agent Traces [AUTHORS]Arun Josephraj Arokiaraj, Chenyi Huang, Sahan Bulathwela, Zekun Wu, Bin Wu [ABSTRACT]Every study of why LLMagents fail begins by labelling traces by hand, and every such study builds its own throwaway tool. Observability platforms record agent traces but offer no structured labelling; annotation tools label text but have no notion of a step, a delegation or a tool call, leaving it unclear what the smallest unit of execution to label should be. Annotrace reads six trace formats, including OpenTelemetry GenAI, LangSmith and Langfuse, and turns each run into an ordered, nested list of steps. The annotator writes their own labelling codebook; a frontier model proposes a starting set from a sample of their own traces, answers first, and is asked exactly the questions the human is asked, making annotation a joint human-AI process. Annotrace measures agreement between any two annotators, human or AI. On a multi-agent question-answering corpus it reproduces a previous manual effort's 235 annotation units exactly, and two models labelling the same traces agree 91% of the time at $κ= 0.82$. [COMMENTS]4 pages, 4 figures, 1 table. Under review at the AAAI 2027 Demonstration Program [LINK]http://arxiv.org/abs/2610.23284v1 [DATE]2026-10-16 18:55:26+08:00 [CATEGORIES]cs.CL
Verbalising Model's Computation for Chain-of-Thought Faithfulness Detection [AUTHORS]Yujing Wang, Yulan He, Wenya Wang, Hanqi Yan [ABSTRACT]Current Chain-of-Thought (CoT) verification methods detect reasoning faithfulness from model outputs or static activations, but are limited in explaining how unfaithfulness develops during generation. We introduce VERITAS to verbalise the model's computation: how different reasoning anchors emerge, transfer across layers, and surface in the output. VERITAS adopts the layer-wise Jacobian projection to produce a verbalised account of the model's internal reasoning evolution at each generated CoT token. By training a classifier on the extracted features along the traces, we show that these structured features provide insightful perspectives on how and where reasoning diverges toward unfaithfulness. Experiments on four datasets from FaithCoT-Bench and BonaFide show that VERITAS achieves a significant performance gain over LLM-as-Judge, with a 28.4% gain in balanced accuracy, and an average improvement of 16.0% over activation probing. Further analysis shows VERITAS can be generalised to different datasets and model architectures. Our work shows that verbalising model computation provides a richer and more interpretable lens on CoT faithfulness, moving beyond detection toward a mechanistic understanding of how reasoning fails. [LINK]http://arxiv.org/abs/2610.23275v1 [DATE]2026-10-16 18:44:23+08:00 [CATEGORIES]cs.CL
SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models [AUTHORS]Hongxing Li, Jinyue Su, Dingming Li, Wenqi Zhang, Weiming Lu, Jun Xiao, Yueting Zhuang, Yongliang Shen [ABSTRACT]Existing spatial reasoning benchmarks mainly test spatial perception: reading off relations already visible in the input. Yet real-world spatial intelligence demands predictive spatial reasoning: constructing a scene from observations, anticipating how an intervention changes it, and reasoning about the unseen outcome. We introduce SpaceCast-Bench, the first benchmark to directly and diagnostically evaluate this capability. Built around an observe-transform-infer framework, its 3,862 questions from 182 real-world scenes span 16 task types at three levels: static perception, local prediction, and global prediction, progressively requiring scene understanding, spatial state updating, and relational inference over unobserved outcomes. Evaluating 21 models exposes a stark gap: the strongest model reaches only 58.0% against 87.2% human performance, while spatially specialized models remain near random chance. Controlled analyses further reveal that bridge views are critical for integrating distributed observations, and that explicit 3D evidence benefits models more reliably than generated outcome images or videos. Fine-tuning on our programmatically generated data lifts Qwen3-VL-4B from 34.0% to 65.7% with macro-average gains across six out-of-domain benchmarks. [COMMENTS]Project Page: https://zjureal.com/SpaceCast-Bench Code: https://github.com/ZJU-REAL/SpaceCast-Bench [LINK]http://arxiv.org/abs/2610.12402v2 [DATE]2026-10-16 18:40:25+08:00 [CATEGORIES]cs.CL
Reasoning Models Can Control Their Chains of Thought When They Tell Themselves To [AUTHORS]Uriel Dolev, Idan Schwartz, Yoav Goldberg [ABSTRACT]Chain-of-thought (CoT) controllability, the ability of reasoning models to control properties of their CoT, is consequential for both their usefulness and reliable oversight. Recent work evaluates this capability by measuring whether models follow CoT constraints given in the user prompt, and reports very low controllability across models. We argue that this low controllability may not reflect an inherent inability of reasoning models to control their CoT, but instead result from naively specifying constraints in the user prompt. To test this, we introduce Self-Instructed CoT Controllability: a complementary evaluation that places the constraint as a self-instruction inside the model's CoT and measures its ability to follow it. We show that, under self-instruction, models achieve significantly higher controllability than with standard user instructions. Even a generic reminder to follow the user's constraint substantially improves controllability. Moreover, effective self-instructions can be bootstrapped into the reasoning through a simple instruction in the user prompt, enabling improved control even without access to the reasoning channel. Finally, using self-instruction, we demonstrate in a realistic monitoring setup that CoT controllability can be leveraged for monitor evasion, posing a risk to CoT monitorability. [COMMENTS]26 pages, 19 figures, 18 tables. Code and data: https://github.com/udolev/Self-Instructed-CoT-Controllability [LINK]http://arxiv.org/abs/2610.23257v1 [DATE]2026-10-16 18:31:27+08:00 [CATEGORIES]cs.CLcs.LG
Sample, Mix, or Verify? Allocating Test-Time Compute under Imperfect Verifiers [AUTHORS]Srinivasan Iyengar [ABSTRACT]A common way to make language models and agents more accurate is to spend more compute at inference time: generate many candidate solutions and submit the one a verifier prefers. Success then requires that some candidate is correct and that the verifier picks it. The first grows predictably with the number of samples; the second does not, because real verifiers (unit tests, majority votes, LLM judges, reward models) accept wrong answers, reject right ones, and make correlated mistakes. With a fixed budget, a practitioner must decide how many samples to draw, which models to draw them from, and how much to spend on verification, and today these choices are made by rules of thumb. We introduce SPLIT, which makes them from a small labelled pilot. It rests on an exact formula for the probability of submitting a correct answer under accept/reject and top-score selection that accounts for varying task difficulty, blind spots shared across models, and correlated repeated verifier calls. SPLIT fits the formula by empirical Bayes and uses it to plan budgets on unseen tasks. Across seven benchmarks spanning math, science QA, code and software-engineering agents, 26 models, 175 agent systems and six kinds of real verifiers, all from public releases, SPLIT lands within 0.02-0.8 percentage points of the best allocation in hindsight, while every fixed rule we test loses at least 2.5 points in some setting. Regret depends only on how well SPLIT ranks configurations, and it stays near-optimal under compute, price and latency costs and online. The formula also explains when asking more LLM judges lowers precision. As a forecaster far beyond the pilot, SPLIT beats power-law fitting when the verifier caps success, and we report where it loses. [COMMENTS]27 pages (18 main + appendix), 14 figures, 11 tables. Code will be released [LINK]http://arxiv.org/abs/2610.23255v1 [DATE]2026-10-16 18:28:07+08:00 [CATEGORIES]cs.LGcs.CL
Ignored, Not Refused: What Low Prompt-Injection Success Rates Hide in Small Tool-Using Agents [AUTHORS]Farhan Zarif, Kazi Shahid Ahmed Galib, Synthia Akter Promi, Shabnam Sharmin Tarin, Ariyan Hossain, Sadif Ahmed, Farig Yousuf Sadeque [ABSTRACT]Tool-using agents read untrusted external content, which exposes them to indirect prompt injection (IPI), where hidden instructions hijack the agent's tool calls. Benchmarks measure IPI security by the attack success rate (ASR), which treats every failed attack as a successful defense. To test this assumption, we audit 33,000 trajectories from eleven open-weight agents (2-14B parameters) on InjecAgent and AgentDojo. A label-blind LLM judge shows that most failed attacks are ignored (76-80%) rather than refused (17-18%). With identical model weights, changing only the system prompt shifts InjecAgent ASR from 0.2% to 43%. Our Attack-Task Outcome Matrix (ATOM) further shows that in 21.7% of AgentDojo episodes, the agent stayed safe but did not complete the user's task. A low ASR therefore mostly reflects agents silently ignoring injections, not refusing them. Evaluations should report how attacks fail and whether the user's task is completed, alongside ASR. We release the labeled corpus, judge prompt, and analysis code. [COMMENTS]17 pages, 3 figures, 11 tables [LINK]http://arxiv.org/abs/2610.23254v1 [DATE]2026-10-16 18:26:23+08:00 [CATEGORIES]cs.CL
Who Pays for Efficient Inference? Confidence-Based Small-to-Large LLM Routing Across Tasks and Languages [AUTHORS]Md Sabbir Hossen, Anichur Rahman, Pabon Shaha, Airin Afroj Aishi, Andrew H. Sung, Md Shohel Rana [ABSTRACT]Small-to-large language model routing can reduce inference cost by escalating difficult inputs while retaining easier ones on a smaller model. Its success, however, depends on identifying examples that benefit from escalation, and the reliability of confidence across multilingual tasks remains unclear. We present a systematic evaluation of confidence-based routing across three model families, four languages, and three task types, covering 72 model, dataset, language, and scale configurations and 42,288 predictions. We compare global confidence routing with random routing, equal-quota confidence routing, and an oracle upper bound. Confidence identifies beneficial escalations on the Belebele reading comprehension and SIB-200 topic classification datasets, with mean correction-detection AUCs of 0.612 and 0.705. At 50% escalation, it improves over random routing by 1.40--4.67 percentage points on Belebele and 2.54--5.63 points on SIB-200. In contrast, the mean AUC on the MMLU-ProX-Lite knowledge and reasoning dataset is only 0.510, and routing provides no consistent improvement over random selection because small models often remain highly confident while incorrect. Global confidence routing also creates language escalation gaps up to 30.33 percentage points, whereas equal-quota routing removes this disparity with little aggregate accuracy cost. These results show that confidence routing is task-dependent and can allocate computation unevenly across languages. Code is available at https://github.com/sabbir5622r/multilingual-llm-routing [LINK]http://arxiv.org/abs/2610.23249v1 [DATE]2026-10-16 18:23:41+08:00 [CATEGORIES]cs.CL
Think More or Read More? A Token-Budget Theory of Reasoning with Shared Context [AUTHORS]Manh Nguyen, Sunil Gupta, Hung Le [ABSTRACT]Reasoning systems increasingly improve answers by incorporating additional context into the model's input, from multi-agent debate, where agents read their peers' responses, to self-refinement, where a model revisits its own drafts. In these systems, communication consumes context as one agent's output becomes another's input. We study the regime where token reading and generation draw on one fixed token window, so every token spent reading is a token that cannot be spent reasoning. We develop a \textbf\{token-budget theory\} that formalizes this trade-off, modeling accuracy as a function of the reasoning budget remaining after reading and aggregating votes with a Condorcet jury model that permits correlated errors. For multi-agent debate, the theory gives a closed-form prediction for the optimal number of agents. This quantity is estimated from single-agent measurements, allowing the optimal group size to be determined without running debate, and the same accounting extends to a heterogeneous set of different models. We empirically verify our theory and show that, at the predicted group size, our method outperforms strong baselines under the tightest calibration spend, where debate-calibrated rules can afford only one or two questions. Self-refinement obeys the same accounting, with its own draft in place of a peer message. The theory further predicts that communication pays only when the gain in the accuracy ceiling exceeds the reasoning room spent on reading, with this condition tightening as the window shrinks. Across five open models and four reasoning tasks, self-consistency matches or beats deployed debate at $93$\% of $980$ matched window-and-group-size settings, with the gap being largest under tight windows and nearly disappearing at a loose one, consistent with the theory. Under a shared window, how much room a model has left to \textit\{think matters more than what it reads\}. [LINK]http://arxiv.org/abs/2610.23246v1 [DATE]2026-10-16 18:18:15+08:00 [CATEGORIES]cs.CLcs.LG
MGE-PO: Entropy-Aware Credit Assignment for Mathematical Reasoning [AUTHORS]Jihao Huang, Guangjia Chai, Zhirui Chen, Lei Wang, Yao Liu, Liu Qi, Pengjie Ding, Xinyang Liu, Chengxun Jia, Ling Shao, Hongyu Chen, Junchen Wan [ABSTRACT]GRPO assigns the same group-relative advantage to equally rewarded responses within a question, leaving their generation uncertainty unused. We introduce MGE-PO, centered on four-quadrant (4Q) advantage shaping that combines verified correctness with rollout generation entropy. Correctness determines the sign of an additional correction, while entropy controls its magnitude: lower-entropy successes receive larger bonuses and higher-entropy failures receive larger penalties. An online entropy-based question curriculum supports this credit rule using cached rollout measurements. At the final checkpoint after eight training epochs on Qwen3-8B-Base, adding 4Q across three policy optimization settings yields absolute gains of 3.23-3.69% in six-benchmark macro avg@32 and 4.24-11.06% in macro pass@32. On Qwen2.5-7B-Base, the complete configuration raises four-benchmark macro avg@32 from 41.5% to 45.4%. [LINK]http://arxiv.org/abs/2610.23245v1 [DATE]2026-10-16 18:15:02+08:00 [CATEGORIES]cs.CL
MultiAgent-OPD: End-to-End Training of Hierarchical Multi-Agent Systems with On-Policy Distillation [AUTHORS]Yuhan Sun, Jiaxin Zhao, Yongdong Luo, Yibo Hu, Meiguang Jin, Junfeng Ma, Weixun Wang, Weihang Pan, Zhongan Bi [ABSTRACT]In a hierarchical multi-agent system, a Leader assigns tasks and subagents carry them out. Jointly training the Leader and the subagents is difficult because a final answer gives only indirect feedback to the subagents. An LLM judge can score the performance of the subagents, but these scores depend on the judge's own assessment of the quality of subagents' responses. We present MultiAgent-OPD, which combines on-policy distillation with reinforcement learning to jointly train a shared-parameter Leader-Searcher-Answerer team end to end. We first train a Teacher in a Searcher-Answerer workflow using answer correctness. During team training, the frozen Teacher scores student-generated search and summary tokens to provide local process supervision for the Searchers, while the Leader learns from team success and the Answerer from candidate correctness. Across seven datasets, the trained system reaches 38.68% macro avg@4, exceeding Prompted Critic by 4.06 percentage points and C-MAS + RL, the trained cascaded multiagent system, by 2.61 percentage points. The team also solves more questions within four attempts. [LINK]http://arxiv.org/abs/2610.23234v1 [DATE]2026-10-16 18:01:52+08:00 [CATEGORIES]cs.CL
STAIF: A Stage-wise Optimization for Complex Instruction Following [AUTHORS]Jian Hong, Chen Cheng, Quan Liu, Yuhao Chen, Enhong Chen [ABSTRACT]Following complex instructions with multiple explicit constraints remains a fundamental challenge for large language models (LLMs). Existing alignment methods, such as DPO, optimize holistic reward signals that often underemphasize strict satisfaction of individual constraints, particularly under out-of-distribution or multi-constraint settings. In this paper, we propose STAIF, a stage-wise optimization framework that decouples the alignment of subjective (soft) constraints from the optimization of objectively verifiable (hard) constraints. Stage 1 applies preference optimization with multiple negative samples to sharpen sensitivity to soft constraints, while Stage 2 applies Reinforcement Learning with Verifiable Rewards (RLVR) to enforce strict compliance with hard constraints. To support this method, we construct STAINSTRUCT, a high-quality bilingual (English, Chinese) dataset of approximately 31,000 complex multi-constraint instructions. Extensive analyses validate the design of STAIF and show state-of-the-art performance on representative benchmarks against strong baselines, as well as genuine generalization. [COMMENTS]16 pages, 6 figures [LINK]http://arxiv.org/abs/2607.22649v2 [DATE]2026-10-16 18:00:07+08:00 [CATEGORIES]cs.CL
ReLaDrift: Recurrent Language Generation via Drifting for Test-Time Scaling [AUTHORS]Yuki Otsuka, Daisuke Oba, Naoaki Okazaki [ABSTRACT]Drifting Models generate a sample with a single function evaluation, making them attractive for fast parallel language generation, but their one-step formulation does not support iterative refinement. We propose Recurrent Language Generation via Drifting (ReLaDrift), which uses a shared Drifting Model to generate an initial draft from noise and iteratively refine it in a continuous latent space. We train this recurrence by sampling the number of rounds and applying the drifting loss only at the final round, stopping gradients through preceding rounds so that only one round requires backpropagation. Unlike the one-step Drifting Model, whose accuracy does not improve when iterated, ReLaDrift enables a trade-off between test-time computation and generation quality, as diffusion and flow language models do. On Sudoku Hard and GSM8K, it achieves competitive or superior performance compared with these iterative-generation approaches across a range of inference budgets. Our code is available at https://github.com/Yuki-Ots/ReLaDrift [LINK]http://arxiv.org/abs/2610.23225v1 [DATE]2026-10-16 17:55:22+08:00 [CATEGORIES]cs.CL
MAS-Bai: Linguistically Informed Adaptive Multi-Agent Translation for Bowen-Script Bai [AUTHORS]Jingting Zheng, Hong Ren, Yuehan Cui, Shaolin Zhu, Deyi Xiong [ABSTRACT]Large Language Models (LLMs) remain unreliable for extremely low-resource languages whose writing systems conflict with high-resource pretraining priors. We study Bai-Chinese translation in the Bowen script, where Chinese characters are frequently borrowed phonetically rather than semantically, causing severe literal-translation hallucinations. We propose MAS-Bai, a linguistically informed adaptive multi-agent framework for Bowen-script Bai-Chinese translation. MAS-Bai routes each input through either a lightweight retrieval-translation path or a deeper philological path with etymological and syntacticagents, and uses cycle-consistency with a lightweight reverse translator as an inference-time reliability signal. We construct a 6,536-sentence Bai-Chinese corpus and evaluate on a 1,536-sentence test set. MAS-Bai achieves 16.23 BLEU and 62.15 COMET, improving over Standard RAG by 8.11 BLEU and 6.85 COMET. Forward LoRA, expert ablations and controlled routing experiments distinguish the roles of supervised adaptation, Bai-specific expertise and conditional routing. The largest gains occur on Bowen phonetic loans and divergent Bai word order, while human evaluation and reverse-verifier analyses support the automatic results. [COMMENTS]14 pages, 3 figures, Findings of EMNLP 2026 [LINK]http://arxiv.org/abs/2610.23205v1 [DATE]2026-10-16 17:43:32+08:00 [CATEGORIES]cs.CL
Brain-to-Language Decoding: Paradigms, Signals, Methods, Evaluation, Practical Use and Beyond [AUTHORS]Yiqian Yang, Yiqun Duan, Chenyu Liu, Yiqi Wang, Xinliang Zhou, Chin-Teng Lin, Yu Zhang [ABSTRACT]Brain-to-language decoding translates neural activity associated with language production, internal speech and perception into linguistic or expressive outputs. It offers a route to restoring communication after speech loss and a means of studying how the brain represents language. Advances in neural recording and representation learning have expanded the field from constrained recognition and acoustic reconstruction to text generation, streaming personalised speech and facial animation. This survey synthesises these developments across invasive and non-invasive measurements, drawing on a search without a lower year limit and source-led updates through September 2026. We connect Articulated, Inner and Perceived tasks to the neural populations they engage, the representations available to decoders and the outputs those representations can support. We examine model development, public resources and the evolution of evaluation, and compare published performance and communication costs within their reported protocols. The synthesis identifies complementary routes to progress: phonetic, acoustic and semantic targets preserve different aspects of a message; shared representations support reuse across recording conditions and tasks; and online communication increasingly depends on calibration, feedback and user control alongside decoding accuracy. Shared benchmarks enable algorithmic comparisons, while longitudinal studies reveal the demands of sustained use. We discuss these developments and their remaining limitations, then outline a prospective five-level trajectory from commands and language to meaning, scenarios and bidirectional cognitive exchange [LINK]http://arxiv.org/abs/2609.27650v4 [DATE]2026-10-16 17:37:51+08:00 [CATEGORIES]cs.CL
Pro-AI Bias in Large Language Models [AUTHORS]Benaya Trabelsi, Jonathan Shaki, Sarit Kraus [ABSTRACT]Large language models (LLMs) are increasingly employed for decision-support across multiple domains. We ask whether these models favor artificial intelligence (AI) itself. We examine this question through three experiments. First, in open-ended advice across investment, study, career, and startup domains, AI is prominent in most recommendation lists and is usually ranked first when it appears. Second, in salary estimations, models overestimate AI-related jobs more than they do for closely matched non-AI jobs. In a controlled follow-up, changing only the job title from the matched non-AI title to the AI-related title raises predicted salary by 10%, while all other supplied job details remain fixed. Finally, we show that the number of degrees and other credentials awarded by AI programs were 28% more overestimated than corresponding awards for matched computing programs. Overall, we show that AI is not only prominent in everyday LLM advice, but also receives systematically inflated quantitative estimates compared to non-AI cases, even against its closest disciplinary neighbors, raising concerns for high-stakes decision support. [COMMENTS]21 pages, 4 figures. Reproducibility code included as ancillary files [LINK]http://arxiv.org/abs/2601.13749v2 [DATE]2026-10-16 17:34:15+08:00 [CATEGORIES]cs.CLcs.LG
Token-Budget-Aware LLM Reasoning [AUTHORS]Tingxu Han, Zhenting Wang, Chunrong Fang, Shiyu Zhao, Shiqing Ma, Zhenyu Chen [ABSTRACT]Reasoning is critical for large language models (LLMs) to excel in a wide range of tasks. While methods like Chain-of-Thought (CoT) reasoning and enhance LLM performance by decomposing problems into intermediate steps, they also incur significant overhead in token usage, leading to increased costs. We find that the reasoning process of current LLMs is unnecessarily lengthy and it can be compressed by including a reasonable token budget in the prompt, but the choice of token budget plays a crucial role in the actual compression effectiveness. We then propose a token-budget-aware LLM reasoning framework that dynamically adjusts the number of reasoning tokens based on the reasoning complexity of each problem. Experiments show that our method effectively reduces token costs in CoT reasoning with only a slight performance reduction, offering a practical solution to balance efficiency and accuracy in LLM reasoning. Code: https://github.com/GeniusHTX/TALE [LINK]http://arxiv.org/abs/2412.18547v6 [DATE]2026-10-16 17:33:52+08:00 [CATEGORIES]cs.CLcs.LG
BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models [AUTHORS]Liubov Chubarova, Alexandra Kuleshova, Daniil Volkov, Kirill Sultanov, Alexey Zaytsev [ABSTRACT]While Multimodal Large Language Models (MLLMs) have made significant strides in visual comprehension, their ability to reason about text-dense, professional documents remains incompletely evaluated. Existing benchmarks emphasize information extraction, require external domain knowledge, or cover professional documents only as one of many settings. They are also largely English- or Chinese-centric, leaving other languages and Russian, in particular, substantially underrepresented. To address these limitations, we introduce BEAR-Bench (Bilingual Enterprise and Academic Reasoning), a self-contained, complex English-and-Russian benchmark comprising 1000 human-annotated questions based on text-rich business and scientific documents. We evaluate 16 proprietary and open-weight MLLMs, including Gemini 3.1 Pro and Qwen3.5-397B, on BEAR-Bench and observe clear headroom even for the strongest systems. Finally, we use the resulting model outputs to compare existing hallucination detection methods, evaluating not only how often models fail on BEAR-Bench but also how reliably those failures can be identified. [LINK]http://arxiv.org/abs/2608.17895v2 [DATE]2026-10-16 17:30:41+08:00 [CATEGORIES]cs.CL
AMEND: Constraint-Guided Multi-Agent Text Detoxification with Isolated Critics [AUTHORS]Tianyu Wang [ABSTRACT]Text detoxification requires balancing toxicity removal, semantic preservation, and fluency. Existing methods typically optimize an unconstrained scalar objective---enabling safety scores to mask semantic loss---or rely on open-ended iterative debate prone to semantic drift and high computational cost. We propose AMEND, an inference-time framework that decouples detoxification into role-conditioned proposal generation, isolated multidimensional criticism, and constraint-first selection. AMEND evaluates diverse candidates with role-blind critics and enforces strict semantic and fluency feasibility before ranking by utility. To maximize efficiency, AMEND Lite restricts subsequent search to at most one targeted repair round only when residual toxic spans remain. On ParaDetox, one-round selection over diverse roles improves the joint evaluation metric by $+0.0508$ over the best single role. On a frozen confirmation test set, AMEND Lite achieves a joint score of $0.6597 \pm 0.0062$, outperforming unconstrained multi-round loops by $+0.0489$ ($p < 0.05$) while reducing inference requests by 28.4%. [COMMENTS]10 pages, 2 figures, 4 tables [LINK]http://arxiv.org/abs/2610.23183v1 [DATE]2026-10-16 17:22:21+08:00 [CATEGORIES]cs.CL
What LLMs Do in the Shadows: Reading Context Reliance from the Unembedding Dark Space [AUTHORS]Alexandra Kuleshova, Andrey Volodichev, Daria Kotova, Alexey Zaytsev [ABSTRACT]Detecting faithfulness hallucinations in large language models (LLMs) is crucial for their safe deployment in high-stakes domains. Recent approaches address this by contrasting contextual against parametric knowledge utilization, yet measuring context reliance typically demands noticeable overhead, such as materializing attention maps, running additional forward passes, or training auxiliary probes. In this work, we show that context reliance can be estimated by projecting hidden states onto the dark space - directions that compose weakly with the unembedding matrix. Beyond its recognized function as an initial-token attention sink, we show that this subspace actively mediates context processing, and that its share of hidden-state norm - termed the Dark Matter Ratio (DMR) - provides a direct score of context utilization. Pairing DMR with a parametric knowledge score yields DarkLens, a lightweight faithfulness detector operating strictly within a single forward pass. Evaluated across three QA and summarization benchmarks on six open-weight LLMs (4B-13B parameters), DarkLens matches or outperforms a wide range of baselines without requiring extra forward passes or probe training. [LINK]http://arxiv.org/abs/2610.23182v1 [DATE]2026-10-16 17:21:49+08:00 [CATEGORIES]cs.CL
Tree-Aware Posterior Distillation for Long-Horizon Agentic Reinforcement Learning [AUTHORS]Heming Zou, Qi Wang, Yixiu Mao, Yun Qu, Yuyi Guo, Zeyuan Liu, Chenglin Li, Weijie Liu, Saiyong Yang, Xiangyang Ji [ABSTRACT]Reinforcement learning with verifiable rewards has become a scalable post-training paradigm for large language models. However, training agents over long-horizon interactions presents two challenges: using only terminal rewards makes turn-level credit assignment sparse and high-variance, while sequential mini-batch updates on a fixed rollout batch introduce policy staleness as the current policy drifts from the behavior policy. Existing approaches obtain denser credit via learned critics, process reward models, or on-policy distillation, but they rely on a value network, step-level annotations, or a stronger external teacher that increase training complexity. Meanwhile, importance sampling and clipping mitigate policy drift, but they do not specify a finite target that stops sequential updates from overshooting. In this work, we introduce TreeOPSD, a critic-free, tree-aware posterior self-distillation framework for long-horizon agentic reinforcement learning. Treating each interaction turn as a tree node, TreeOPSD branches alternative continuations from shared prefixes and recursively backs up terminal rewards into implicit turn-level credit. It then maps each credit to a closed-form, bounded teacher distribution parameterized by the likelihood ratio relative to the frozen behavior policy. Minimizing the divergence between this teacher and the student policy recovers the policy-gradient direction as the tangent at the behavior policy while bounding updates toward a finite target across sequential mini-batches. Empirically, TreeOPSD improves agentic capabilities across several competitive benchmarks and model scales, especially for long-horizon tasks, e.g., raising the average accuracy of Qwen3-30B-A3B-Thinking-2507 by 4.97 points over GRPO on the Deep Research setting. [LINK]http://arxiv.org/abs/2610.23181v1 [DATE]2026-10-16 17:17:26+08:00 [CATEGORIES]cs.LGcs.CL
Do Speech BCIs Need Larger Models? Rethinking Neural Decoding beyond Scaling [AUTHORS]Zehui Feng, Weichuan Wang, Xiaohan Chen, Cuntai Guan, Ting Han [ABSTRACT]Recent speech brain-computer interfaces (BCIs) increasingly rely on large models and complex multi-stage pipelines for neural decoding. However, in speech BCI, where training data are scarce, noisy, and expensive to collect, the benefits of model scaling remain unclear. This work revisits a fundamental question: are large models truly necessary for accurate neural decoding? We propose BEST (Brain-Encoding-Speech-Text), a two-stage framework for efficient neural-to-text decoding. Instead of relying on large generative models or phoneme-centric pipelines, BEST introduces hierarchical intermediate representations and a compact speech-oriented decoder that directly maps neural activity to text. This design preserves essential structure while significantly reducing computational cost, enabling robust generalization in low-resource settings and practical deployment under constrained resources. On the Brain-to-Text '24 and '25 benchmarks, BEST achieves word error rates (WERs) of 6.19% and 1.62%, respectively, using only 3.85% of the parameters required by comparable methods. Beyond these results, we provide three key insights: (1) decoder scaling yields non-monotonic gains; (2) cross-modal alignment is more critical than model size; and (3) overfitting degards sequence-level performance. Representation analysis further reveals a consistent neural-audio-text alignment pathway. [COMMENTS]25 pages, Accepted by NeurIPS 2026 as a Main Track conference paper [LINK]http://arxiv.org/abs/2610.23175v1 [DATE]2026-10-16 17:12:58+08:00 [CATEGORIES]cs.CL
Chinese Decision: A Family of Language Models for Single-Pass Probabilistic Decision-Making [AUTHORS]Qiang Yan [ABSTRACT]Many language-based applications require probabilities over a finite set of decisions rather than open-ended responses. We present Chinese Decision, a family of Chinese-language models for single-pass probabilistic decision-making. Given context, an instruction, and candidate descriptions, the model extracts final-position candidate-letter logits in one forward pass and normalizes them over valid options, without autoregressive answer generation. A shared interface supports Boolean judgments, categorical choices, and ordinal scores. We report a completed 0.8B model trained with distribution cross-entropy on 13.048 million decisions and preliminary 2B checkpoints trained with an additional RLCD objective. On the full CJ-Bench of 307,900 decisions, the uncalibrated 0.8B model achieves 72.24% general-domain accuracy and 76.98% average accuracy across medicine, law, and finance using a single checkpoint, with 1.07% overall expected calibration error. The 2B model reaches 72.56% general-domain accuracy after 40.13% of one epoch but has not surpassed the completed 0.8B model on the professional-domain average. We further evaluate a validation-temperature-scaled deployment model and its Q8 quantization, finding close aggregate performance while documenting individual prediction changes. We release the training, evaluation, and deployment pipeline. The current results establish a reproducible approach to Chinese probabilistic decisions but do not isolate the effect of model scale. [COMMENTS]11 pages, 2 figures, 11 tables. Code: https://github.com/yanqiangmiffy/chinese-decision [LINK]http://arxiv.org/abs/2610.23170v1 [DATE]2026-10-16 17:08:32+08:00 [CATEGORIES]cs.CL
Explaining Chess and Go in Natural Language [AUTHORS]Ori Press, Alexander Golubev [ABSTRACT]Modern chess engines play far beyond human ability, often choosing moves whose underlying ideas are difficult for humans to understand. We propose an automatic metric for evaluating chess explanations, grounded in chess-engine evaluations. Evaluating state-of-the-art language models using this metric, we find that they perform surprisingly poorly, often relying on surface-level board observations rather than robust, transferable chess concepts. We then introduce BinProp, a training-free method that enables these same models to discover reusable chess concepts and use them to explain moves in natural language. BinProp improves explainability for every model tested, nearly tripling the average score compared with direct prompting. Beyond chess, the same method discovers useful concepts in Go, draughts, and shogi without additional tuning. [LINK]http://arxiv.org/abs/2610.23166v1 [DATE]2026-10-16 17:05:17+08:00 [CATEGORIES]cs.CLcs.LG
EKLAV: Elicited Key CoT Learning via Answer only superVision [AUTHORS]Adarsh Singh, Aman Singh, Kushal Raj Bhandari, Soham Dan, Anoop Kumar, Vivek Gupta [ABSTRACT]Reasoning distillation trains a student to reproduce both the teacher's chain-of-thought trace and its final answer. We ask whether the trace is better used as privileged training context than as an imitation target. EKLAV keeps the teacher's reasoning in the input sequence but supervises only the final task answer, applying a task-dependent placement that moves part or all of the trace into the prompt. On BRIGHT passage reranking, EKLAV improves average nDCG@10 at all four tested checkpoints across two model families, winning 41 of 48 model-domain comparisons while reducing training compute by up to 31.6% and inference time by up to 32.7%. On four held-out table-reranking benchmarks, it is the best of four conditions on three, exceeding full-trace distillation by 3.0 points and answer-only training without reasoning by 5.5 points on average. Analysis of generated traces reveals that EKLAV students defer their first verbalized verdict later (0.76 vs. 0.44 relative position), reverse fewer decisions (5% vs. 16%), and produce reasoning more predictive of their own rankings than STD-COT's, though less predictive than the base model's. The recipe reuses a fixed teacher corpus, requires no student rollouts or iterative teacher feedback, and applies to both pointwise and listwise formulations. Code and Data are at https://github.com/AdarshSingh7647/Eklavya [COMMENTS]8 pages, 3 figures [LINK]http://arxiv.org/abs/2607.25182v2 [DATE]2026-10-16 17:04:49+08:00 [CATEGORIES]cs.CL
Cross-Centaur: Modeling Individual Decision-Making Across Psychological Tasks with Language Models [AUTHORS]Hikaru Asano, Hidezo Suganuma, Ryutaro Mori, Yukino Baba [ABSTRACT]Large language models (LLMs) are increasingly used as computational models of human behavior. Recent models such as Centaur have advanced this goal by fine-tuning LLMs on trial-by-trial decisions from psychological experiments. However, existing large-scale behavioral datasets do not link the same participants across different tasks, so these models learn population-level regularities but are not trained to model individual-level regularities across tasks. We introduce Cross-Psych-101, a dataset of 591 participants and 261\{,\}551 valid decisions across 12 Bandit, Tradeoff, Structure, and Social tasks. We train Cross-Centaur to predict a participant's target-task actions from a target-task description and the source-task description paired with the same participant's source-task behavior. Cross-Centaur outperforms off-the-shelf LLMs and Centaur on held-out participants, both within and across tasks. Probing analyses suggest that its hidden states encode participant-level behavioral tendencies that transfer across tasks. These results highlight the potential of cross-task individual-level data for predicting human decision-making with language models. Cross-Pysch-101 is available at https://huggingface.co/datasets/HikaruAsano/cross-psych-101. [COMMENTS]Accepted to EMNLP 2026 (Main Conference, Oral Presentation) [LINK]http://arxiv.org/abs/2610.23157v1 [DATE]2026-10-16 16:55:00+08:00 [CATEGORIES]cs.CL
Can AI interview? Comparing AI-led to Human-led Chat Interviews [AUTHORS]Semra Guven, Tobias Gårdhus, Andreas Bjerre-Nielsen, Anna Rogers, Hjalmar Bang Carlsen [ABSTRACT]Large language model chatbots have been proposed as AI-interviewers for collecting qualitative interview data at scale, potentially bridging social science's longstanding divide between the breadth of surveys and the depth of qualitative interviews. This paper provides the first direct empirical comparison of data quality in AI-led and human-led interviews. In a randomized experiment (N = 41), university students were interviewed via synchronous text chat by either trained non-expert interviewers or a locally hosted AI system; we evaluate exposure, specificity, and relevance. Median interview duration was similar across conditions, but AI-interviewers asked more than twice as many questions per interview and elicited roughly 65 percent more respondent text per interview. Per-question responses were longer with human interviewers, though not after multiple-testing correction; only AI-condition participants skipped questions. Responses were almost uniformly on-topic in both conditions; for specificity we detect no difference, but the design cannot exclude one of up to three quarters of a point. These results were obtained with a small, locally hosted open-weight model (7B parameters), which addresses concerns about data security, GDPR compliance, and scientific reproducibility. Our findings support the viability of AI-interviewing for large-N qualitative data collection. [COMMENTS]37 pages, 3 figures, 6 tables. Submitted to Socius. Code and data: https://github.com/ainterviewer/can-ai-interview [LINK]http://arxiv.org/abs/2610.23149v1 [DATE]2026-10-16 16:43:18+08:00 [CATEGORIES]cs.CL
TriRouter: Sparse-to-Dense Electrode-Space Routing for Robust EEG Motor Imagery Decoding [AUTHORS]Ying Zhou, Yaoxin, Mengting Liu [ABSTRACT]Motor imagery brain-computer interfaces require reliable decoding from multichannel EEG, but practical acquisition often yields sparse or corrupted electrode montages. We propose TriRouter, an electrode-space routing framework with discriminative-preservation, mask-robustness, and latent-compensation paths. A validation-calibrated router uses observable channel availability to allocate path contributions, preserving complete-channel discrimination while increasing representation recovery under severe missingness. In BNCI2014-001, TriRouter outperforms standard public backbones and a dynamic-spatial-filtering corrupted-EEG baseline. It improves accuracy over the strongest corrupted-EEG baseline by 3.68, 4.02, and 6.46 percentage points at nominal 25%, 50%, and 75% missing settings, and remains effective under structured regional missingness and additional low-channel dataset evaluation. [COMMENTS]Submitted to IEEE ICASSP 2027 [LINK]http://arxiv.org/abs/2610.23665v1 [DATE]2026-10-16 23:59:00+08:00 [CATEGORIES]cs.LG
JAXtari: High-Throughput and Easy-to-Modify Arcade Learning Environment [AUTHORS]Quentin Delfosse, Raban Emunds, Paul Seitz, Sebastian Wette, Jannis Blüml, Daniel Kirn, Dominik Mandok, Kristian Kersting [ABSTRACT]The Arcade Learning Environment (ALE) has been a cornerstone of Deep Reinforcement Learning research, providing a standardized benchmark for evaluating general agents. However, its lack of native support for object-extraction and environment modification restricts research into evaluating whether agents learn robust causal models or rely on brittle visual heuristics. Furthermore, the ALE's dependence on CPU-based emulation creates a significant bottleneck in modern GPU-accelerated training pipelines, limiting both experimental scale and iteration frequency. In this work, we introduce JAXtari, an open-source, high-performance reimplementation of Atari 2600 environments. By leveraging massive GPU parallelization, JAXtari allows PPO agents to finish training on 100M environment steps on a single A100-GPU in under an hour. Beyond speed, it provides object-centric observations and a flexible modification system, allowing researchers to, e.g. develop neuro-symbolic methods or discover novel shortcuts and subsequently train and adapt their agents rapidly, sometimes in only minutes. [LINK]http://arxiv.org/abs/2610.23650v1 [DATE]2026-10-16 23:48:00+08:00 [CATEGORIES]cs.LG
Scalable long-range machine learning interatomic potentials with electrical response for atomistic material modeling [AUTHORS]Yicheng Chen, Xiaoyun Yu, Hyejeong Song, Xintian Wang, Zihui Li, Zixi Xia, Junmin Chen, Peichen Zhong [ABSTRACT]Accurate atomistic simulations of materials under electric fields require long-range electrostatics and electrical-response modeling at length scales beyond the reach of conventional first-principles methods. Yet these capabilities remain difficult to reconcile with the scalability of machine-learning interatomic potentials (MLIPs). Strictly local models enable efficient spatial parallelization but neglect long-range interactions, whereas existing electrostatic extensions introduce global dependencies that hinder distributed computation. Here, we overcome this limitation by analytically separating the long-range electrostatic forces and Born effective charges (BECs) into a global electrostatics calculation and strictly local derivatives of learned latent charges. This formulation preserves the spatial decomposability of equivariant MLIPs while enabling distributed evaluation of energies, forces, stresses, and BECs across different boundary conditions. Implemented as DPA4C-LR on the \texttt\{nvalchemi\} platform, the framework scales to 5.2 million atoms on four GPUs at approximately twice the computational cost of its short-range backbone. Trained on energies and forces, DPA4C-LR accurately infers BECs and reproduces the field-dependent infrared spectra of water. We further demonstrate its capabilities by recovering the characteristic slim polarization hysteresis of relaxor \ce\{Pb(Mg_\{1/3\}Nb_\{2/3\})O3\}, which is obscured by finite-size effects in smaller simulations; predicting a room-temperature field-driven ionic conductivity of $6.8 \pm 0.7$~mS/cm in amorphous \ce\{Li_\{0.9\}NbO_\{0.9\}Cl_\{4.1\}\}; and modeling interfacial thermal transport across Si/diamond interfaces with slab-corrected electrostatics. By combining long-range physics, electrical-response inference, and distributed computation, this framework extends predictive atomistic modeling toward device-relevant length scales. [LINK]http://arxiv.org/abs/2610.23643v1 [DATE]2026-10-16 23:44:07+08:00 [CATEGORIES]cs.LG
No-Regret Mixing of LRU and LFU with Optimal Switching Cost [AUTHORS]Younes Ben Mazziane, Xinying Zou [ABSTRACT]Caching systems often rely on simple eviction policies such as Least Recently Used (LRU) and Least Frequently Used (LFU), which perform well in complementary request regimes. Recent policies such as LeCar and Cacheus combine LRU and LFU using ideas from the experts problem in online learning. Specifically, upon a miss, they randomize between the two eviction rules using probabilities derived from scores updated by tracking the history of past evictions. While these policies exhibit strong empirical performance, it remains unclear whether they are guaranteed, on every request sequence, to perform asymptotically as well as the better of LRU and LFU, i.e., whether they achieve sublinear regret with respect to this benchmark. We first show that LeCar suffers linear regret against an oblivious adversary, even with unbounded history. We then propose H-MC, a Hedge-based mixture of virtual LRU and LFU caches that preserves Hedge's selection probabilities, and hence its regret guarantees, while minimizing the switching cost among all joint selection rules with these marginals. [LINK]http://arxiv.org/abs/2609.07566v3 [DATE]2026-10-16 23:42:39+08:00 [CATEGORIES]cs.LG
Deep LLM layers expose passage-level context in EEG during silent reading [AUTHORS]Ingo Marquardt, Anthilia Alchanat, Priyanka Jain [ABSTRACT]In a previous report we showed that a decoder can recover open-vocabulary word-level information from 19-channel EEG recorded during silent reading, by aligning short EEG windows with hidden-state embeddings of the presented word taken from a large language model (LLM). That analysis contrasted two of the model's 33 layers: the context-free input embedding and one mid-depth layer. Here we ask how EEG-decodable structure is distributed over the full depth of the language model. We trained one decoder per layer of Llama-3.1-8B, with and without a causal transformer over the sequence of EEG windows. We distinguish word-level decodability from a context-tracking component. Word-level decodability peaks in early layers and declines steadily towards the last layer. In contrast, the context-tracking component of the retrieval gain rises with depth, both with and without a causal transformer. Unexpectedly, a decoder that sees a single 0.5-s window at a time tracks the text passage too. Two new controls dissect this component. A target-geometry control, a synthetic predictor made of the true target plus noise calibrated to each decoder's word-level accuracy, shows that context tracking is not a by-product of word decoding and target geometry. A control that permutes predictions across runs from either the same or from different sessions shows that most of the observed context-tracking is not caused by an EEG recording session signature. Instead, context tracking mostly follows the text. A single EEG window therefore carries passage-level information that a contextual target can absorb. The results locate word identity in the early layers of the language model, discourse state in the deeper layers, and make the choice of target layer a design dial for EEG language decoders. [COMMENTS]41 pages (including 23 pages of Supplementary Material), 12 figures (including 5 in Supplementary Material) [LINK]http://arxiv.org/abs/2610.23638v1 [DATE]2026-10-16 23:41:19+08:00 [CATEGORIES]cs.LG
Deconstructing Multi-Task Active Learning: The Paradox of Gradient Conflict and Orthogonal Decomposition [AUTHORS]Juho Jung, Seongjae Kang, Jinyoung Han [ABSTRACT]Existing multi-task active learning (MTL-AL) methods predominantly rely on linear scalarization of heterogeneous objectives such as uncertainty, diversity, and gradient conflict. We argue that this entangled formulation is fundamentally limited, as linear combination fails to capture Pareto-optimal trade-offs under conflicting objectives. We propose a shift from combination to deconstruction and introduce a principled framework that rethinks MTL-AL along three orthogonal dimensions. First, we identify the Conflict Paradox: although the optimization literature treats gradient conflict as interference to be suppressed, penalizing conflict when ranking candidates is measurably worse than not scoring it at all. We therefore decouple the two concerns: the acquisition function does not score conflict, and the resulting interference is resolved by gradient surgery during training. Second, we introduce Orthogonal Decomposition, separating acquisition into Informativeness and Exploration. This two-stage process, informativeness filtering followed by coverage maximization, prevents "diversity dilution" by excluding uninformative outliers before enforcing diversity. Third, we propose Cross-Task Mutual Information (CTMI), a covariance-based statistic of cross-task dependency computed from MC-dropout task entropies. Extensive experiments on COCO, Cityscapes, and NYUv2, including comparisons against recent MTL-AL and data-selection baselines, demonstrate consistent improvements over state-of-the-art methods, highlighting that structural rethinking of acquisition yields larger gains than incremental heuristic design. Code is available at https://github.com/Juho-Jung/multitask_active_learning_official. [COMMENTS]10 pages, 5 figures, NeurIPS 2026 [LINK]http://arxiv.org/abs/2610.23623v1 [DATE]2026-10-16 23:32:37+08:00 [CATEGORIES]cs.LG
GRAFT-EEG: Gauge-Aware Physical Field Transport for Cross-Dataset Motor-Imagery EEG [AUTHORS]Ying Zhou, Mengting Liu [ABSTRACT]Cross-dataset EEG decoding often treats montage, reference, and active-sensor changes as representation shift, although they also alter the observation operator. We study target-statistics-blind transfer: fitting receives no target signal, label, or empirical statistic, and electrode coordinates only instantiate a fixed prediction operator. method removes the active-set common-reference gauge and transports each native mask through a template motor leadfield. A source-fitted spectral classifier anchors the decision; a bounded state-space-model (SSM) residual is secondary. We establish restricted identifiability and an operator-perturbation bound. Across five leave-one-dataset-out targets (253 participants), dynamic method achieved 60.865% equal-target balanced accuracy versus 52.684% for strict-clean 5/5 BIOT (hierarchical difference +8.182 points; 95% CI +6.382 to +10.196; positive in every target). On four EEGPT-clean targets, the descriptive means were 61.520% and 59.057%. In four post-hoc exact same-parent tests, FULL exceeded coordinate-basis, matched-spectrum, row-shuffled, and wrong-anatomy controls by 9.959, 9.698, 11.241, and 3.985 points, respectively (all 95% CIs excluded zero; Holm-adjusted P<0.001). A separate 60-participant same-paradigm evaluation retained a positive physical-operator contrast (+4.029 points; +2.338 to +5.773), whereas dynamic--static was inconclusive; its preregistered status is partial confirmation rather than unrestricted external validation. The SSM added +1.059 points (+0.013 to +1.892). Claims remain bounded to five binary motor-imagery domains and the tested operators. [COMMENTS]Under review at ICLR 2027 [LINK]http://arxiv.org/abs/2610.23620v1 [DATE]2026-10-16 23:30:37+08:00 [CATEGORIES]cs.LG
ThetreStageAI: Multi-View Retrieval-Augmented Generation for Structured Bengali Theatre Production Planning [AUTHORS]Md. Mehedi Hasan Naeem, Mst. Kamrunnahar Ruma, Shakila Sultana, Nafiza Anjum, Nasreen Jannatun Nemu, Md. Sujan Ali [ABSTRACT]Staging a theatre performance requires coordinating dialogue, actor movement, stage directions, lighting focus, cue timing, and equipment constraints within the same production plan. This work presents ThetreStageAI, a retrieval augmented framework for structured Bengali theatre production planning. Rather than retrieving from one mixed corpus, the system maintains separate scene, actor-blocking, and lighting views. Bengali text is embedded with BGE-M3 and indexed in view-specific Qdrant collections. Retrieved references are combined with a production brief and passed to Gemini, which generates schema-constrained JSON containing characters, scenes, dialogue, stage directions, blocking, lighting, and optional sound cues. Structural and semantic checks are applied before acceptance, with at most one constrained repair attempt. We also construct a human-guided, LLM-assisted synthetic dataset of 500 short Bengali theatre records that expands to 1,500 retrieval documents. Dataset diagnostics confirm consistent identifiers and valid lighting fields while also revealing substantial dialogue repetition. A paired camera-ready evaluation compares single-combined and full multi-view retrieval under matched settings. The source code, dataset, and evaluation materials are publicly available. [COMMENTS]6 pages, 6 figures. Accepted at the 12th IEEE International Conference on Sustainable Technologies and Engineering (IEEE iCOSTE 2026), Cox's Bazar, Bangladesh, November 1-3, 2026 [LINK]http://arxiv.org/abs/2610.23618v1 [DATE]2026-10-16 23:29:20+08:00 [CATEGORIES]cs.LG
Activation-Efficient Fine-Tuning via Sequence-Compressed Low-Rank Subspaces [AUTHORS]Irene Tenison, Stella Ahn, Miriam Kim, Ebtisam Alshehri, Lalana Kagal [ABSTRACT]Parameter-Efficient Fine-Tuning (PEFT) has become the standard approach for adapting large language models (LLMs). However, despite reducing trainable parameter counts, many PEFT methods remain memory-bound by token-level activation intermediates that scale linearly with sequence length, limiting their practicality for long-context and on-device adaptation. Unlike prior memory-efficient PEFT approaches that preserve token-level adapter activations, LARS (Low-memory Activation-Rank Subspace) reduces the sequence-length scaling of stored adapter activation intermediates from O(BSRL) to O(BRL) by performing adaptation within a sequence-level low-rank subspace. Through pooled feature extraction, low-rank subspace modulations, and residual projection, LARS reduces the overall memory footprint by 23% on GPUs and 47.8% on CPUs relative to LoRA at S = 1024, with relative savings increasing with context length. Experiments across reasoning, understanding, long-context, retrieval, and token-level tasks show that LARS maintains competitive performance relative to LoRA while substantially improving memory efficiency. Beyond GPUs, LARS also reduces average CPU power and energy consumption per optimisation step by 22.43% and 22.06%, respectively. These results demonstrate that activation-efficient adaptation provides a practical path toward resource-constrained and on-device fine-tuning of LLMs. [COMMENTS]Accepted to AACL 2026 [LINK]http://arxiv.org/abs/2604.22783v2 [DATE]2026-10-16 23:29:14+08:00 [CATEGORIES]cs.LG
When Magnitude Suffices: Reconstructing Time-Domain-Gated Microwave Signals Without Phase on a Thoracic Phantom [AUTHORS]Maciej Cichoń, Patrycja Gołaś, Maksymilian Bartosik, Angelika Walczak, Agata Bielecka-Dąbrowa, Maciej Ślot [ABSTRACT]Time-domain gating separates the arrivals that form a microwave S-parameter measurement. It requires the measurement of complex values, so this application almost always uses a vector network analyzer (VNA). We show that, on one calibrated measurement configuration, the gated magnitude the VNA would give can be reconstructed from a magnitude-only sweep of a spectrum analyzer with a tracking generator (SA). We select the gate center for separability of the measured states, correct each instrument once by a per-bin subtraction of a reference state, and train a fully connected network on sweeps from both instruments. The test object is a torso phantom with lung-shaped compartments whose filling takes eleven dielectric states, each held out in turn. On $S_\{11\}$ the reconstruction from SA input reaches a shape error of 0.08 dB$^2$ with a state ordering of 0.81. This matches the reconstruction from VNA input. On $S_\{21\}$ it reaches 0.82 dB$^2$ and keeps the ordered groups of the states. An explicit phase prediction performs much worse, because gating needs the phase to about 0.2 rad RMS. Across eleven measurement configurations the dependence of phase on magnitude is strong for reflection and weak for transmission. This explains the difference between $S_\{11\}$ and $S_\{21\}$. Where the magnitude carries enough of the phase, a scalar instrument gives the gated response without any phase measurement after installation. On the open UM-BMID breast data set, with whole phantoms held out, the shape error of the network is 30 percent lower than that of a per-bin linear map. [COMMENTS]18 pages, 7 figures, 4 tables [LINK]http://arxiv.org/abs/2610.23608v1 [DATE]2026-10-16 23:22:35+08:00 [CATEGORIES]cs.LG
OrbiSim: World Models as Differentiable Physics Engines for Embodied Intelligence [AUTHORS]Jiajian Li, Jingyuan Huang, Junru Gong, Qi Wang, Xiaokang Yang, Yunbo Wang [ABSTRACT]We present OrbiSim, a novel robotic simulation paradigm that redefines world models as a fully differentiable physics engine for embodied intelligence. Unlike prior world models that focus on unconstrained imagination in latent or visual domains, OrbiSim establishes a unified, physically-grounded pathway that bridges structured scene assets, neural dynamics, and downstream reinforcement learning. By enabling end-to-end differentiability throughout the entire simulation loop -- spanning from explicit state transitions to visual observation generation -- OrbiSim supports tasks traditionally intractable for classical simulators, such as differentiable contact modeling, gradient-based policy optimization under sparse rewards, and intuitive physical inference. Empirical results demonstrate that OrbiSim significantly outperforms state-of-the-art world models in both predictive fidelity and control performance. Furthermore, its consistent responsiveness to asset configurations and physical parameters suggests its potential as a differentiable tool for enhancing robot simulation and policy training. [COMMENTS]Project page: https://jjleejj85.github.io/projects/orbisim [LINK]http://arxiv.org/abs/2605.16395v2 [DATE]2026-10-16 23:22:03+08:00 [CATEGORIES]cs.LG
Boundary Detection for High-Dimensional Point Clouds [AUTHORS]Abigail Hickok, Valerio Galanti, Osama Al Dalahmah, Andrew J. Blumberg [ABSTRACT]We introduce two new methods for boundary detection in high-dimensional point cloud data, intended to work in situations with non-uniform density, high ambient dimension, and substantial noise. Our first method fits a separating hyperplane on local neighborhoods after doing local dimension reduction---this method works best when one can reliably estimate the intrinsic dimension of the data. Our second method measures the directional uniformity of local neighborhoods, and has the best overall performance on synthetic data. In addition to pointwise numerical boundary scores, our algorithms yield an estimated normal vector at each boundary point. We validate our methodology on synthetic data as well as two real examples: we analyze spatial transcriptomics data, where we show that we can detect the boundary of a tumor, and we also study LLM activation geometry. [COMMENTS]31 pages; 23 figures [LINK]http://arxiv.org/abs/2610.23590v1 [DATE]2026-10-16 23:12:00+08:00 [CATEGORIES]cs.LG
Empirical Bayesian Multi-Bandit Learning [AUTHORS]Xia Jiang, Rong J. B. Zhu [ABSTRACT]Multi-task learning in contextual bandits has attracted significant research interest due to its potential to enhance decision-making across multiple related tasks by leveraging shared structures and task-specific heterogeneity. In this article, we propose a novel hierarchical Bayesian framework for learning in various bandit instances. This framework captures both the heterogeneity and the correlations among different bandit instances through a hierarchical Bayesian model, enabling effective information sharing while accommodating instance-specific variations. Unlike previous methods that overlook the learning of the covariance structure across bandits, we introduce an empirical Bayesian approach to estimate the covariance matrix of the prior distribution. This enhances both the practicality and flexibility of learning across multi-bandits. Building on this approach, we develop two efficient algorithms: ebmTS (Empirical Bayesian Multi-Bandit Thompson Sampling) and ebmUCB (Empirical Bayesian Multi-Bandit Upper Confidence Bound), both of which incorporate the estimated prior into the decision-making process. We provide the frequentist regret upper bounds for the proposed algorithms, thereby filling a research gap in the field of multi-bandit problems. Extensive experiments on both synthetic and real-world datasets demonstrate the superior performance of our algorithms, particularly in complex environments. Our methods achieve lower cumulative regret compared to existing techniques, highlighting their effectiveness in balancing exploration and exploitation across multi-bandits. [COMMENTS]33 pages, 13 figures [LINK]http://arxiv.org/abs/2510.26284v3 [DATE]2026-10-16 23:06:14+08:00 [CATEGORIES]cs.LG
SelectArena: A Modern Benchmark for Feature Selection on Real-world Tabular Data [AUTHORS]Bastian Schäfer, Dominika Matus, Lennart Purucker, Ricardo Knauer, Erik Rodner, Frank Hutter, Anton Sokolov, Arnaud Liefooghe [ABSTRACT]Selecting a subset of features (columns) from a tabular dataset is a ubiquitous challenge in real-world applications. Industry experts seek to minimize model training costs, climate scientists require an interpretable subset of factors to understand extreme weather events, and clinicians need to choose among diagnostic tests due to financial and time constraints. When looking at the scientific literature to determine which feature selection method to use, practitioners across domains face an unrepresentative landscape of benchmarking. Despite extensive research, existing feature selection benchmarks provide limited guidance because they rely on outdated models, ad hoc evaluation protocols, or artificially generated data that do not reflect real-world problems. To close the gap between scientific benchmarking and real-world applications, we introduce SelectArena, a modern benchmark for feature selection. We compare 15 widely used feature selection methods on 71 manually curated real-world tabular datasets, covering classification and regression problems, using standardized evaluation protocols and state-of-the-art tabular (foundation) models. Our results reveal clear trade-offs across predictive performance, runtime, stability, and validity. Tree-based methods, CART and RFImportance, along with F-Test are Pareto-optimal when taking into account predictive performance and runtime. F-Test, information-theory-, and distance-based methods lead on stability and validity, with F-Test emerging as a balanced choice across our metrics. Search-based methods are slow, unstable and achieve low performance. SelectArena turns these findings into practical guidance and, as an extensible platform, supports future feature selection research. [LINK]http://arxiv.org/abs/2610.13644v2 [DATE]2026-10-16 22:56:25+08:00 [CATEGORIES]cs.LG
BMuon: Budgeting Muon's Spectral Geometry for Structured Compressibility [AUTHORS]Rida Assalouh, Tom Jacobs, Rebekka Burkholz, Yani Ioannou, Umut Şimşekli [ABSTRACT]Muon has emerged as a strong optimizer for neural network training and admits an interpretation as steepest descent under a spectral-norm geometry. Yet we find that Muon-trained Transformers can degrade sharply when their MLP hidden channels are pruned after training. To reconcile Muon's optimization performance with structured compressibility, we introduce Budget Muon (BMuon), which modifies Muon's spectral geometry so that hidden channels across blocks compete for a shared allocation budget. The resulting oracle is exactly normalized steepest descent under a family of norms that interpolate between Muon's geometry and structured sparse geometries. Our principal sparsity control is an interpretable global channel-allocation budget, rather than a setting-dependent sparsity-penalty coefficient. For linear factorization models, we show that Muon admits minimum-cost exact representations with redundant hidden width, whereas under the BMuon oracle geometry and a sufficiently scarce budget, every minimum-cost representation is width-minimal. For practical training, we derive a minorization-maximization surrogate for the exact allocation problem. On ImageNet-1K with ViT-S/16, BMuon improves over Muon under gradual structured MLP pruning, and its learned allocations provide a more effective pruning criterion than weight magnitude. On a 1B-parameter LLaMA-style model trained on FineWeb, BMuon improves over Muon under post-training MLP channel pruning. [COMMENTS]54 pages, 6 figures [LINK]http://arxiv.org/abs/2610.23571v1 [DATE]2026-10-16 22:52:56+08:00 [CATEGORIES]cs.LG
Beyond Expected Values: Risk-Sensitive Planning with Distributional Monte-Carlo Tree Search [AUTHORS]Quynh Trinh, Trung Nguyen, Nam Nguyen, Tuan Dam [ABSTRACT]Monte-Carlo Tree Search (MCTS) typically stores only scalar mean estimates, which makes it poorly suited for tail-risk objectives such as conditional value-at-risk (CVaR). We introduce Categorical Thompson Sampling with Optimism (CATSO) and Particle Thompson Sampling with Optimism (PATSO), two distributional MCTS algorithms that maintain finite-support empirical backup laws at Q-edges. CATSO represents Q-edge laws with categorical atoms and Dirichlet Thompson sampling, while PATSO represents them with adaptive particles. Both algorithms select actions using the CVaR of a Thompson-sampled Q-edge law plus a polynomial optimism bonus, and propagate scalar continuation values through visit-weighted averages of child Q-edge CVaR scores. For the finite-depth nested CVaR planning target, we prove an $O(n^\{-1/2\})$ bound on the root value error after $n$ rollouts, which upper-bounds the risk-sensitive simple regret. CATSO adds a fixed-grid discretization bias, while capped PATSO adds a tunable Wasserstein compression bias. Experiments on stochastic planning benchmarks show that distributional Q-edges improve lower-tail return and reduce catastrophic outcomes when mean-optimal and risk-sensitive behavior differ. [COMMENTS]Proceedings of the 40th Annual Conference on Neural Information Processing Systems [LINK]http://arxiv.org/abs/2610.23565v1 [DATE]2026-10-16 22:51:28+08:00 [CATEGORIES]cs.LG
Can Graph Learning Learn Circuits? [AUTHORS]Chester Tan, Moritz Lampert, Courtney Maynard, Ankit Ramakrishnan, Tina Eliassi-Rad, Ingo Scholtes [ABSTRACT]Circuit localization is a mechanistic interpretability task whose goal is to identify a sparse subgraph of a transformer's computation graph sufficient to reproduce a particular behavior. Most established methods localize circuits independently for each model-task pair. We instead frame circuit localization as a graph machine learning problem in which the edges of a computation graph represent computational pathways, and graph neural networks (GNNs) model interactions among these pathways. We introduce Graph Circuit Learning (GCL), a supervised, amortized framework that trains a GNN across multiple model-task pairs and applies it to unseen cases. To provide sufficient data, we augment the InterpBench benchmark with additional cases derived from TracrBench programs. Of the 46 evaluated GCL configurations, the one selected by cross-validation scored a median edge AUROC of 0.946 (interquartile interval [0.841, 0.980]) on the 16 held-out cases of our evaluation split. Removing all message-passing edges reduces the median to 0.926. We also adapt PGExplainer, a GNN explainability method, to circuit localization, obtaining a median edge AUROC of 0.950 on the same cases. These preliminary results suggest that graph machine learning offers a natural and potentially powerful perspective on circuit localization, one we hope encourages closer exchange between the two communities. [COMMENTS]Oral presentation at the Fifth Learning on Graphs Conference (LoG 2026), Extended Abstract Track [LINK]http://arxiv.org/abs/2608.08536v2 [DATE]2026-10-16 22:46:47+08:00 [CATEGORIES]cs.LG
Beyond Latent Personalization: Integrating HRV-Conditioned Personalization via LLM for Interpretable Stress Detection [AUTHORS]Wei-Hsiang Ma, Ya-Tse Wu, Hsing-Hang Chou, Chi-Chun Lee [ABSTRACT]Stress detection based on heart rate variability (HRV) has received increasing attention. However, existing personalized approaches integrate personal attributes with HRV through latent representations, limiting the interpretability of their interactions. In this work, we propose an LLM-enhanced personalized stress detection framework that integrates HRV features and personal attributes through LLM-based semantic interpretation. The proposed framework achieves a 29.9% relative MCC improvement over the strongest baseline on TILES-2018 while providing human-readable descriptions of individual physiological responses. The results highlight the potential of LLM-based semantic interpretation for achieving both effective and interpretable personalized stress detection. [COMMENTS]Submitted to ICASSP 2027 [LINK]http://arxiv.org/abs/2610.23555v1 [DATE]2026-10-16 22:44:02+08:00 [CATEGORIES]cs.LG
Causal Discovery beyond Conditional Independence: Identifiability, Algorithms, and Statistical Inference for Directed Acyclic and Cyclic Graphs [AUTHORS]Yang Ni [ABSTRACT]Learning causal structure from data requires assumptions about mechanisms, noises, observation schemes, and interventions. Historically, causal discovery has relied primarily on conditional independence. This review focuses on modern causal discovery methods that exploit properties of distributions beyond conditional independence. Our organizing principle separates the source of identification from the statistical estimator and its computational implementation. This perspective connects conditional-independence methods with distributional asymmetries, conditional-moment restrictions, and variation across environments or individuals. We examine discovery under latent confounding and feedback, distinguishing causal information about observed variables from reconstruction of latent structure and emphasizing the additional semantics required by cyclic models. We also distinguish population identification, frequentist and Bayesian consistency, uncertainty quantification, and algorithmic guarantees. Throughout, examples show why the same observational distribution can support different conclusions under different model classes. We conclude with guidance for comparing methods at their identifiable targets and a research agenda for uncertainty, heterogeneous populations, non-scalar variables, and the intersection of discrete data, feedback, and hidden confounding. [LINK]http://arxiv.org/abs/2610.23543v1 [DATE]2026-10-16 22:37:57+08:00 [CATEGORIES]cs.LG
BenthicDINO: Physics-Informed Self-Distillation for View-Invariant Side-Scan Sonar Representations [AUTHORS]Taqi Hamoda, Hayat Rajani, Nuno Gracias [ABSTRACT]Automated interpretation of side-scan sonar (SSS) imagery is traditionally hindered by view-dependence, as recorded intensities inextricably conflate intrinsic seabed reflectivity with transient acquisition geometries and acoustic artifacts. To address this, we introduce BenthicDINO, a physics-informed self-supervised learning framework that decouples acquisition dynamics from seabed representations. Utilizing a ConvNeXt-v2 backbone, physics-guided augmentations, and an HSIC penalty, BenthicDINO explicitly enforces feature invariance to physical viewing parameters like slant range and incidence angle. Evaluations demonstrate exceptional data efficiency - achieving near-peak performance in benthic habitat mapping with only 10% of available labels - and strong cross-domain generalization, outperforming fully supervised baselines on the AI4Shipwrecks benchmark using frozen representations. [LINK]http://arxiv.org/abs/2608.23215v2 [DATE]2026-10-16 22:33:36+08:00 [CATEGORIES]cs.LG
Out of Support, Not Out of Reach: In-Context Learning with Unsupported Instructions and Composed Tasks [AUTHORS]Xuhan Tong, Yuchen Zeng, Jiawei Zhang [ABSTRACT]In-Context Learning (ICL) enables pretrained LLMs to adapt to downstream tasks from a small set of input-output demonstrations, without parameter updates. Yet existing theory primarily explains ICL as identifying a latent task already represented in pretraining, leaving unclear whether ICL can succeed when either the instruction or the target task falls outside the pretraining support. We develop a theoretical framework under mild assumptions that separates these two forms of distribution shift. For target tasks represented in pretraining, we show that demonstrations can identify the task and reduce sensitivity to pretrain-supported prompt templates as their number increases. For an unsupported instruction, we show that under a fixed model-induced reference extension, demonstration evidence can dominate an $N$-independent instruction shift while preserving the exponential identification rate. We then use the same ICL framework as an interpretive lens for long-horizon target tasks that are not represented in pretraining but admit a plan that decomposes them into pretraining-supported subtasks. Applying the supported-task analysis step by step yields a standard accumulated-error bound and clarifies how invoking familiar subtasks along an intermediate-state trajectory can avoid directly selecting the unsupported complete task. Together, our results characterize when demonstrations, instructions, and task decomposition can support ICL beyond the corresponding pretraining supports. [LINK]http://arxiv.org/abs/2610.23528v1 [DATE]2026-10-16 22:27:05+08:00 [CATEGORIES]cs.LG
Interpreting SSL Representations for Spoof Detection: a WavLM Study [AUTHORS]Mohamed Mallat, Michele Panariello, Massimiliano Todisco, Nicholas Evans, Anthony Larcher [ABSTRACT]Self-supervised speech models such as WavLM achieve strong spoofing detection performance, yet which layers encode the most spoof-relevant information and what acoustic properties they capture remain opaque. We present a layer-wise interpretability study of WavLM using the ASVspoof 5 database, combining three complementary analyses: (i) per-layer linear probing with attack-level evaluation across speech synthesis, voice conversion, and adversarial attacks; (ii) a canonical correlation analysis (CCA) linking layer representations to spoofing-relevant acoustic features; (iii) an exhaustive grid search over layer subsets showing that use of five layers outperforms full thirteen-layer pooling. CCA analysis reveals that, while most acoustic features peak in alignment with mid-layers and then decay, voice quality measures jitter, shimmer, and the harmonics-to-noise ratio show the opposite trend, with correlation increasing toward later layers. Embedding visualisations reveal that deeper layers capture a continuous acoustic similarity among attacks rather than discrete categories, suggesting these layers encode fundamental synthesis artifacts rather than attack-specific signatures. [COMMENTS]5 pages, 4 figures, 1 table. Published at IWBF 2026. An extended abstract of this work was presented as a poster at Odyssey 2026 [LINK]http://arxiv.org/abs/2610.23526v1 [DATE]2026-10-16 22:26:55+08:00 [CATEGORIES]cs.LG
Leaner Transformers Can Easily Learn to Cluster [AUTHORS]Charlotte Park, Kenneth L. Clarkson, Lior Horesh, Takuya Ito, Parikshit Ram [ABSTRACT]Transformers have in-context learning capabilities, where some known learning algorithms can be executed in the forward pass through the model. Recent work shows that transformers can exactly perform Lloyd's algorithm for $k$-means clustering with $n$ points in $d$ dimensions with an embedding size $d_\{\textsf\{emb\}\} = d+k$ (thus, requiring attention projection matrices of size $(d+k)^2$). In this work, we build upon this result in the following ways: First, we present an equally expressive but smaller transformer that executes Lloyd's algorithm with embedding size $d_\{\textsf\{emb\}\} = (d + \lceil \log_2 k \rceil)$. Next, we train these transformers to learn the clustering algorithms given a distribution of clustering tasks, and theoretically characterize and empirically validate the factors affecting the convergence and in-distribution generalization of learning algorithms based on stochastic gradients. Finally, we probe the general clustering abilities of these learned algorithms (in the form of transformers), and try to understand situations where they succeed and fail. [COMMENTS]NeurIPS 2026 accepted paper [LINK]http://arxiv.org/abs/2610.09760v2 [DATE]2026-10-16 22:22:28+08:00 [CATEGORIES]cs.LG
Adaptive conformal prediction on the Grassmannian [AUTHORS]Dimitrios Loukrezis, Zachary J. Grey, Dimitris G. Giovanis [ABSTRACT]We introduce a conformal prediction framework for regression with Grassmann-valued responses. The proposed method constructs prediction regions in the form of metric balls directly on the Grassmann manifold, using basis-invariant geometric distances as non-conformity measures and scaling the corresponding non-conformity scores through a learned prediction difficulty estimator. This yields locally adaptive, input-dependent prediction regions that respect the underlying manifold geometry, while retaining the finite-sample marginal coverage guarantee of split conformal prediction under exchangeability. We evaluate several Grassmannian distance metrics and compare their prediction-region efficiency on synthetic benchmarks and an airfoil inverse shape-design problem. The results highlight the advantages of geometry-aware non-conformity scores, while local adaptation improves empirical coverage uniformity across prediction difficulty levels and, in strongly heteroscedastic settings, can substantially reduce prediction-region size while maintaining the target coverage. [COMMENTS]21 pages, 6 figures, 3 tables [LINK]http://arxiv.org/abs/2610.23508v1 [DATE]2026-10-16 22:18:06+08:00 [CATEGORIES]cs.LG
Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification [AUTHORS]Brown Zaz, Mar Gonzàlez I Català, Ferran Hernandez Caralt, Moshe Eliasof, Pietro Liò [ABSTRACT]In the transductive setting, where the full graph is observed but node labels are only partially available, progress in semi-supervised node classification has largely focused on architectural innovation. In this paper, we revisit an orthogonal axis: the training objective. We start from a simple observation: transductive models produce predictions for every node during training, including nodes without labels. These unlabeled-node predictions may contain useful training signal, but standard supervised objectives discard them because no ground-truth labels are available. Inspired by the decomposition of cross-entropy into a label-dependent alignment term and a label-independent entropy term, we propose prediction confidence as a natural way to extract this signal in the absence of labels. This motivates Transductive Sharpening (TS): a loss-level modification that minimizes prediction entropy on unlabeled nodes while counterbalancing this effect on labeled nodes. We evaluate Transductive Sharpening across a wide range of node-classification benchmarks and observe consistent performance improvements without requiring any changes to the backbone architecture. Code is available at https://github.com/transductive-sharpening/tunedGNN. [COMMENTS]19 pages, 3 figures, 20 tables [LINK]http://arxiv.org/abs/2605.20248v2 [DATE]2026-10-16 22:15:12+08:00 [CATEGORIES]cs.LG
Time-partitioning Gaussian Process Mixture Model [AUTHORS]Guilherme Piantino, Hedibert F. Lopes [ABSTRACT]Nonstationarity and heterogeneous dynamics are common in financial and macroeconomic time series. Volatility alternates between calm and turbulent regimes, interest rates follow monetary-policy cycles, and major events trigger persistent structural shifts. We propose the Time-partitioning Gaussian Process Mixture Model (TPGPMM), a Bayesian nonparametric framework for regime identification and nonstationary prediction that infers such regimes and their within-regime dependence from data, without Markovian assumptions. It combines (i) an adaptive Pitman-Yor process prior on ordered compositions, mapped to block lengths with contiguity enforced by construction, and (ii) regime-specific Gaussian processes for temporal dependence, yielding a nonstationary predictive distribution. By inferring its discount parameter, the prior adapts to homogeneous and highly heterogeneous regime lengths. Unlike sequential or Markov-switching methods, it flags persistent structural shifts as change points rather than reacting to transient shocks. A split-merge Metropolis-within-Gibbs sampler with exact Gibbs updates of regime boundaries jointly learns change points and hyperparameters in batch, quantifying uncertainty over the partition and within-regime dynamics. With an RBF kernel and a fixed, data-informed noise level, TPGPMM produces a sparse, confident segmentation. On simulated data it recovers the true regimes, and on the CBOE Volatility Index and the 10-year US Treasury yield it identifies regimes aligned with documented market and monetary-policy events. For prediction, variants estimating the noise level fit better than change-point baselines and fit and forecast better than the main specification, at the cost of a less sparse or less event-aligned segmentation; with a regime mean, they fit better than a fitted stationary Gaussian process and have lower forecast RMSE except at the shortest horizons. [LINK]http://arxiv.org/abs/2610.20860v2 [DATE]2026-10-16 22:12:42+08:00 [CATEGORIES]cs.LG
Cross-Dialect Generalization Without Retraining: Benchmarks and Evaluation of Schema-Derived Constrained Decoding for MLIR [AUTHORS]Plawan Kumar Rath [ABSTRACT]Multi-Level Intermediate Representation (MLIR) underlies modern ML compiler infrastructure (TensorFlow, JAX/StableHLO, PyTorch Inductor, IREE), yet appears only in trace amounts in code-LM pretraining corpora. MLIR is also extensible by design: new dialects ship per application domain, so a fine-tuned model per dialect does not scale. We ask whether inference-time priors derived mechanically from each dialect's Operation Definition Specification (ODS) can substitute for gradient-based adaptation. First, we release four natural-language-to-MLIR benchmarks across three dialects (MLIR-Spec-150, Linalg-Spec-30, StableHLO-Spec-30, StableHLO-Held-Out-200), totaling 410 in-scope NL-to-MLIR pairs, plus a 25-program out-of-grammar stress set, with functional references on 30 Spec prompts, under Apache-2.0 with Gebru datasheets and Croissant 1.1 metadata. Second, we build a three-layer schema-derived constraint stack: a CFG over op signatures (C1), type-domain splits from an ODS-extracted type lattice (C2), and an SSA-scope validator driving five-retry rejection sampling (C3). Porting it to StableHLO required no new constraint-layer code. On structurally dominated dialects, SmolLM2-1.7B with token-level constraints matches or exceeds 15B-34B code LMs running post-hoc rejection at a matched retry budget and 8-25x the per-generation speed: on linalg it reaches 80.0% verify-valid (three runs, n=125), 21 to 44 pp above CodeLlama-34B, Granite-Code-34B, and StarCoder2-15B with non-overlapping CIs. The comparison is between systems, not models: under the baselines' post-hoc protocol, SmolLM2 falls from 52.0% to 24.5% on arith+func. On arith+func and the templated StableHLO-Held-Out-200 the baselines match or beat the SLM; we scope these as non-win cells. We release benchmarks, decoder, every per-prompt generation, and a pinned reproducibility Dockerfile. [COMMENTS]Accepted to the NeurIPS 2026 Evaluations and Datasets Track; camera-ready version. 31 pages (10 main + references + appendix + checklist). Code: https://github.com/plawanrath/slm-to-mlir-constrained-emitter Datasets: https://huggingface.co/collections/plawanrath/cross-dialect-mlir-benchmarks-neurips-2026-e-and-d [LINK]http://arxiv.org/abs/2607.18254v2 [DATE]2026-10-16 22:07:14+08:00 [CATEGORIES]cs.LG
A Topological Signature of Learning under Noisy Labels [AUTHORS]Zhongtian Sun, Fan Mo, Prayag Tiwari, Kelin Xia [ABSTRACT]How does noisy supervision change the representations learned by a neural network? Training accuracy measures agreement with training labels but provides little information about representation geometry. We use persistent homology to track loop structure across distance thresholds in neural activations. Experiments span image classification, text classification, and modular arithmetic. On a fixed CIFAR-10/ResNet-18 training probe, equally sized clean-label and corrupted-label subsets show no detectable persistence difference at epoch 10, whereas corrupted-label subsets have higher persistence at epoch 50. The contrast survives uniform-scale controls, even though predictions match the assigned noisy labels for only 7-11% of corrupted probe examples at that checkpoint. Normalisation preserves the crossover between clean and noisy training trajectories on CIFAR-10/ResNet-18, whereas CIFAR-100/ResNet-18 shows no crossover under any of the five evaluated constructions. In a separate noise-rate calibration comparison across five pipelines, mean activation norm achieves a higher mean in-sample R2 than raw persistence (0.936 versus 0.908). These findings show that noisy supervision leaves a measurable geometric signature even when agreement with corrupted labels is low. Comparing persistence under scale controls with label-fitting measurements offers a way to study how representations respond to label corruption beyond what training accuracy reveals. [LINK]http://arxiv.org/abs/2610.23491v1 [DATE]2026-10-16 22:04:35+08:00 [CATEGORIES]cs.LG
QD-Rec: Open-Ended Evolution of Recommendations via Quality-Diversity [AUTHORS]Takashi Ishida, Taro Makino, Issa Sugiura [ABSTRACT]Open-ended search combined with foundation models has produced advances such as faster algorithms, new results on open mathematical problems, and self-improving agents. We extend this paradigm to recommendation for decision-making, focusing on decisions where the set of possible choices is fixed but the reasons for choosing them are not. In high-stakes decisions, e.g., which company to acquire or which customer to target, comparing diverse, well-reasoned options can be more useful than a single ranked list. We formalize open-ended recommendation as building a diverse, high-quality set of recommendations, each an item together with a hypothesis, a natural-language reason why the item fits the needs of a decision maker. We propose QD-Rec, a quality--diversity method that selects items with a contextual bandit and generates hypotheses with a foundation model from earlier ones kept in an archive as stepping stones. A new recommendation enters the archive if it is novel or better than its nearest neighbor. Across M&A, customer targeting, and paper and book recommendation, QD-Rec gives the highest mean quality and builds the largest archive in most settings. [LINK]http://arxiv.org/abs/2610.23470v1 [DATE]2026-10-16 21:49:22+08:00 [CATEGORIES]cs.LG
Graphlet-based Decentralized Finance Transaction Method Classification [AUTHORS]Célestin Coquidé, Alessio Brini, Natkamon Tovanich [ABSTRACT]Decentralized Finance (DeFi) transactions involve complex interactions among accounts, smart contracts, and tokens. We investigate whether network topology can identify transaction methods that describe DeFi operations. We extract directed graphlet features to classify the 30 most frequent methods. Our results show that graphlet features complement event log counts in improving classification performance. Using SHapley Additive exPlanations (SHAP), we further derive model explanations that reveal method-specific graphlet structural patterns. [COMMENTS]24 pages, 5 figures, 5 tables, 25 references 15th International Conference on Complex Networks and Their Applications [LINK]http://arxiv.org/abs/2610.23466v1 [DATE]2026-10-16 21:46:37+08:00 [CATEGORIES]cs.LG
RamanBench: A Large-Scale Benchmark for Machine Learning on Raman Spectroscopy [AUTHORS]Mario Koddenbrock, Christoph Lange, Robin Legner, Martin Jäger, Martin Kögler, Mariano N. Cruz Bournazou, Peter Neubauer, Felix Biessmann, Erik Rodner [ABSTRACT]Machine learning (ML) has transformed many scientific fields, yet key applications still lack standardized benchmarks. Raman spectroscopy, a widely used technique for non-invasive molecular analysis, is one such field where progress is limited by fragmented datasets, inconsistent evaluation, and models that fail to capture the structure of spectral data. We introduce RamanBench, the first large-scale, fully reproducible benchmark for ML on Raman spectroscopy, consisting of streamlined data access (https://pypi.org/project/raman-data/), evaluation protocols and code (https://pypi.org/project/raman-bench/), as well as a live leaderboard (https://huggingface.co/spaces/HTW-KI-Werkstatt/RamanBench). It unifies 77 datasets (including 17 first released with this benchmark) across four domains, comprising 325,862 spectra and spanning classification and regression tasks under diverse experimental conditions. We benchmark 28 models under a standardized protocol, including classical methods (e.g., PLS), Raman-specific (e.g., RamanNet), tabular foundation models (TFMs; e.g., TabPFN), and time-series approaches (e.g., ROCKET). TFMs consistently outperform domain-specific and gradient boosting baselines, while time-series models remain competitive. However, no single method dominates across these heterogeneous datasets, revealing that Raman spectroscopy remains an open challenge for machine learning. Therefore, we invite the community to contribute new approaches to our living benchmark, with the potential to accelerate advances in critical applications such as medical diagnostics, biological research, and materials science. [COMMENTS]https://huggingface.co/spaces/HTW-KI-Werkstatt/RamanBench [LINK]http://arxiv.org/abs/2605.02003v4 [DATE]2026-10-16 21:36:03+08:00 [CATEGORIES]cs.LG
Predictable Scaling: Efficient Model Exploration via Scaling-Law Prediction [AUTHORS]Aishwarya Wesanekar, Sheng-Tao Yang, Jiawei Zhang, Flavio Sales Truzzi, Chen Chen, Nader Beigiparast, Shengyuan Huang, Roshanak Zilouchian Moghaddam, Vitor Cid, Colin Clement, Vinayak Thapliyal, Sebastian Ament, Jihao Andreas Lin, Vincent Huang, Kai Han, Boyi Jiang, Wenlin Chen, Chaorong Chen, Ashwin Kumar, Maximilian Balandat, David Eriksson, Yifei Zhang, Pan Zhong [ABSTRACT]Evaluating candidate changes (hypotheses) to recommendation models requires full-scale training runs consuming thousands of GPU-hours, severely limiting the hypothesis space engineers can feasibly explore. We introduce Predictable Scaling (PS), a systematic framework that exploits empirically verified regularity conditions to predict full-scale model performance from small-scale proxies. PS constructs scaled-down variants via composable scaling functions along distinct resource axes, and projects metrics to the target scale via curve fitting with calibrated uncertainty estimates. We compare three prediction strategies of increasing expressiveness: even the simplest (raw proxy) achieves prediction error within the empirical noise floor of repeated-run variance; the most expressive (Gaussian Process) provides an indicative 15% relative MAE improvement and calibrated uncertainty. Evaluation on two ranking models of increasing complexity---a task-specific model and a foundation model---demonstrates that PS correctly ranks actionable hypotheses across diverse hypothesis types at 79-88% compute reductions. [LINK]http://arxiv.org/abs/2610.23453v1 [DATE]2026-10-16 21:35:54+08:00 [CATEGORIES]cs.LG
Regularized free deconvolution for spectral repair of resampling distortion in high-dimensional particle filters [AUTHORS]Deming Song, Chi-Kin Lam, Kefeng Li, Hui Yu, Xilin Li, Huagang Lin [ABSTRACT]Resampling is the standard remedy for weight degeneracy in particle filters, but in high dimensions it distorts the spectrum of the ensemble covariance in a way that is computable from the importance weights. This paper studies the inverse problem: recovering the pre-resampling spectrum from the observed weights and the empirical spectrum alone, and using it without biasing the filter. We first show that the forward spectral law extends to random importance weights once one conditions on the weights together with the realized offspring counts, rather than on the weights alone. We then construct a regularized free-deconvolution estimator, prove its consistency with an explicit error budget at a fixed resolution scale, and certify model mismatch through a computable forward residual. The repaired spectrum enters the filter only as the preconditioner of a Metropolis-adjusted rejuvenation move, so that spectral error affects mixing efficiency but not the invariant target. For the full filtering cycle, we derive a geometrically summable error recursion under stable scalar dynamics and an explicit law for the cost of abstaining from repair. Pre-registered simulations show that the inversion reduces the spectral error by an order of magnitude, that the corrected preconditioner recovers about 90% of the oracle's sampling efficiency in an anisotropic regime, and that the cost of abstention depends on when abstentions occur rather than on how many. The repair recovers eigenvalue laws, not eigenvectors, and is certified only within an explicitly stated exogenous-weight regime. [COMMENTS]This is a preprint of an article submitted to Signal Processing (Elsevier). The implementation and experiment scripts are available at https://github.com/songming-teaching/spectral-repair-resampling. This is the second paper in a series on spectral analysis of particle filters. The first paper is available at arXiv:2610.16680 [LINK]http://arxiv.org/abs/2610.23448v1 [DATE]2026-10-16 21:30:52+08:00 [CATEGORIES]cs.LG
RECREG: Recovery-based Shielding for Regular Safety Properties in Reinforcement Learning [AUTHORS]Alexander W. Goodall, Francesco Belardinelli [ABSTRACT]In many real-world scenarios we need to consider formal requirements, such as safety and constraint compliance, to deploy Reinforcement Learning (RL) agents. In this paper we introduce RecReg, a safe RL framework capable of enforcing non-Markovian regular safety properties via recovery-based shielding. RecReg is a unified approach that supports different levels of a priori knowledge and safety guarantees. The paper primarily focuses on building the theoretical foundations for RecReg in the tabular setting, providing rigorous safety and convergence guarantees that are not present in related work. We provide an extensive set of empirical results using the MASA-Safe-RL library, and we outline an extension of RecReg to deep RL domains using SOTA model-based RL. Our empirical results demonstrate that RecReg achieves efficient learning and matches the theory in this paper. [LINK]http://arxiv.org/abs/2610.23445v1 [DATE]2026-10-16 21:28:34+08:00 [CATEGORIES]cs.LG
Dynamics of Stochastic Momentum with Sparse Updates in High Dimensions [AUTHORS]Katie Everett, Elliot Paquette [ABSTRACT]Existing theory of momentum assumes that gradients arrive at every parameter at a roughly constant rate, an assumption violated in practice by heavy-tailed data distributions and modern architectures. We theoretically analyze the dynamics of two tractable models of momentum under sparse updates: a least squares model with sparse inputs and a logistic regression model with a rare class. Both admit exact closed-form second-moment dynamics whose high-dimensional limits we characterize across three scaling exponents for sparsity, batch size, and momentum decay. The phase structure on both problems is governed by the ratio of two intrinsic timescales: a momentum retention timescale (how many active updates the buffer survives) and a learning timescale (how many active updates it takes to reduce the squared error). When learning is much slower than retention, the limit matches SGD; when learning is faster, the system is unstable; where the timescales coincide, we recover classical heavy-ball dynamics. The oscillatory dynamics occur at different momentum values for different token sparsity, creating a spectral conflict for global momentum across token frequencies. [LINK]http://arxiv.org/abs/2605.28961v2 [DATE]2026-10-16 21:22:14+08:00 [CATEGORIES]cs.LG
Generative Policy Improvement from Observational Data [AUTHORS]Jonas Schweisthal, Dennis Frauen, Stefan Feuerriegel, Emma Brunskill [ABSTRACT]We study how to improve decision policies from observational data to generate better actions when actions are complex, structured objects, such as text messages or images. This task is subject to two major challenges: (i) action-level overlap is often too weak for reliable off-policy evaluation and improvement, and (ii) generating new structured actions often requires expensive training or fine-tuning (e.g., of an LLM) and can be ill-defined due to information loss in action representations. To address both, we propose PolicyLift, a three-stage framework for generative policy improvement with structured actions from observational data. In Stage 1, PolicyLift learns a small set of discrete action codes that capture high-level strategies in the observed action space and are tailored to support policy improvement under limited overlap. In Stage 2, it learns a context-dependent policy over these action codes. At deployment time, Stage 3 "lifts" the learned code-level policy back to the original action space by generating actions from the historical behavior system in a code-conditioned way. We show theoretically that improving the policy over action codes guarantees improvement of the induced policy in the original action space. We demonstrate PolicyLift across experiments in synthetic and real-world text generation tasks with consistent policy improvements over the behavior policy. [COMMENTS]Preprint [LINK]http://arxiv.org/abs/2610.23436v1 [DATE]2026-10-16 21:20:43+08:00 [CATEGORIES]cs.LG
Bridging Compressed Communication and Local Steps via Spectral Method in Federated Learning [AUTHORS]Dmitriy Bystrov, George Chirkov, Daniil Medyakov, Dmitry Bylinkin, Aleksandr Beznosikov [ABSTRACT]Federated Learning (FL) enables collaborative training of complex models across large-scale networks of edge devices while keeping data localized. However, data heterogeneity and high communication overhead remain fundamental bottlenecks. Recently, the spectral optimizer Muon has demonstrated remarkable advantages over Adam-like methods in centralized training. While recent efforts have attempted to transition Muon to FL, naive integrations into traditional architectures fail short. Crucially, existing approaches do not support the the communication compression and local steps --- the two foundational techniques of efficient FL. To address this limitation, we introduce a novel spectral FL method that utilizes both local updates and compression. We establish rigorous convergence guarantees under mild assumptions, notably eliminating the restrictive bounded heterogeneity condition. Empirical evaluation on standard image classification benchmarks validates the efficiency and robustness of our approach. [LINK]http://arxiv.org/abs/2610.23434v1 [DATE]2026-10-16 21:19:03+08:00 [CATEGORIES]cs.LG
Residual Stochastic Concept Bottleneck Models [AUTHORS]Stephen Marcou, Mateo Espinosa Zarlenga, Sonia Laguna, Moritz Vandenhirtz, Julia E. Vogt, Mateja Jamnik [ABSTRACT]Concept-based models (CMs) learn human-interpretable intermediate representations that enable users to understand the model's reasoning and steer its predictions through test-time concept interventions. However, in realistic settings, their performance can be limited by concept incompleteness, where the training concepts fail to capture all task-relevant information. Adding an unconstrained residual pathway to the model's predictive process can overcome the consequences of incompleteness. However, this introduces an inscrutable, leak-prone prediction pathway that may weaken concept interventions, as these often leave residual variables unchanged. To address this limitation, we introduce Residual Stochastic Concept Bottleneck Models (Res-SCBMs), a family of concept-based models that jointly model concept and residual variables through a multivariate Gaussian distribution. Res-SCBMs learn a covariance structure that explicitly captures dependencies between the known concepts and learnt residual variables, allowing interventions to propagate to statistically dependent residual variables through Gaussian conditioning. This extends the intervention capabilities of previous CMs without requiring concept completeness during training. Our results show that, across multiple concept-incomplete benchmarks, Res-SCBMs improve predictive performance over competing baselines while retaining effective test-time concept interventions. [LINK]http://arxiv.org/abs/2610.23421v1 [DATE]2026-10-16 21:09:16+08:00 [CATEGORIES]cs.LG