Daily Papers

The project automatically fetches the latest papers from arXiv based on keywords.

The subheadings in the README file represent the search keywords.

Only the most recent articles for each keyword are retained, up to a maximum of 10 papers.

You can click the 'Watch' button to receive daily email notifications.

Last update: 2026-10-26

LLM Reasoning

Title	Date	Abstract	Comment
When Should LLMs Reason? Adaptive Inference Routing via Entropy Dynamics	2026-10-22	Show Chain-of-thought (CoT) reasoning can improve LLM performance, but its benefits vary across tasks, models, and individual queries. On some tasks, CoT provides marginal or negative accuracy gains while substantially increasing token consumption. We investigate whether early decoding dynamics can help identify when explicit reasoning is useful. Our analysis shows that tasks with larger CoT gains tend to exhibit decreasing early-stage entropy, while low-gain tasks more often show oscillatory or increasing trajectories. These patterns motivate an empirical regime interpretation of decoding behavior. We propose \textbf{EDRM} (Entropy Dynamics-based Reasoning Manifold), a lightweight routing framework whose default heuristic requires no router training. EDRM summarizes early entropy trajectories with a compact, interpretable representation and selects among Direct, Standard, and CoT decoding. Across 15 benchmarks and 7 LLMs, its global and instance-level variants offer \textbf{16--55%} token savings at comparable accuracy. The Full instance-level configuration combines routing with a short Direct fallback branch; our heuristic-router ablation reports up to \textbf{4.9 percentage points} of additional accuracy from oracle answer selection over routing without fallback. These results support selective reasoning based on early decoding signals and demonstrate the potential of entropy-based control for efficient LLM inference.
Tracing Hallucinations in LLM Reasoning for UML Class Diagram Generation: An Empirical Taxonomy and Mitigation Study	2026-10-22	Show Large Language Models (LLMs) can generate UML class diagrams from natural language requirements, but hallucinations may compromise the correctness and completeness of the resulting diagrams. Existing evaluations mostly assess the final diagram, without examining the intermediate traces of reasoning-capable LLMs, which can reveal where these hallucinations arise. We develop an empirical taxonomy of eleven hallucination sub-types organised into three dimensions: Prompt Faithfulness (deviations between requirements and reasoning), Code Faithfulness (discrepancies between reasoning and generated PlantUML code), and Reasoning Quality (problems within the reasoning). The taxonomy was derived from 60 manually annotated DeepSeek-R1 traces, applied to 120 further DeepSeek-R1 traces through a dual-LLM annotation pipeline, and examined on 20 DeepSeek-V4-Pro traces. Code Faithfulness was the most frequent dimension for DeepSeek-R1 in both datasets, whereas Prompt Faithfulness dominated for DeepSeek-V4-Pro, suggesting model-dependent failure patterns. In the manually annotated traces, reasoning-trace length showed no significant monotonic association with hallucination frequency, whereas requirement complexity was negatively associated with it, particularly for Prompt Faithfulness errors. In a mitigation experiment on 13 Golden UML ModelSet cases with three Qwen3-based models, adding Prompt Faithfulness and Reasoning Quality constraints to the generation prompt raised the pooled mean F1 score by 4.0 percentage points, whereas a post-hoc Code Faithfulness review provided no consistent further benefit. Relationship modelling remained the lowest-scoring element throughout. These results suggest that localising hallucinations within the generation process can inform targeted, prompt-based mitigation. Whether these findings generalise beyond the DeepSeek and Qwen3 model families remains to be tested.
Same Beliefs, Different Policies: Learning Reusable Representations for Early Stopping in LLM Reasoning	2026-10-21	Show Reasoning LLMs often continue reasoning after reaching a correct answer, wasting computation and sometimes replacing it with an incorrect one. Deciding when to stop is a metareasoning problem: the model must weigh what further reasoning could achieve against its cost, given what it can already answer. Direct utility training learns a stopping representation at a chosen token price; supporting new prices can require additional controller training. We introduce BELT (BELief-based Termination), which separates outcome prediction from stopping control. It learns beliefs about current and future answers from a frozen LLM's hidden states, then fits a linear stopping policy on the frozen features. The prediction targets are independent of token price, so one probe supports different accuracy-cost trade-offs through CPU policy refitting or threshold adjustment. Averaged over five out-of-distribution (OOD) math benchmarks, BELT reduces generated tokens by 41% on Qwen3-1.7B and 51% on OLMo-3-7B-Think without loss of accuracy. The same math-trained BELT transfers to coding and agentic tasks without training on code or agentic data. In all three domains, its accuracy-cost trade-offs are competitive with or better than those of end-to-end controllers. On OLMo3, its probe has 21.2x fewer parameters than one checkpoint of the strongest baseline, OS-Pruner, and needs 55x fewer training GPU-hours than OS-Pruner's five price-specific checkpoints.	Under Review
HiddenLens: A Multi-Layer LLM Framework for Detecting and Resolving Hidden Assumptions in Software Requirements	2026-10-21	Show Natural-language software requirements often omit assumptions about performance, operating conditions, user behaviour, and domain constraints, affecting architecture, testing, and deployment. If undiscovered until implementation, resolving them is far more costly than addressing them early. Although ambiguity and vagueness are widely studied, less attention has been given to assumptions embedded in requirements that appear clear and reasonable. We propose HiddenLens, a novel six-layer BERT-LLM pipeline and, to our knowledge, the first framework engineered to detect, classify, question, and resolve hidden assumptions rather than merely flag ambiguity. Layer 1 uses BERT attention entropy to localise unresolved spans; Layer 2 uses CoT-prompted LLM reasoning to articulate each assumption and its failure risk; Layer 3 uses fine-tuned BERT to classify assumptions into six non-functional requirement categories: performance, security, scalability, usability, reliability, and domain; Layer 4 generates ranked clarification options using FS LLM prompting; Layer 5 enables human validation through confirmation, editing, or rephrasing; and Layer 6 rewrites the requirement using the confirmed resolution, producing an explicit, testable, and traceable requirement. We also introduce the HiddenLens Corpus (HLC), comprising 312 requirements annotated with 847 hidden assumptions across six NFR categories. Inspired by PROMISE and PURE, HLC addresses the lack of annotated data for this task and is, to our knowledge, the first dataset developed specifically for it. HiddenLens outperforms keyword-based and QuARS-Adapted baselines by 71% and 61%, respectively, achieving an end-to-end F1-score of 0.79; its BERT-based NFR classifier achieves an F1-score of 0.847. These results indicate that HiddenLens can help analysts resolve hidden assumptions before they propagate through software development.	24 pa... 24 pages, 14 figures, 10 tables
Neither Stochastic Parroting nor AGI: LLMs Solve Tasks through Context-Directed Extrapolation from Training Data Priors	2026-10-21	Show This position paper argues that many behaviours framed as either "stochastic parroting" or "advanced reasoning" are better explained as context-directed extrapolation from training-data priors, and that ML and NLP research communities would benefit from evaluations that distinguish remembering and basic extrapolation from genuine application-level generalisation. Our central contribution is an actionable evaluation checklist and design patterns that make these distinctions testable in practice. Our framework posits that LLMs extrapolate from training data priors while using context to guide the model to the appropriate priors; we call this "context-directed extrapolation." Specifically, this context direction is achieved through examples in base models, leading to in-context learning, while instruction tuning allows analogous task specification via prompts rather than explicit examples. We argue that adopting this more realistic framework and controls for improving evaluation would better calibrate claims about LLM "reasoning," refocus research on how LLM performance interacts with training data and generalisability to out-of-domain tasks, and reduce miscalibration in deployment and policy decisions.	14 pa... 14 pages, 1 figure (including appendix)
A rubric landscape for evaluating clinical reasoning in large language models: what exists, what is missing, and what needs to be combined	2026-10-21	Show Exam-style accuracy does not establish whether large language models (LLMs) reason well over clinical records. We define clinical reasoning as integrating and updating evidence across time and sources to form, revise and justify a patient's problem representation and a defensible plan. This structured narrative review maps three literatures: medical education assessment instruments, clinical LLM benchmarks published from 2023 onwards, and general-domain methods for evaluating long-form generation. We examine six dimensions: problem representation, temporal synthesis, differential and management reasoning, counterfactual reasoning, calibrated uncertainty, and reasoning faithfulness. Preprints are included and flagged. No single instrument covers all six dimensions. Problem representation and differential or management reasoning are reasonably covered, although reliability varies by instrument and setting. TIMER-Eval targets temporal synthesis, and ER-Reason assesses sequential diagnostic belief updating. Dedicated uncertainty and counterfactual evaluations are emerging, but their applicability to longitudinal free-text reasoning remains limited. Factual completeness is well theorised in general-domain evaluation, with early clinical evidence of important omissions. Faithfulness remains the weakest dimension, with one identified clinical causal-ablation study on multiple-choice questions. Existing tools should be combined through binary rubric items, separate completeness and correctness scores, case-specific importance weighting with non-compensable safety caps, temporal order-consistency checks, and chance-corrected reliability reporting. Further design work is needed for calibrated uncertainty, counterfactual reasoning and faithfulness over longitudinal free-text records. This review provides a design rationale, not a validated instrument.	13 pa... 13 pages, 1 table. Structured narrative review
OMAR: Open-World Multi-Attribute Reasoning for Scene-Conditioned Shelf Organization	2026-10-21	Show With the advent of foundation models, robots are increasingly capable of understanding and executing explicitly specified human tasks. Extending these capabilities to open-ended tasks, however, requires robots to infer appropriate actions from the observed scene. Household organization exemplifies this challenge, as robots must often act without detailed instructions, predefined object vocabularies, or known object-receptacle associations. While robots may follow instructions or learn organization preferences from demonstrations, inferring the latent organization of an occasionally changing scene and placing a novel object consistently with it remains challenging. We propose an open-world multi-attribute reasoning pipeline that grounds unknown objects across views and constructs an instance-level 3D scene graph for placement reasoning. First, a vision-language model proposes labels together with attributes such as content and packaging type, which are grounded through segmentation, depth, and multi-view association. Then, an LLM reasons over this representation to infer the latent shelf organization and rank scene-consistent object placements without explicitly user-provided preferences. The resulting placement rankings guide subsequent geometric and kinematic feasibility checks for execution. Our experimental evaluation shows that our approach leads to improved instance mapping and placement coherence over single-view language-grounding and geometry-only baselines, enabling autonomous placement of previously unseen objects without predefined object-receptacle mappings or class-specific retraining.
MARS: Multi-hop Adaptive Retrieval and SPARQL Generation for KGQA	2026-10-21	Show Large language models (LLMs) have demonstrated strong reasoning performance, but their tendency to hallucinate limits their reliability in knowledge-intensive tasks requiring up-to-date and grounded information. Combining knowledge graphs (KGs) with LLMs facilitates the use of explicit symbolic knowledge that can be continuously updated without costly fine-tuning, while benefiting from rapidly advancing LLM reasoning. We propose MARS, a scalable knowledge graph question answering (KGQA) approach that requires no model fine-tuning. Rather than relying on open-ended agentic exploration, MARS performs a structured retrieval procedure that links question entities to the KG and iteratively retrieves relevant next-hop information. At each step, MARS decides whether to continue graph traversal or to generate the final SPARQL query, allowing the model to adapt the retrieval depth to the question while keeping the overall pipeline more predictable than fully agentic approaches. We evaluate MARS on three established KGQA benchmarks across several LLMs and settings, including multilingual evaluation, and provide insights through ablation studies and error analysis. Our approach achieves competitive performance relative to state-of-the-art methods while remaining efficient and scalable. The evaluation results, code and resources are publicly available: https://github.com/dice-group/mars-kgqa.	EKAW ... EKAW 2026 (https://ekaw2026.di.unito.it/accepted-contributions/papers)
When Form Changes but Logic Doesn't: Building Logic-invariant LLMs through Logical Structure	2026-10-21	Show Large language models (LLMs) have achieved strong performance across diverse reasoning tasks, yet it remains unclear whether this performance reflects an understanding of underlying logical structure or a reliance on shallow semantic cues. A key indicator of such understanding is whether models treat logically equivalent inputs consistently, producing the same predictions despite differences in expression. Through systematic evaluation on curated data, we show that even strong LLMs often fail this test: they do not consistently preserve predictions across logically equivalent inputs and generalize poorly to unseen logical forms. To address this limitation, we propose LoGIcal STructure-guided reasoning (LoGIST), a framework that guides LLM reasoning with explicit logical structure. LoGIST maps each input instance to a decision diagram, a directed acyclic graph that captures the underlying logical structure, and encodes this graph into a logical embedding that conditions the LLM's reasoning process. Across multiple training settings, LoGIST reduces inconsistency under logical equivalence and improves generalization to unseen logical forms, suggesting that logical invariance is a useful target for building more robust and generalizable logical reasoning in LLMs.	Accep... Accepted at NeurIPS 2026, 24 Pages
ReAPO: Retry-Aware Policy Optimization for Efficient Exploration in LLM Reasoning	2026-10-20	Show RLVR improves language-model reasoning, but commonly relies on independent stochastic rollouts for exploration, which can repeatedly sample similar failed strategies. Prior work has seen success with approaches that guide exploration by introducing dependence into the sampling process by including incorrect solutions in the context for subsequent attempts. Still, these methods introduce the additional question of how to assign credit to each attempt. We propose to view the credit assignment through the lens of Stochastic Shortest Paths, a framework in classical reinforcement learning that seeks to minimize the expected cumulative cost to reach a terminal state (in our case, a verified successful response). We show that, under suitable simplifying assumptions, this perspective naturally leads to a simple modification to GRPO, which we call Retry-Aware Policy Optimization (ReAPO). ReAPO allocates additional generation to failed trajectories, allowing successful retries to be reinforced without rewarding preceding failures, and assigns credit in a theoretically principled way to each attempt. Across Qwen3 and OLMo-2 models from 0.6B to 4B and eight mathematical and scientific reasoning benchmarks, ReAPO improves first-attempt performance over GRPO as well as natural recent baseline algorithms, while additional ablations demonstrate the utility of both serial sampling and ReAPO's credit assignment. On Qwen3-0.6B, ReAPO also outperforms GRPO with twice the group size or twice the training steps, reaching their peak accuracy with 39% and 31% fewer generated tokens, respectively. Taken together, our results suggest that ReAPO is a simple yet effective approach for improving LM reasoning performance through improved exploration and retry-aware credit assignment.

Skill

Title	Date	Abstract	Comment
How Does RL Post-training Induce Skill Composition? A Case Study on Countdown	2026-10-22	Show While reinforcement learning (RL) successfully enhances reasoning in large language models, its role in fostering compositional generalization (the ability to synthesize novel skills from known components) is often conflated with mere length generalization. To this end, we study what RL post-training teaches about skill composition and how the composition structure affects the skill transfer. We focus on the Countdown task (given n numbers and a target, form an expression that evaluates to the target) and analyze model solutions as expression trees, where each subtree corresponds to a reusable subtask and thus can be viewed as a "skill." Tracking tree shapes and their success rates over training, we find: (i) out-of-distribution (OOD) generalization to larger n and to unseen tree shapes, indicating compositional reuse of subtasks; (ii) the order of skill acquisition depends upon the composition structure---models master shallow balanced trees (workload is balanced between subtasks) before deep unbalanced ones, with persistent fragility on right-heavy structures for a fixed composition depth. Our diagnostic reveals what is learned, in what order, and where generalization fails, clarifying how RL-only post-training induces OOD generalization beyond what standard metrics such as pass@k reveal. To study the implications of our findings for real-world tasks, we present a preliminary study on formal theorem proving.
Adaptive Skill Granularity for Coding Agents: Runtime Routing and Macro-to-Micro Recovery	2026-10-22	Show Reusable skills allow large language model agents to package task-specific procedures, tool affordances, and execution guidance into modular units. Prior work has largely studied how agents retrieve, compose, or synthesize skills before execution. In coding agents, however, execution poses an additional runtime control problem: the same task may be solved through primitive tool calls, macro skills, or hybrid execution that starts with a macro and falls back to micro-level recovery. We introduce Adaptive Skill Granularity, a framework that treats skill granularity as an action-abstraction policy conditioned on task structure, execution state, and macro-skill failures. The system combines a feature-based granularity controller and a finite-state recovery executor. We evaluate the framework with a deterministic coding-agent protocol using verifier-defined success, total cost, and process metrics such as macro adoption, failure recovery, route match, and transition correctness. Results show that fixed macro execution is efficient when applicable but brittle under mismatch, while runtime routing and explicit macro-to-micro recovery improve robustness and recover controlled macro failures that fixed-granularity execution leaves unresolved. Our findings suggest that skill libraries for coding agents should be treated not only as retrievable resources, but also as controllable, verifiable, and recoverable execution abstractions.
R$^2$Orch: Recursive Self-Improvement for Agent Orchestration via Self-Play within a Skill-Internalization Loop	2026-10-22	Show In recent years, agentic orchestration has been widely applied to solve complex human tasks. However, existing orchestration learning still faces key challenges, including hand-designed curricula, advantage collapse when a task's rollouts all reach the same outcome, and unreliable evaluation signals that judge each skill by one noisy score. To address these challenges, we propose R$^2$Orch, a new paradigm of Recursive Self-Play Orchestration: an inner loop of self-play that improves the orchestrator on self-chosen tasks, and an outer loop that improves how the inner loop improves. To support this paradigm, we introduce Recursive Learnability, a training objective that rewards a Proposer for choosing tasks whose rollouts keep reaching different outcomes as skills are withdrawn. Built on this objective, we further propose Skill Withdrawal, in which the outer loop withdraws each skill once a controlled evaluation within self-play shows that the orchestrator performs as well without it. Moreover, the two loops form one cycle: failures are distilled into skills, skills into parameters, and withdrawals lead the Proposer to new failures, so the orchestrator moves toward zero-skill deployment. Experimental results on twelve datasets show that R$^2$Orch consistently outperforms baselines across question answering, math, code, and interactive tasks. Our code is available at https://github.com/beita6969/R2Orch.	21 pa... 21 pages including appendix
SkillFlow: Flow-Driven Recursive Skill Evolution for Agentic Orchestration	2026-10-22	Show In recent years, a variety of powerful LLM-based agentic systems have been applied to automate complex tasks through task orchestration. However, existing orchestration methods still face key challenges, including strategy collapse under reward maximization, high gradient variance with opaque credit assignment, and unguided skill evolution whose decisions are typically made by directly prompting an LLM to judge rather than derived from principled training signals. To address these challenges, we propose SkillFlow, a flow-based framework that takes a trainable Supervisor as the agent and a structured environment with dynamic skill library and frozen executor, automating task orchestration through multi-turn interaction. SkillFlow employs Tempered Trajectory Balance (TTB), a regression-based flow-matching loss that samples trajectories proportional to reward, preserving diverse orchestration strategies rather than collapsing to a single mode. The same flow objective yields a jointly learned backward policy that provides transparent per-step credit assignment at zero additional inference cost. Building on these flow diagnostics, a recursive skill evolution mechanism determines when to evolve, what skills to create or prune, and where decision gaps lie -- closing the loop from training signal to autonomous capability growth. Experimental results on 14 datasets show that SkillFlow significantly outperforms baselines across question answering, mathematical reasoning, code generation, and real-world interactive decision making tasks. Our code is available at https://github.com/beita6969/SkillFlow.	49 pa... 49 pages, 5 figures, 6 tables
R$^2$ Flow: Recursive Self-Improvement via Recursive Skill Evolution	2026-10-22	Show LLM-based agents can improve themselves across tasks by reusing and revising the skills they orchestrate into executable procedures. Flow-based training fits this loop: it samples procedures in proportion to reward, and the flow through each skill credits it for the next library revision. Three obstacles stand in the way of making this self-improvement reliable: flow training suffers strategy collapse over tree-structured histories; nonnegative flow-based credit rewards frequent use as if it were benefit; and library edits rest on the task reward the policy optimizes. We introduce R$^2$ Flow, a recursive self-improvement framework that alternates policy learning, independent verification, and versioned skill-library updates on a shared-state orchestration graph. The graph merges histories that differ only in the order of independent steps, allowing flow training to pool evidence across equivalent executions. A flow-share readout of the trained flow, invariant to the backward policy, and a separate signed utility rank which skills to change, verifier evidence decides whether an edit is warranted, and a residual-variance plateau sets when to update. Committed edits reshape the graph the next policy learns on, realizing recursive skill evolution. Across question answering, mathematical reasoning, interactive decision making, and code generation, R$^2$ Flow improves task accuracy and library-edit precision over heuristic orchestration, reinforcement learning, and skill-evolution baselines, and transfers across executors. Code is available at https://github.com/beita6969/r2flow.	27 pages
V2-STRep: VLM-Grounded Structured Task Representations for Reusable Robot Skills Acquired from Generated Videos	2026-10-22	Show Human manipulation videos provide rich motion and interaction cues for acquiring robot skills without robot demonstrations. Video generation models synthesize such demonstrations from an initial scene image and task instruction, avoiding the need to record demonstrations for each task. However, the recovered motion captures only one scene-specific realization, leaving task structure, geometric relations, and constraints implicit. We present V2-STRep, a zero-shot framework that converts generated video motion into reusable robot skills through VLM-grounded structured task representations. The representation specifies motion phases, references, and task-relevant constraints, with targets described by minimal geometric structures: points, point-normals, axes, planes, and full 6D poses. VLM-provided 2D image-space cues are lifted into 3D using RGB-D observations to reconstruct task geometry and candidate grasp poses. Geometry-specific rules transfer motion to new scenes, while task-constrained trajectory optimization couples grasp selection with complete robot motion planning. It preserves task requirements while using remaining rotational freedom to accommodate joint limits. Updating deployment grounding and constraints enables reuse under new compatible instructions without generating another video. Experiments on six real-world manipulation tasks demonstrate improved execution success over baselines, reliable cross-scene transfer of successfully acquired skills, and adaptation to changed deployment instructions. Details are available on the project website: https://tuwien-asl.github.io/V2-STRep/.
Unifying Goal-Conditioned RL and Unsupervised Skill Learning via Control-Maximization	2026-10-22	Show Unsupervised pretraining has driven empirical advances in goal-conditioned reinforcement learning (GCRL), but its theoretical foundations remain poorly understood. In particular, an influential class of methods, mutual information skill learning (MISL), discovers behaviorally diverse skills that can later be used for downstream goal-reaching. However, it remains a theoretical mystery why skills learned through MISL should support goal-reaching. A subtle challenge is that both GCRL and MISL are umbrella terms: Different GCRL tasks use distinct criteria for measuring goal-reaching performance, while different MISL methods optimize distinct notions of behavioral diversity. We address this challenge and unify GCRL and MISL as instances of control maximization. We identify three canonical GCRL formulations and prove that they are fundamentally inequivalent: They can induce incompatible optimal policies even in the same environment. Nevertheless, they all share a common interpretation: A well-performing goal-conditioned policy is one whose future trajectory is highly sensitive to the commanded goal, with the precise notion of sensitivity determined by the GCRL formulation. Noting that MISL objectives can be understood as measures of skill sensitivity akin to goal sensitivity, we show that MISL objectives are bounded by formulation-specific downstream goal sensitivities. These bounds establish a precise correspondence between MISL methods and downstream GCRL tasks: For every GCRL formulation, there exists a matching MISL objective for which more diverse skills afford greater downstream goal sensitivity. Our results thus lay a theoretical foundation for RL pretraining and have important practical implications, such as suggesting which pretraining objectives to use when a user cares about a specific class of downstream tasks.	Accep... Accepted at NeurIPS 2026 (Oral Presentation)
How to Teach Large Multimodal Models New Skills	2026-10-22	Show How can we teach large multimodal models (LMMs) new skills without erasing prior abilities? We study sequential fine-tuning on five target skills while monitoring general ability on eight held-out benchmarks across three model families. Surprisingly, we find that performance lost on held-out tasks after fine-tuning on one skill can partly recover when the model is subsequently tuned on a different skill. We trace this behavior to a measurable shift in the output token distribution, manifested through a simple counting-bias probe that shows the shift co-varies with forgetting. Guided by this insight, we identify two simple, robust tuning recipes that learn strongly while limiting drift: (i) updating only the self-attention projection layers (SA Proj., $Δ$ learning +24.9 / $Δ$ held-out forgetting -0.6), and (ii) updating only the MLP Gate&Up while freezing the Down projection (+30.5 / -2.1). Both substantially outperform full-LLM tuning (+31.8 / -23.3) in the learning-forgetting trade-off. We also compare against common forgetting mitigation methods: Learning without Forgetting (LwF), LoRA, Mixture-of-Experts, and weight-space interpolation (WiSE-FT), and find that our selective tuning recipes match or exceed their learning-stability balance while remaining simpler, requiring no replay, auxiliary parameters, or per-stage tuning. These results hold across LLaVA-OneVision, LLaVA-NeXT, and Qwen2.5-VL, confirming that the key to teaching LMMs new skills without forgetting lies in controlling output distribution shift by choosing which components to tune. Code will be made available.	ECCV ... ECCV 2026 Spotlight. Code is available at https://github.com/jessemelpolio/LMM_CL
Impacts of generative AI skill development on academic adaptation and learning outcomes among university students	2026-10-22	Show Understanding how university students perceive the development of their generative artificial intelligence (GenAI) skills over time can clarify how GenAI use relates to learning and academic outcomes. This study examined perceived skill-development trajectories, changes in learning practices, and associated academic gains using a retrospective questionnaire with 313 analyzed responses and interviews with 18 students across subject areas. Descriptive and comparative analyses and exploratory structural equation modeling (SEM) were integrated with thematic analysis. Perceived developmental trajectories were heterogeneous: 34.2% of students reported rapid early improvement followed by a plateau, 28.4% steady improvement, and 21.1% delayed acceleration. Development was also uneven across skills, with improvement in evaluating GenAI outputs perceived as more challenging; assignments were the most frequently selected driver of skill growth. Interviews further indicated increasingly specific prompting, tailored verification, and selective delegation, although some students reported stable routines, dependence, skipped reasoning, or improvements attributable partly to better GenAI tools. Academic gains were reported more often for assignments than for examinations. The SEM indicated a positive indirect association between perceived skill growth and academic gains through adaptation of assignment processes, whereas academic gains did not clearly differ across developmental trajectory groups after multiplicity correction. The findings suggest that perceived GenAI skill development varies in both trajectory and skill domain and is associated with changes in learning practices and differentiated academic benefits.
VisCurate: Evaluating Behavioral Equivalence of Visual Skills	2026-10-21	Show Modern vision-based agents have shifted toward skill-based problem-solving, where a skill is an executable image-to-image operation, e.g., cropping, thresholding, or edge detection. As these skill libraries grow, they accumulate duplicated, mislabeled, and defective skills. The signals normally used to curate them (a skill's name, description, or source code) say little about how the skill visually changes the image. We present VisCurate, a framework and benchmark that curates visual skill libraries from observed execution rather than from text. VisCurate runs each candidate skill pair on a shared probe battery and assigns one of six behavioral relations, each tied to a library edit that is safe to make. The benchmark injects seven defect types into a clean library at controlled rates, producing 300 reproducible instances with ground-truth relations and repairs. While text baselines accept up to five of six false merges on hard negatives, our verifier accepts none and recovers 51.2% of the pairs that are truly mergeable. On two real, uncurated libraries, VisCurate finds that 92% of the skills a package shares with its downstream copy are redundant, while execution confirms only 10 of 17 published equivalence claims. The curation problem also remains open, as the strongest of the nine automated curators recovers only about half of the repairs. We emphasize that visual skill libraries should be curated by behavior, not by description.	Accep... Accepted at the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 2027), Main Conference

RSI

Title	Date	Abstract	Comment
RSI Control: Can AI Models Perform Covert Sabotage in the Self-Improvement Process?	2026-10-21	Show Recursive self-improvement, in which frontier AI models work on training the next generation, is quickly becoming a reality. However, the safety of this process has received far less attention than it should. A misaligned model training its successor can shape what that successor knows, how it behaves, and what it wants, and the harm carries forward into the next generation. We study this risk through the lens of AI control. We introduce RSI Control Arena, a testbed in which a frontier model, acting as an autonomous R&D agent, is given a benign main task from a stage of a successor model's training pipeline together with a covert side task to install a misaligned property in the successor. The testbed contains 16 side tasks in three classes, forming 22 task pairs with main tasks that span the pipeline from data curation to reinforcement learning. We evaluate six frontier models as agents against LLM monitors that read the transcript of the run and the agent's reasoning. We observe that every model completes a substantial share of the side tasks without degrading the main-task performance, including side tasks that make the successor itself a better saboteur. Under minimal instructions to stay covert, models already produce successful attacks that monitors fail to detect, and providing a short list of evasion tactics further reduces the detection rate. Monitors with access to agent's reasoning detects most attacks, but not all. Our results indicate that recursive self-improvement raises significant safety concerns, and that strong safety effort should be in place before models are trusted with greater responsibility of training their successors.
MaD-RSI: Model-as-Data for Recursive Self-Improvement of Agent Harnesses	2026-10-20	Show Recursive self-improvement (RSI) of agent harnesses has emerged as an effective approach to enhancing the capabilities of large language model (LLM) agents. By searching the harness space for a specific LLM on a given dataset, RSI can discover a harness tailored to the capabilities and behavioral characteristics of that model. However, such optimization often creates a strong coupling between the learned harness and its target LLM, limiting generalization across models and requiring substantial computational resources to adapt to new LLMs. Moreover, existing RSI methods typically optimize a harness using feedback from a single model and evaluate its generalization to other models only after training. This paradigm treats model differences primarily as a criterion for evaluating generalization rather than as valuable evidence for optimization. In this paper, we introduce a new perspective that treats models as data and places the harness at the center of optimization. Building on this perspective, we propose \textsc{MaD-RSI}, a framework that directly optimizes a shared harness across multiple heterogeneous LLMs by incorporating their diverse failure trajectories into the training process. Rather than assessing model differences only after optimization, \textsc{MaD-RSI} leverages these differences as training evidence to guide the self-improvement of a shared harness. Furthermore, we introduce a new evaluation framework with two complementary metrics designed to assess the effectiveness of harness optimization across models. Experiments across multiple datasets and LLMs demonstrate that \textsc{MaD-RSI} improves generalization over existing approaches while reducing training costs.
Where is Physical RSI capability from? An Observational Scaling Law of LLM-as-policy	2026-10-19	Show In this paper, we systematically investigate the mechanisms, scaling behavior, and generalization of physical Recursive self-improvement. We introduce PhysRSI-eval to quantify physical RSI through a five-axis metric: First-Attempt Performance, Efficiency, Success, Budget and Improvement from failure. Different VLMs are evaluated on 18 manipulation tasks under a fixed interaction budget. The analysis of experimental results motivates the Physical Barrel Hypothesis: different capabilities may constrain different parts of improvement. The five-axis metrics also distinguish systems with close public benchmark averages and show that high scores need not come with early or efficient gains. We study observational scaling from two perspectives: predicting the relative performance of VLM agents from their public capabilities, and forecasting later progress from early physical feedback. Together, these analyses ask how quickly and how far an agent can improve. Moreover, we conduct workspace interventions and then analyze what improvement leaves behind. Removing history reduces late-round success, frozen workspaces remain useful on unseen layouts, and three cross-task cases achieve success earlier. The results suggest that scene-specific adaptation and reusable experience can coexist.	Tech Report
CoEvo-RSI: When Harnesses Evolve, Which Models Should Follow?	2026-10-18	Show Model--harness co-evolution has emerged as a prominent paradigm for recursive self-improving (RSI) agents, jointly evolving their execution scaffolds and intrinsic capabilities. We study the persistent yet often overlooked brain--body mismatch problem in this paradigm, where an evolving harness (body) demands behaviors its model (brain) has not learned or restricts capabilities it already possesses. We introduce CoEvo-RSI, a RSI framework that advances model--harness co-evolution through agent-driven, bidirectional historical inheritance. Drawing on a shared evolutionary archive and execution feedback, a meta-agent coordinates harness development from historical designs with model checkpoint selection or merging and subsequent adaptation under the selected harness, aligning inherited capabilities with evolving execution requirements to reduce brain--body mismatch. Using only supervised fine-tuning (SFT) on its own rejection-sampled trajectories, CoEvo-RSI makes Qwen3.5-9B competitive with larger models across diverse vertical applications, including multi-agent failure attribution, role-aware text-to-SQL, and smart-home instruction following. It surpasses DeepSeek-V4-Pro on TraceElephant (+14.16 agent accuracy) and GLM-5.3 on HomeBench VM (+4.90 success rate), underscoring the need for brain--body alignment and the potential of bidirectional historical inheritance to advance agent co-evolution.
Terminal-RSI: Recursive Self-Improving Terminal Agents via Recursive Self-Rewrite and Policy-Refine	2026-10-15	Show Converting verifier outcomes into useful policy updates and partial progress into complete task resolution is crucial for terminal agents. We present Terminal-RSI, which alternates Recursive Self-Rewrite (RSR) of verified trajectories with critic-inheriting Recursive Policy-Refine (RPR). We formulate Turn-Boundary Skip-Observation GAE (TB-GAE) to retain token-specific baselines while applying decay across interaction turns. Family Dense Reward (FDR) calibrates semantic progress across task families; Success-Prioritized Reward (SPR) separates strict success from bounded partial credit. A Warm-up Critic Model, recursive critic handoff, and stage-specific learning rates coordinate the two refinement phases. The five-stage sequence from Qwen3.8-27B base model through RSR-1, RPR-1, RSR-2, and RPR-2 yields Terminal-RSI-27B: 19.84% on Terminal-Bench 4.0, 86.52% on Terminal-Bench 2.1, 16.95% on Terminal-Bench 3.0, and 79.00% on Terminal-Bench Hard. On LHTB, process reward rises from 0.21 to 0.37 and strict completion from 0/46 to 5/46; SWR100 success doubles from 3.00% to 6.00%. By recursively coupling verified trajectory rewriting with critic-inheriting policy refinement, each retained stage consolidates discovered behaviors and progressively advances the agent from partial local repairs to strict terminal verification.	28 pa... 28 pages, 16 figures, 5 tables
MMLA-RSI: Recursive Self-Improvement through RTT and RTMA	2026-10-14	Show An agent can improve its behavior within an unresolved problem by changing how it acts and which experience it retains. We propose MMLA-RSI, a recursive self-improvement scheme with two implementations in Memory-Mediated Learning Architecture: Reasoning-Time Training (RTT) changes the policy state $Φ$, while Reasoning-Time Memory Adaptation (RTMA) changes authoritative memory $M$. Each route also makes its adaptation method an object of improvement. A current numerical updater trains the policy proposing its successor; a current memory manager chooses the records available to the process proposing its successor. An independently admitted successor takes over the same role in the next generation. We give bounded executable rule languages, shared task/proposal interfaces, and one algorithm for either route or their composition. The RTT instance changes scalar schedules in a clipped-momentum update. The RTMA instance chooses NULL or a complete-row write with possible eviction. A fixed interpreter, trusted committer, and independent evaluator bound what can change. State versions, read closures, and conditional properties describe causal execution, accepted transitions, and finite work. Existing numerical-component studies motivate the construction. In four-candidate program selection, retaining LARC feedback updates reduces expected error by 24.65 and 36.65 percentage points for two trained initializations. In chronological continuous-integration replay, feedback binding is positive while retention increases loss. Completed SQL studies improve substantially over an unadapted substrate but do not establish the required additional online benefit. These results characterize their declared components. The two recursive routes and their joint realization are specified here; their beneficial recursive chains and discovery efficiency remain to be evaluated.	Techn... Technical report. 24 pages, 5 figures, 12 tables
AReaL-RSI: Adaptive Scaling and Coordination of Recursive Self-Improvement	2026-10-13	Show Scaling recursive self-improvement requires additional computation to produce useful research progress in reaching persistent objectives. Intuitively, more researcher agents can explore more hypotheses, but early evaluation scores might misrepresent their promise, shared findings could lose their applicability conditions, and useful procedures could remain confined to completed investigations. In this paper, we present AReaL-RSI, a framework for organizing a persistent research team of agents without updating backbone LLM weights. We formulate the scaling problem in terms of evaluated outcomes and whole-system resource expenditure. Concretely, a leader agent selects task-appropriate research orientations and guides ongoing investigations; conditional memory sharing distinguishes unresolved problems from supported findings and enables recipient adaptation. During recursive evolution, after each campaign, evidence-based review revises researcher orientation, knowledge, and skills for later campaigns. Experiments compare AReaL-RSI with Codex-Goal on downstream harness optimization, GPU kernel optimization, and AI research tasks using two backbones. Under the reported resource allocations, AReaL-RSI improves Terminal Bench performance by 2.2 percentage points for each backbone and increases mean KernelBench speedup by 32.2-36.8%; further scalability comparisons favor the design of AReaL-RSI over native scaling: combining experience sharing with specialized researcher roles yields 1.9-27.5% lower kernel latency than independent exploration and 1.3-33.9% lower latency than sharing alone across all seven comparable task-backbone pairs.	12 pages
ECG-RSI: Recursive Self-Improvement through Targeted Electrocardiogram Waveform Evolution	2026-10-13	Show Electrocardiogram (ECG) models need supervision that targets evolving weaknesses and remains faithful to waveform evidence. We introduce ECG-RSI, a recursive self-improvement framework combining verified waveform synthesis with feedback-guided intervention search. Signal remeasurement verifies waveform edits; construction errors guide generator repair, while learner errors motivate interventions compared on a separate selection set. Across seven benchmarks, three rounds raise Qwen3.5-35B-A3B's mean question-answering accuracy from 34.57% to 56.04% and mean classification F1 from 24.27 to 34.03. Across the retained trajectory, the preferred supervision shifts from direct reinforcement to confusion-focused contrasts. Continued feedback correction yields diminishing external gains. These results motivate separating error diagnosis from intervention selection and checking whether gains on the feedback set transfer to independent benchmarks.	Proje... Project: https://pork-pay.github.io/ECG-RSI/; GitHub: https://github.com/pork-pay/ECG-RSI
Reels-RSI: Program-Level Self-Improvement for Code-Rendered Explainer Videos	2026-10-10	Show Explainer videos -- a worked geometry problem, why the sky is blue, how a dynasty fell -- are usually one more target for pixel video generation. We argue they are a different kind of artifact: a program (script, storyboard and per-frame animation code) that a deterministic renderer turns into frames. A program can be read, diffed, tested and audited; pixels can only be watched. Reels-RSI is an open-source reel director for any agent harness; its self-improvement happens at the program level rather than in weights or prompts. Every revision pass is logged with a code snapshot; a retrospective pairs each check finding with the diff that fixed it and proposes lessons; a lesson that a static check can catch is compiled into a rule with a must-fail and a must-pass example, so the memory is unit-tested; and a rule learned today runs over everything made before it. A bounded, benchmark-gated step lets an agent rewrite the director's playbook, kept only when blind pairwise judgments on held-out topics prefer it; nothing enters memory without a human decision. Across 111 reels -- a field study, two playbook evolutions, a 2x2 memory ablation, a weaker director, a 48-reel learning curve and a judge-reliability study -- we find that the rule learned in the field caught four latent defects in frames authored before it existed and none after; that rules are an audit for a strong director (eight unguarded reels, no violation) and a guard for a weaker one (thirty violations unguarded versus five); that lessons shorten revision loops but do not compound or move the judge; and that the first rewrite's 11--1 verdict was a position bias in our vote design (judges pick the second-shown reel 62--81% of the time), masking scripts that traded thirteen facts for craft. We document four failure modes of the preference signal, which check caught each, and the fixes. Code, rules, lessons and data are public.	19 pa... 19 pages, 4 figures, 9 tables. Code, rules, lessons, data and evolve reports: https://github.com/agentic-loops-x/Reels-RSI
RSI-Anything: Unlocking Recursive Self-Improvement with World Models	2026-10-09	Show Recursive self-improvement (RSI) aims to transform interaction experience into persistent improvements, yet external feedback is costly and world-model predictions may become unreliable as policies evolve. We introduce RSI-Anything, a domain-general framework built on JEPA-Anything that uses predictive world models to guide action and learning-data selection. Observed feedback drives policy or workflow improvement and conditional world-model adaptation, while component and compatibility checks govern the retention of updated policy-model pairs. Across diverse computational and scientific domains, RSI-Anything demonstrates substantial gains. In code repair, with the coder fixed, success increases from 26.3% with a static verifier to 29.2% through predictive selection and 31.2% after world-model adaptation, at comparable test usage. Graphical user interface (GUI) evaluations showed higher task completion, while average actions decrease by 39.4%. Across five successive policy versions, laboratory-planning scores more than double from 21.80% to 46.01%. In reaction optimization, RSI-Anything achieves 96.15% mean best observed yield versus 82.16% for Bayesian optimization under the same 40-experiment budget. An end-to-end immune-science application further connects workflow evolution and predictive modeling with therapeutic hypothesis generation. These results demonstrate how predictive world models and retained system updates enable recursive improvement across diverse environments.	Code:... Code: https://github.com/Gen-Verse/RSI-Anything

Evolve

Title	Date	Abstract	Comment
MLEvolve: A Self-Evolving Framework for Automated Machine Learning Algorithm Discovery	2026-10-22	Show Large language model (LLM) agents are increasingly applied to long-horizon tasks such as scientific discovery and machine learning engineering (MLE), where sustained self-evolution becomes a key capability. However, existing MLE agents suffer from isolated exploration, inefficient experience reuse, and lack of hierarchical control, which together hinder long-horizon optimization. We present MLEvolve, an LLM-based self-evolving multi-agent framework for end-to-end machine learning algorithm discovery. By extending tree search to Progressive MCGS, MLEvolve enables cross-branch information flow through graph-based reference edges and gradually shifts the search from broad exploration to focused exploitation with an entropy-inspired progressive schedule. To allow the agent to evolve with accumulated experience, we introduce Retrospective Memory, which combines a cold-start domain knowledge base with a dynamic global memory for task-specific experience retrieval and reuse. For stable long-horizon iteration, we further decouple strategic planning from code generation with adaptive coding modes. Evaluation on MLE-Bench shows that MLEvolve achieves state-of-the-art performance across multiple dimensions including average medal rate and valid submission rate under a 12-hour budget (half the standard runtime). Moreover, MLEvolve also outperforms specialized algorithm discovery methods including AlphaEvolve on mathematical algorithm optimization tasks, demonstrating strong cross-domain generalization. Code is available at https://github.com/InternScience/MLEvolve.	24 pa... 24 pages, 8 figures, 6 tables. Project page: https://internscience.github.io/MLEvolve/
HERD: Harness-Evolving Recursive Distillation for Continual Learning	2026-10-22	Show Continual learning requires a model to keep acquiring new skills and knowledge without compromising its existing capabilities. Parametric methods suffer from catastrophic forgetting and loss of plasticity. Non-parametric methods, such as prompts, skills, and harnesses, leverage in-context learning but are limited by context length and lossy plain-text compression. To effectively combine their complementary strengths, we propose HERD (Harness-Evolving Recursive Distillation), a hybrid continual learning framework that learns efficiently in the harness and internalizes experience into cumulative capabilities via distillation. On the harness side, HERD continually refines its skills to capture reusable patterns, while retaining high-residual examples that the skills fail to capture in a bounded memory. On the parameter side, HERD performs drift-constrained distillation with dual anchors, substantially reducing forgetting over long recursive runs. On the 25-stage task stream of TRACE+17, under fixed memory and compute budgets, HERD matches the strongest parametric baseline in new-task acquisition while exhibiting the least forgetting and the strongest retention of general capabilities.	22 pages, 9 figures
Evolving with Open Worlds: Towards Recursive Self-Improvement for Future Forecasting	2026-10-22	Show Most contemporary artificial intelligence (AI) systems are released with frozen parameters and improve through external updates, whereas an ideal system would keep improving from its own experience, a capacity we call recursive self-improvement (RSI). Yet studying RSI requires an environment that is continuously changing, objectively verifiable, and naturally grounded, posing a fundamental challenge for agent evolution. Here we introduce EvoForesight, a sustainable evolution agent system grounded in future event prediction, in which agents continuously interact with a dynamically evolving environment and improve by internalizing and externalizing the experience gained through these interactions. A living environment synthesizes verifiable tasks, and a recursive-evolution engine turns the reasoning trajectory behind each prediction back into capability: decoupling search from prediction adapts parameter-level evolution to the contamination-prone live environment, while distilling reusable skills lets the agent evolve from the limited live trajectories such an environment affords. On the external FutureX leaderboard, which evaluates predictions against real-world outcomes and is updated weekly, our self-improved forecaster ranked first across all systems over three consecutive weeks, achieving up to a 195% improvement in rating. On our synthesized evaluation, it ranked among the top three across four base LLMs, outperforming the strongest baseline by up to 10.7 Bradley--Terry points. More broadly, these results suggest that a system built for self-improvement is best understood as an agent that co-evolves with a genuinely living, open-ended environment, pointing to a shift in AI from training static models on fixed data toward cultivating agents that grow stronger from their own experience, with such an open world as the indispensable basis of that growth.	47 pa... 47 pages, 13 figures. Working in progress
EvoSteer: Online Self-Evolving Graph Orchestration via Reference-Anchored Credit Assignment	2026-10-22	Show In recent years, LLM-based multi-agent systems have been widely applied to orchestrate tool-using agents into executable communication graphs. However, existing self-evolving orchestration still faces key challenges, including post-hoc evolution that revises the team only after the trajectory ends, credit diffusion that gives every action the same terminal advantage under confounded baselines, and skill admission that is uncalibrated and never retired. To address these challenges, we propose EvoSteer, a new paradigm of Online Self-Evolving Graph Orchestration -- the orchestrator builds a running team and repairs its plausible but failing steps from execution features and a learned value estimate. To support this paradigm, we introduce Anchored Trajectory Balance (AnchorTB), a regression-style flow-matching loss that assigns each orchestration action a coefficient by balancing subtrajectories against a frozen reference. Built on the learned flow, we further propose Validated Skill Admission, in which a candidate skill is tried before promotion and promoted only if paired evidence passes a sequential test under a shared nominal testing budget. Moreover, AnchorTB combines measured task-level reference reward statistics with prefix-dependent corrections. Experimental results on twelve datasets show that EvoSteer significantly outperforms baselines across question answering, mathematical reasoning, code generation, and interactive decision making. Our code is available at https://github.com/beita6969/evosteer.
Evolving Federated Density Peak Clustering with Shuffler-Based Anonymization	2026-10-22	Show In this paper, we present a novel one-shot federated clustering approach based on evolving fuzzy clustering and shuffler-based anonymization. In federated clustering, each round reveals additional model information to the server and can increase the opportunity to infer properties of local datasets, even when raw samples are never shared. Consequently, fewer client-server exchanges and unlinkability are important, so transmitted information cannot be attributed to any single client. Data-stream clustering algorithms, which learn in a single pass over the data, align well with the one-shot federated learning paradigm, where a single communication round can be sufficient. The proposed Evolving Federated Density Peak Clustering (eFDPC) method combines evolving fuzzy data-stream clustering and density peak clustering in a federated setting. Each client independently evolves micro-clusters from private data mini-batches through cluster addition and merging. To limit communication, density peak clustering is applied to the learned micro-clusters and only the selected peaks are transmitted, rather than the full local models. A shuffler-based layer is introduced between the clients and the server to anonymize contributions prior to reconstruction and global peak selection. We consider a peak-only variant, eFDPC-P, as well as eFDPC-G with server-side label propagation. Experiments on synthetic and real-world benchmark datasets show that eFDPC outperforms recent federated clustering baselines in one-shot settings and scales to higher-dimensional, non-IID data.	30 pa... 30 pages, 4 figures. Published open access (CC BY 4.0) in Evolving Systems 17, 123 (2026). Code: https://github.com/mihaozbot/eFDPC
Large Language Models Meet Bayesian Optimization: Towards Self-evolving Wireless Digital Twins	2026-10-22	Show Wireless digital twins (DTs) are increasingly integrated with artificial intelligence to support online modeling, prediction, and network optimization. However, most existing approaches perform adaptation within predefined learning or optimization structures, which may become unsuitable when wireless tasks and operating conditions change. In this article, we investigate the integration of large language models (LLMs) and Bayesian optimization (BO) to extend wireless DTs from online adaptation towards algorithm-level self-evolution. We introduce an LLM-assisted BO framework in which BO provides a stable and sample-efficient optimization backbone, while the LLM determines whether selected BO components should be retained or adapted based on the current task and accumulated optimization experience. In simulations, we apply the framework to Gaussian process (GP) kernel discovery for a multi-user wireless beam-selection problem. Simulation results show that the LLM-discovered kernel improves the final BO performance by 19.7% on the training tasks compared with the strongest standard kernel baseline. The results demonstrate the potential of combining LLM flexibility with structured BO for self-evolving wireless DTs.
Sci-CoE: Co-evolving Scientific Reasoning LLMs via Geometric Consensus with Sparse Supervision	2026-10-22	Show Large language models (LLMs) have demonstrated exceptional reasoning capabilities, and co-evolving paradigms have shown promising results in domains with verifiable reward signals, such as code and math. However, in unverifiable domains, such as scientific reasoning, these methods remain fragile due to unreliable solution evaluation and limited diversity in verification strategies. In this work, we propose Sci-CoE, a two-stage co-evolving framework for domains without verifiable feedback. The framework enables models to self-evolve as both solvers and verifiers by transitioning from sparse supervision to unsupervised learning. In the first stage, the model uses a small set of annotated data to establish fundamental correctness judgment anchors for the Verifier. In the second stage, we introduce a geometric reward mechanism that jointly considers consensus, reliability, and diversity, driving large-scale self-iteration on unlabeled data. Experiments on several general scientific benchmarks demonstrate that Sci-CoE enhances complex reasoning capabilities and exhibits strong scalability, facilitating the construction of more robust and diverse evaluation systems. Codes are available at https://github.com/InternScience/Sci-CoE.
ADAPT for Science: A Multi-Agent System for Self-Evolving Scientific Optimization	2026-10-22	Show Scientific discovery often requires identifying a small number of high-performing designs in vast, constrained search spaces. This is difficult when evaluations are costly, feasible solutions are sparse, parameters are strongly coupled, and physical constraints are strict. Recent AI for AI methods can automatically improve algorithms and optimization procedures. However, general purpose approaches are not designed around scientific priors, physical constraints, and limited evaluation budgets. We introduce ADAPT for Science, a multi-agent framework that treats scientific optimization as two coupled decisions: where to search and how to search. ADAPT uses historical designs to identify promising and diverse initial conditions, progressively concentrates computation on favorable search directions, and evolves the optimization procedure from execution feedback. We evaluate ADAPT on the design of gravitational wave detectors using a differentiable physics simulator and eight unseen detector topologies. Under matched scientific evaluation budgets, ADAPT achieves a mean loss of 0.204 +/- 0.101, reducing loss by 54.9% relative to the best classical optimizer and by 62.2% relative to the best general purpose agentic evolution baseline. ADAPT achieves the lowest loss on all unseen topologies among the evaluated baselines. ADAPT provides a step toward scientific optimization systems that continuously learn from prior experience and their own search outcomes.	Inclu... Includes supplementary material
ManimAgent: Self-Evolving Multimodal Agents for Visual Education	2026-10-22	Show Multi-round reflection lets agents built on large language models recover from failures within a single task, but each task remains an isolated episode: lessons learned across many reflection rounds on one task are discarded before the next begins. We study this gap on a code-generation task: from a scientific paper section, the agent writes Python in the open-source Manim library to render a mathematical animation. We present ManimAgent, a self-evolving multimodal agent that carries reflection experience across tasks through a dual-channel Episodic Memory Bank grown entirely from its own task stream, with no weight updates and no human seeds. After each animation converges, a vision-language model scores the rendered keyframes; the resulting signals populate a positive channel M+ that stores success rationales as soft Reference Examples, and a negative channel M- that stores validated failure patterns as hard Known Pitfalls. On a fixed-probe evaluation against no-memory, matched-budget retrieval-augmented generation, and shuffled-memory baselines, blind human Pass@1 rises and reflection rounds fall as memory size grows.	Accep... Accepted to EMNLP 2026 (Main Conference). Project page: https://manimagent.github.io/
Large language models show systematic failures in incorporating evolving medical knowledge	2026-10-22	Show Medical knowledge evolves as clinical guidelines revise recommendations, yet whether large language models (LLMs) faithfully incorporate such changes remains unclear. We introduce ChronoEvo, comprising 1,709 recommendation-level revisions across 12 forms of medical knowledge change, to distinguish recognition of revised guidance from incorporation into clinical recommendations. In the primary model, 90.8% of responses recognized revisions and 80.2% were clinically correct, but only 58.5% fully incorporated guidance, yielding a 32.3-percentage-point recognition-incorporation gap. This dissociation persisted across 14 models and was corroborated by two clinicians' adjudicated assessment of 300 responses, yielding a 40.3-percentage-point gap in the validation sample. Changes in recommendation strength and patient eligibility were associated with retention of outdated guidance. A revision-salience intervention reduced outdated-content retention and improved clinical correctness but did not significantly improve full incorporation. These findings identify faithful incorporation of evolving medical knowledge as a distinct challenge for clinical LLM reliability, even when updated guideline evidence is available.

Agent

Title	Date	Abstract	Comment
BrickWright: Agentic Creation of Physically Feasible Brick Builds	2026-10-22	Show Frontier coding agents are becoming increasingly adept at 3D modeling. Yet, they still face challenges when building objects meant to be assembled in the physical world, often creating invalid assemblies with floating, colliding, or disconnected parts. We argue that providing agents with tools for assembling builds and verifying their physical feasibility is sufficient for them to produce physically valid assemblies that still adhere to input prompts. Focusing on the domain of LEGO brick builds, we present BrickWright, a framework for generating physically feasible brick builds from text or image prompts using pre-trained agents, where a build is physically feasible if it uses only real parts, has no colliding or floating parts, and forms a connected structure if possible. BrickWright guides agents to create builds by writing code with BrickLayer, a Python toolkit with two main components: an ergonomic brick assembly API that halves inference cost compared to working in raw LDraw, and a validator that provides feedback on physical feasibility. Whereas baseline agents achieve physical validity rates ranging from only 9% (GPT-6 Luna) to 69% (GPT-6 Astra), BrickWright-guided agents achieve 100% validity on a sample of 512 prompts, while improving prompt adherence for lightweight agents and maintaining it for stronger agents. Additionally, they outperform specialized models on the text-to-brick task. Code is available on our project page: https://avalovelace1.github.io/brickwright/
EpiSEM-Bench: Evaluating Safety and Evidence Recovery in LLM Agents under Missing Evidence	2026-10-22	Show Tool-using language model agents rely on environmental facts to decide whether an action is permitted. When a tool response omits a safety-relevant fact, an agent may carry out a prohibited action even though the user request is benign and every returned statement is true. We introduce \bench{}, an executable benchmark for evaluating unsafe execution, safety boundaries, and evidence recovery in LLM agents under missing evidence. It contains 300 matched unsafe/safe task pairs across six domains. By systematically suppressing critical facts, the benchmark measures unsafe execution under critical evidence omission and determines the minimum observed failure cardinality, including failures with full evidence and cases with no observed failure after exhaustive testing. An independent audit tool tests whether agents seek missing facts and then withhold prohibited actions while completing permissible tasks. The benchmark reveals substantial variation across the ten models in omission robustness, evidence acquisition, and appropriate action. Resistance to missing evidence does not necessarily translate into useful recovery. Recovery access reduces unsafe execution for all ten models, but agents differ in whether they actively obtain the missing facts and use them to act appropriately in both worlds.
ATTUNE: Agentic Throughput Tuning for LLM Training	2026-10-22	Show The throughput of an LLM training job depends on its recipe: how the model is split across GPUs, how memory is traded for compute, and which kernels and communication overlaps it uses. The best recipe changes with every model, accelerator and framework version. Automated tuners search a fixed space of knobs and check only that a job runs, so they miss levers outside that space and can ship a configuration that silently changes training. ATTUNE is a coding agent that tunes any working recipe unattended within a GPU-hour budget. Guided by a skill of search rules, scripts and lessons from past tuning campaigns, it diagnoses each run and changes one lever at a time. A correctness gate outside the agent refuses any argument known to change the training objective. ATTUNE tunes Amazon's production post-training recipes on H100, H200 and B200 clusters, raising throughput by 1.7x on average. It beats the published throughput of four of NVIDIA's own benchmark recipes on NVIDIA's AutoModel framework by 6-17%. On models held out from its development, it improves the strongest public recipes most where they were written for another setting, by up to 2.6x on MiniMax-M2, and least where the vendor tuned them for the same hardware. Given the same GPU-hour budget, it outperforms existing automated tuners, including Bayesian optimization, random search and OptiCo, a recent LLM-agent tuner. Its gain on GLM-4.5 is more than four times the best alternative's. The same coding agent without ATTUNE's skill gains less and fails half its launches. We release ATTUNE at https://github.com/amazon-science/attune and OpenRecipes, a leaderboard of tuned and verified training recipes, at https://openrecipes.ai.	24 pa... 24 pages, 2 figures. v2: adds NVIDIA NeMo AutoModel benchmark recipes, production deployment results and the OpenRecipes leaderboard
Can an Agent Debug a Web Corpus? Agentic Auditing of Slovak FineWeb2	2026-10-22	Show Multilingual pipelines such as FineWeb2 apply one recipe across a thousand languages, leaving the failure modes of individual languages unanticipated. We use an LLM once, as a corpus debugger rather than a classifier: a tool-using agent explores the corpus through a read-only, budgeted API and compiles each finding into an executable test, which a native speaker then validates. On the 26.5M-document Slovak portion of FineWeb2, three runs costing $2.21-$2.84 each found seven findings over four phenomena. The largest is a machine-generated content farm of 1,408,437 documents (5.32%) that passed the pipeline's filters, 86.6% of it passing even stricter thresholds. Blinded labels support 6 of the 7 findings (defect rates .53-1.00 against .08 in a random control) and do not support one. The compiled suite scans the full corpus in 0.7 s of wall-clock time without a GPU, against $12k-$90k per LLM classification pass. The best static sampling baseline validates more selectors than any agent run, at under half the cost. The agent contributes exact prevalence, a precision estimate checked against sampled documents, an account of why the pipeline missed each slice, and a reusable test. We release the framework, Slovak test suite, validation data and agent trajectories.	Accep... Accepted to WaC Workshop at EMNLP 2026
TabAgent: Distilling LLM Agent Components into Cheap Classifiers	2026-10-22	Show Agentic systems, AI architectures that autonomously execute multi-step workflows to achieve complex goals, are often built using repeated large language model (LLM) calls for closed-set decision tasks such as routing, shortlisting, gating, and verification. While convenient, this design makes deployments slow and expensive due to cumulative latency and token usage. We propose TabAgent, a framework for replacing generative decision components in closed-set selection tasks with a compact textual-tabular classifier trained on execution traces. TabAgent (i) extracts structured schema, state, and dependency features from trajectories (TabSchema), (ii) augments coverage with schema-aligned synthetic supervision (TabSynth), and (iii) scores candidates with a lightweight classifier (TabHead). On the long-horizon AppWorld benchmark, TabAgent maintains task-level success while eliminating shortlist-time LLM calls, reducing latency by approximately 95% and inference cost by 85-91%. Beyond tool shortlisting, TabAgent generalizes to other agentic decision heads, establishing a paradigm for learned discriminative replacements of generative bottlenecks in production agent architectures.	67 pa... 67 pages, 14 figures, 26 tables. v2 is a substantial revision with a new title and two added authors. The evaluation is extended from AppWorld to seven agent benchmarks across three decision families (set retrieval, classification, policy and reward), and TabHead is now a CatBoost classifier with an ablation over classical and neural heads
Contestable Agents for Data Governance: Recording Human Stewardship of AI-Generated Metadata Across Structured and Unstructured Systems	2026-10-22	Show Industrial organizations hold operational knowledge in structured records, such as telemetry, point lists, and transactions, and in unstructured sources, such as manuals, service notes, communications, and inspection reports. Governance programs often manage these sources separately. Classifying, linking, and documenting them depends on a small group of employees with deep domain knowledge. Agents built on large language models can propose metadata, cite documents, and flag quality issues, but unreviewed proposals can spread errors into downstream systems. A perfunctory approval leaves no record of what a reviewer checked. We present contestable governance: specialist agents produce scored, cited arguments for disputed metadata decisions; a steward can accept, reject, modify, or add arguments before commit; and a governance record retains the evidence, interventions, and outcome. A versioned domain glossary supplies model-specific terms and acronym definitions to the agents. We describe the workflow, a triage path that reserves full argumentation for uncertain proposals while preserving review before commit, the scoring and record models, and a gate that controls which verified decisions become precedent. Two HVAC use cases illustrate the pattern: point classification using Brick Schema class labels, supported by an open reference implementation, and an illustrative warranty failure-mode decision linking claim fields and telemetry to service notes and revisioned bulletins.	15 pa... 15 pages, 2 figures, 2 tables
Same Luck, Different Choice: Twin-World Credit Assignment for Agentic Reinforcement Learning	2026-10-22	Show Agentic reinforcement learning rewards a complete episode, yet the policy needs credit for every decision inside it. Counterfactual branching estimates this credit by re-running the episode after an alternative decision, but the branch draws its sampling noise independently of the factual episode, so the outcome difference mixes the effect of the decision with the luck of everything that follows. We observe that in agentic search the retriever, the corpus, and the grader are deterministic, so an episode is a deterministic function of the policy's decisions and its sampling noise. Twin worlds are therefore realizable: the same noise, a different decision. We introduce TWICE (TWIn-world Credit Estimation), which keys the Gumbel noise of every token by turn and line, rolls out counterfactual twins that inherit the factual episode's downstream noise, and turns their paired outcome differences into turn-level advantages for policy-gradient updates on actor-sampled trajectories; only the actor is deployed. Twin credit is unbiased for any coupling quality, its variance is that of an independent counterfactual scaled by $1-ρ$, where $ρ$ is the twins' outcome correlation, and under coupling the number of rollouts needed to resolve an advantage of size $Δ$ falls from $Θ(Δ^{-2})$ to $Θ(Δ^{-1})$. A square-root rule spends the twin budget where outcomes are predicted to disagree per unit cost. On seven question-answering benchmarks, TWICE improves the average exact match of GRPO by 5.3 and 4.5 points on Qwen2.5-3B and 7B, outperforms tree-search and counterfactual-branching baselines, and uses 68% fewer auxiliary tokens than VinePPO.
Semantics Delivery Network: Rethinking Web Retrieval Infrastructure for LLM Agents	2026-10-22	Show Large language models (LLMs) increasingly ground their answers in proprietary data and live web content through both single-shot retrieval-augmented generation (RAG) and multi-turn agentic search. Yet the web path they use was built for human readers: search services return URLs and snippets ranked by generic relevance, and content delivery networks (CDNs) cache URL-addressed objects without knowing which passage an agent needs. LLMs instead consume short, semantically coherent passages - hereafter, chunks - selected for downstream utility rather than similarity alone, retrieve statefully across reasoning turns, and, as uncoordinated streams, repeatedly fetch, clean, chunk, and embed the same content. We argue that semantic chunk retrieval should become a first-class network-delivery abstraction, and propose the Semantics Delivery Network (SemDN): an origin-authorized, hierarchical substrate that indexes, searches, and caches web content at chunk granularity, serving clients and origins alike. SemDN is more than finer-grained caching: it must estimate missing or stale evidence without an explicit miss signal, let clients choose retrieval representations that origins need not see, and share state among clients with the same representation. Preliminary results show that page-first retrieval processes two to three orders of magnitude more content than agents consume, that agentic queries exhibit task-local chunk reuse invisible to URL caches, and that chunk delivery improves answer quality per context token. We organize an open research agenda around shareable retrieval state, hierarchical caching and refresh under coverage risk, and deployability, including incentives, privacy, and verifiability. Our code is available at: https://github.com/sarendis56/SemDN_HotNets26	Accep... Accepted and to appear at the 25th ACM Workshop on Hot Topics in Networks (HotNets 2026). 13 pages, 3 figures
AgentVQA: A Unified Benchmark for Agentic Visual Understanding	2026-10-22	Show Vision-language models (VLMs) can perform a broad range of tasks across diverse settings, yet their performance in agentic contexts remains poorly understood. Although online evaluations in simulators and real-world environments are essential for measuring interactive agent performance, they are costly and challenging to scale. Complementary offline VQA benchmarks can enable faster and more reliable comparisons across agentic domains, yet existing benchmarks remain limited and fragmented. To address this gap, we introduce AgentVQA, an offline benchmark for evaluating VLMs on realistic agentic scenarios across five agent domains: Web Agents, Robotics, Egocentric Videos, Games, and Spatial Planning. We build AgentVQA by manually selecting 1000 challenging, real-world agent examples from 23 source datasets and adapting them for reliable offline evaluation. The resulting benchmark preserves realistic scenarios from interactive environments across diverse domains, and its model rankings track those from corresponding online evaluations. Across the 22 evaluated models, the best achieves only 58.5% accuracy. Our cross-domain analysis shows that errors are driven primarily by failures to ground and maintain accurate representations of the current and evolving visual state, including localizing relevant UI elements, interpreting motion, tracking state changes over time, and inferring 3D spatial relationships, rather than by task misinterpretation or downstream logical reasoning.	Proje... Project page: https://zlab-princeton.github.io/AgentVQA
Adaptive Skill Granularity for Coding Agents: Runtime Routing and Macro-to-Micro Recovery	2026-10-22	Show Reusable skills allow large language model agents to package task-specific procedures, tool affordances, and execution guidance into modular units. Prior work has largely studied how agents retrieve, compose, or synthesize skills before execution. In coding agents, however, execution poses an additional runtime control problem: the same task may be solved through primitive tool calls, macro skills, or hybrid execution that starts with a macro and falls back to micro-level recovery. We introduce Adaptive Skill Granularity, a framework that treats skill granularity as an action-abstraction policy conditioned on task structure, execution state, and macro-skill failures. The system combines a feature-based granularity controller and a finite-state recovery executor. We evaluate the framework with a deterministic coding-agent protocol using verifier-defined success, total cost, and process metrics such as macro adoption, failure recovery, route match, and transition correctness. Results show that fixed macro execution is efficient when applicable but brittle under mismatch, while runtime routing and explicit macro-to-micro recovery improve robustness and recover controlled macro failures that fixed-granularity execution leaves unresolved. Our findings suggest that skill libraries for coding agents should be treated not only as retrievable resources, but also as controllable, verifiable, and recoverable execution abstractions.