跳到正文
研踪PaperTrail

论文库

检索计算机领域论文,按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+197新增论文
+151其中 AI 相关
+16已有论文修订
+91新增 AI 导读

arXiv 最近完成扫描:2026-11-22 08:44 · 全站最近入库:2026-11-21 12:19各信源进度 →

统计口径与数据新鲜度

统计 2026-11-21 10:37 — 2026-11-22 10:37(北京时间) · 点击数字查看对应简报;AI 相关包含在新增论文内。

最近完成检查:2026-11-22 08:44最近发现新增版本:2026-11-21 12:19库内论文最新更新时间:2026-11-20 02:59

每分钟刷新已保存的数据。空检查不代表采集异常;新收录可能包含历史论文。AI 导读按新增版本计数。

29,158 已收录论文24,181 AI 相关8,992 有修订112 HF 社区发现

浏览论文库

时间范围跟随排序:版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常,查看状态 →

找到 21,490 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-22 08:42(北京时间)
版本更新 · v3arXiv:2602.08616

Breaking the Grid: Distance-Guided Reinforcement Learning in Large Discrete Action Spaces

Heiko Hoppe, Fabian Akkerman, Wouter van Heeswijk, Maximilian Schiffer

原文摘要

Reinforcement Learning (RL) is increasingly applied to large-scale decision-making problems like logistics, scheduling, and recommender systems, but existing algorithms struggle with the curse of dimensionality in such large discrete action spaces. We propose Distance-Guided Reinforcement Learning (DGRL), combining Sampled Dynamic Neighborhoods and Distance-Based Updates to enable efficient RL in problems with up to $10^{36}$ actions. Unlike prior methods, DGRL performs stochastic volumetric exploration and transforms policy optimization into a stable regression task, decoupling gradient variance from action space cardinality. On structured tasks, DGRL provably guarantees local value improvement. DGRL naturally handles action spaces without regular semantic structure and hybrid continuous-discrete action spaces. We demonstrate performance improvements of up to 66% against state-of-the-art benchmarks across regularly and irregularly structured environments, while simultaneously improving convergence speed and computational complexity.

cs.LG · cs.AI

版本更新 · v2arXiv:2605.21699

X-Token: Projection-Guided Cross-Tokenizer Knowledge Distillation

Sharath Turuvekere Sreenivas, Adithyakrishna Venkatesh Hanasoge, Mingyu Yang, Ali Taghibakhshi, Saurav Muralidharan, Ashwath Aithal, Pavlo Molchanov

原文摘要

Cross-tokenizer knowledge distillation allows a student to learn from teachers with incompatible vocabularies while preserving its tokenizer and inference architecture. Existing logit-based methods either supervise only the observed token or match full distributions through strict token partitions and heuristic ranking. We identify two limitations: an uncommon-token failure, where task-critical tokens outside the shared vocabulary are suppressed, and over-conservative matching, where exact string equality discards useful correspondences. We propose X-Token, which addresses both through a sparse projection initialized from tokenizer-level string rules and optionally refined during distillation. P-KL removes the partition and projects the student distribution into the teacher vocabulary; H-KL retains the hybrid loss while relaxing matching through the projection. A token-coverage audit guides the choice between them, and both extend naturally to heterogeneous teachers. Across Llama, OLMo, and Gemma students, X-Token achieves the highest average accuracy in all four teacher-student settings among all evaluated cross-tokenizer methods. Complementary teachers provide a further 1.30-point gain, and both variants are faster than the evaluated cross-tokenizer baselines in matched profiling.

cs.LG · cs.CL

版本更新 · v2arXiv:2604.10479

Expanding Codes with Polynomial Stretch

Isaac M Hair, Amit Sahai

原文摘要

We give a polynomial-stretch, constant-arity counterpart to the expanding-code construction of Blasiok, Lou, Rosen, and Sudan (FOCS 2026). For every constant $\varepsilon\in(0,1/2)$, we construct binary matrices with $n$ columns and polynomially many rows, each of weight $k=O(\varepsilon^{-6})$, such that the union of the supports of any set of at most $n^{1-\varepsilon}$ rows has size at least $(1-\varepsilon)k$ times the number of rows. Nevertheless, a polynomial-time test accepts every codeword (generated by the columns of this matrix) corrupted in at most a $1/2-\varepsilon$ fraction of its coordinates, while rejecting a uniformly random vector with high probability. We give two applications. First, we obtain a family of (nearly lossless) constant locality expanders on which Goldreich's pseudorandom generator with the XOR-MAJ predicate can be distinguished from uniform in polynomial time, well within the small-bias parameter regime and the hard regime for known algebraic-immunity/annihilator attacks and algebraic-degree linearization attacks described by Applebaum and Lovett (STOC 2016, SICOMP 2018). Second, we obtain public key encryption from either the constant-noise $k$XOR conjecture over random graphs or the Goldreich conjecture with XOR-MAJ over random graphs, together with a new planted-hypergraph conjecture. We prove low-degree lower bounds for the planting conjecture, providing evidence for subexponential hardness, and give lower bounds against dimension tests based on coordinatewise products (Schur products) of public-code vectors. The PKE schemes and lower bounds were discovered by the authors. The code construction was discovered by Codex / ChatGPT 6 Astra Ultra using harness components from the UCLA Moonshot Harness Project, in response to a general strategy suggested by the authors. See the end of the introduction for a discussion.

cs.CR

版本更新 · v3arXiv:2603.00963

Stabilizing Policy Optimization via Logits Convexity

Hongzhan Chen, Tao Yang, Yuhua Zhu, Shiping Gao, Xiaojun Quan, Ting Yao

原文摘要

While reinforcement learning (RL) has been central to the recent success of large language models (LLMs), RL optimization is notoriously unstable, especially when compared to supervised fine-tuning (SFT). In this work, we investigate the stability gap between SFT and RL from a gradient-based perspective, and show that the convexity of the SFT loss with respect to model logits plays a key role in enabling stable training. Our theoretical analysis demonstrates that this property induces favorable gradient directionality during optimization. In contrast, Proximal Policy Optimization (PPO), a widely adopted policy gradient algorithm utilizing a clipped surrogate objective, lacks this stabilizing property. Motivated by this observation, we propose Logits Convex Optimization (LCO), a simple yet effective policy optimization framework that aligns the learned policy with an optimal target derived from the original RL objective, thereby emulating the stabilizing effects of logits-level convexity. Extensive experiments across multiple model families show that our LCO framework consistently improves training stability and outperforms conventional RL methods on a broad range of benchmarks.

cs.LG · cs.CL

新论文 · v1arXiv:2611.10423

Mental Imagery That Matters: Curing Latent Collapse in Multimodal Reasoning

Mi Luo, Rohan Dugad, Zihui Xue, Alex Dimakis, Kristen Grauman

原文摘要

Mental simulation, the ability to reason by constructing and updating internal visual states, is central to flexible human cognition, but remains elusive in current multimodal large language models. Latent multimodal reasoning offers a promising route toward this capability by interleaving text with continuous visual latent blocks, allowing models to maintain internal visual workspaces without decoding them into pixels. In this paper, we ask whether these latent visual states actually matter. Through a direct perturbation analysis across representative latent-reasoning MLLMs, we find that they often do not: dropping, replacing, or shuffling generated visual latents has only minor effects on the model's own final-answer probability, while equivalent perturbations to the linguistic reasoning trace are substantially more damaging. We call this failure mode Mental Simulation Collapse (MSC): the model appears to perform interleaved visual reasoning, but routes most answer-relevant computation through text. We further diagnose that MSC follows naturally from existing training objectives: visual alignment can make generated latents resemble helper-image embeddings, but it does not require the decoder to rely on them; subsequent text-only supervision further leaves the semantic role of the latent pathway unidentifiable. To address this, we introduce AnchorSim (Anchored Mental Simulation), a contrastive distillation framework that trains models to make their generated visual latents consequential. Across multiple visual reasoning benchmarks, AnchorSim improves performance, while diagnostic re-evaluation shows that visual perturbations now meaningfully affect predictions. Our results expose a fundamental failure mode of latent multimodal reasoning and demonstrate that mental simulation must be anchored to model behavior in order to matter.

cs.CV · cs.AI

版本更新 · v3arXiv:2605.25326

Perceive-then-Plan: Layout-as-Policy for Monocular 3D Scene Layout Estimation

Junwei Zhou, Yu-Wing Tai

原文摘要

Building structured 3D scene layouts from a single image requires reconciling visual observations with physical and spatial constraints, a challenge that is difficult to address with direct prediction alone. In this work, we formulate monocular 3D layout estimation as a perceive-then-plan problem with vision-language models, where a Perceiver first grounds the 3D objects and then a Planner iteratively refines the scene hypothesis through actions that improve physical plausibility while preserving consistency with the input image. We propose Layout-as-Policy (LaP), which casts the planning stage as a policy learning problem: 3D layouts are represented as structured states, and refined via discrete actions such as translation, rotation, and rescaling. Starting from an observation-aligned initialization with the geometry-enhanced Perceiver, the LaP Planner is trained to produce action sequences that progressively resolve geometric inconsistencies and enforce realistic spatial relations. To enable effective learning, we combine supervised trajectory initialization with preference-based optimization, allowing the model to learn corrective behaviors without requiring explicit reward engineering. This formulation transforms layout estimation from a one-shot prediction task into an iterative refinement process, enabling better handling of global constraints and complex object interactions. Experiments demonstrate that our approach produces layouts that are more physically coherent and better aligned with visual observations, while naturally supporting downstream tasks such as scene editing and manipulation.

cs.CV

版本更新 · v2arXiv:2607.07625

Embedded Blockchain Infrastructure Management (eBIM): A RISC-V-Empowered Hardware--Software Co-Design Framework Towards Trustworthy Blockchain

Qinglin Yang, Yuan Liu, Yaoyao Zhang, Boya Wang, Zongjian You, Chunming Rong, Zhihong Tian

原文摘要

Blockchain systems are shifting from decentralized ledgers for digital assets to general-purpose trust infrastructures for verifiable computation, decentralized physical resources, and automated infrastructure management. However, deploying these systems through third-party Blockchain-as-a-Service (BaaS) offerings may introduce dependencies involving operational control, confidentiality, and infrastructure governance. Meanwhile, blockchain execution and verification impose demanding requirements on cryptographic processing, isolation, and reproducibility. RISC-V offers an open, extensible instruction set architecture that enables co-design of software execution environments and domain-specific security hardware. Inspired by the embedded identity and provisioning model of an embedded subscriber identity module (eSIM), this survey proposes embedded Blockchain infrastructure management (eBIM) as a conceptual hardware-software co-design framework that organizes existing RISC-V-enabled technologies into a hardware-anchored blockchain infrastructure stack. We systematically review RISC-V-based smart-contract execution, trusted execution environments, zero-knowledge virtual machines, and cryptographic acceleration, and analyze their relationship to blockchain requirements, including deterministic execution, verifiability, extensibility, and secure key management. Based on this evidence, we distinguish currently demonstrated capabilities from the proposed eBIM integration vision and identify major open challenges, including system-level benchmarking, blockchain-oriented cryptographic extensions, TEE security, post-quantum migration, and deployable hardware-software co-design. The survey provides an architectural reference for researchers and designers investigating open and verifiable blockchain infrastructure.

cs.CR · cs.AR

新论文 · v1arXiv:2611.10418

Beyond Query Similarity: Safety-Conditioned Cache Routing for LLM Response Reuse

Shilong Pan, Zhiliang Tian, Weiguo Chen, Zihan Chen, DongSheng Li

原文摘要

Semantic response caching reuses a cached response for a semantically similar query to improve LLM inference efficiency. However, query similarity alone can lead to unsafe reuse when safety conditions change. Existing approaches improve cache safety through stricter similarity checks, context-aware validation, or response verification, yet they may still miss safety conflicts between the cached response and the new request or introduce unnecessary verification overhead. We argue that safe and efficient semantic caching requires a reuse policy that makes the cached response's safety conditions explicit and uses them to guide the reuse decision with a balance between safety and utility. In this paper, we propose SafeCache, a safety-aware framework that separates candidate retrieval from safety approval and formulates semantic response reuse as a safety-conditioned routing problem. SafeCache matches each cached-response condition against the new request and aggregates the per-condition mismatches into an overall risk score. A separate contradiction guard prevents explicit conflicts from being diluted by strong matches. It further estimates matching uncertainty and calibrates the routing policy under a required unsafe-reuse target and a verifier-call budget to choose among direct reuse, selective verification, or regeneration. Experiments show that SafeCache achieves the better safety-utility trade-off among the evaluated baselines, enabling safer response reuse at comparable cache utility.

cs.AI

新论文 · v1arXiv:2611.10416

S2HAND: Accuracy-Preservation Trade-Offs for Hand Specialization in Whole-Body Pose Models

Singyu Muk, Xuanren Song, Kalin Stefanov

原文摘要

Unified whole-body pose estimators jointly predict body, face, and hand landmarks, but often degrade on small, blurred, or occluded hands. Attaching a standalone hand specialist introduces extra detection overhead, coordinate alignment, and wrist-stitching artifacts. We introduce S2HAND, a hand-specialized Sapiens2-5B framework that preserves the native 308-keypoint output interface. Specifically, S2HAND-MT adapts a lightweight 0.87M-parameter pose branch to route 40 finger heatmaps while strictly preserving the shared Transformer, auxiliary task heads, and the other 268 body, foot, and face heatmaps. S2HAND-Pose further fine-tunes three Fisher-selected Transformer blocks (267.13M parameters), trading exact preservation for stronger hand recovery: it lifts macro AUC from 57.83% to 65.73% across HInt TEST, COCO WholeBody VAL, and FreiHAND EVAL, and surpasses HInt-adapted Hand5/HRNet and stitched Hand5 systems on all five Native Pipeline benchmarks. On COCO, S2HAND-Pose improves hand AP by 12.89 points and body-plus-foot AP by 0.19 points. Across a 512-image multi-task audit, it maintains 99.60% foreground segmentation agreement with the public checkpoint, with quantified normal and pointmap drift. Its 2D coordinates also lower 3D mesh error in frozen Pose2Mesh transfer, while our optional TailGuard re-cropping module cuts S2HAND-MT's severe failure rate on NewDays from 13.45% to 10.43%. S2HAND thus resolves the whole-body hand bottleneck with either exact non-hand preservation (MT) or empirically audited trade-offs (Pose).

cs.CV

新论文 · v1arXiv:2611.10414

Track or Follow? A Jev-Compatible Multi-Agent Trading System When Web News Lags the Market

Yebo Feng, Guangyan Gan, Liangrui Pan, Ruichao Liang, Cong Wu, Jiahua Xu, Yang Liu

原文摘要

LLM agents increasingly trade on the Web news they read, yet much Web news about listed firms is published after, and about, the price move it reports, so a reader can agree with the market by following it instead of reading the event. We ask whether LLM judgments of Web news track the market's reaction to the reported event or follow it. To answer, we built a multi-agent trading system whose agents share only point-in-time stores and answer typed questions through the interface of Jev, a hosted typed decision model; code sets every position under a proved bound on how far text can move it. In a hash-committed test of a Jev-compatible reader built from a pinned LLM's option probabilities, item judgments track announcement-window abnormal returns (weekly Spearman 0.119), also after the LLM's release, and event types agree with the issuers' 8-K filings (AUC 0.833); Jev's judgments of news published after its own release track the reaction as well (descriptive). Exploratory analyses show where the association comes from: 72% of earnings news items appear after the issuer's filing, contributor-written items often report the price move, and from the first trade after publication the association falls to 0.023. At a weekly horizon, the confirmatory test detects no rank information over code-only features. Under attack, decision cards traced 88% of successful attacks to the planted item. We propose that market-graded tests of Web-text readers report the criterion from publication, not only from the event.

cs.CL · cs.AI · cs.CR · cs.IR · q-fin.TR

新论文 · v1arXiv:2611.10413

Frequency-Structure Complementary Modeling for Hyperspectral Image Fusion

Song Qiya, Tian Siyu, Li Shutao, Dian Renwei, Sun Long, Liu Jinyang

原文摘要

Existing methods mainly rely on convolutional aggregation or attention-based spatial-spectral interaction. However, conventional feature-domain attention does not explicitly characterize variation patterns along spatial-token and learned latent-channel coordinates, while spatially shared local operators lack the flexibility to adapt neighborhood aggregation to heterogeneous structures. To address these limitations, we propose a Frequency-Structure Complementary Modeling Network (FSCM-Net), which couples frequency-aware dependency learning with structure-adaptive local refinement. Specifically, a Spatial-Spectral Frequency Calibration Attention (SSFCA) module constructs spatial-token affinities from Fourier-interacted query-key representations and builds latent-channel affinities after input-conditioned frequency filtering along learned channel coordinates. The latent-channel response is then calibrated by a learned shared gain and combined with the spatial-token response. Complementarily, a Structure-Coherent Dynamic Feature Fusion (SC-DFF) module combines regularized structure-tensor cues from MSI features with context-modulated HSI detail representations to generate channel- and position-dependent aggregation kernels. This design makes local HSI refinement responsive to smooth regions, edges, and textures while maintaining spectral fidelity. The two modules are progressively integrated within a multiscale encoder-decoder to coordinate representation-level interaction and structure-adaptive reconstruction. Experiments on two simulated benchmarks and one real remote-sensing dataset demonstrate favorable performance against representative fusion methods in spatial detail reconstruction and spectral preservation.

cs.CV · cs.AI

新论文 · v1arXiv:2611.10411

Tail-risk-aware Agent Safety: Constraining Tail Risks Across Multi-step LLM Agent Workflows

Shilong Pan, Zhiliang Tian, Zihan Chen, Wentao Zou, Weiguo Chen, DongSheng Li

原文摘要

LLM-based agents extend LLMs with tool calls, memory, planning, and retrieval to act in external environments, but they are vulnerable to attacks at different workflow steps. Existing defenses, including prompt engineering and post-training alignment, primarily improve LLM-only safety and often overlook risks that arise from multi-step execution. External verification adds guardrails to check each step separately, but it typically ignores how risks can propagate across steps. We argue that we should consider risk transitions across dependent steps through the agent workflow, rather than separate verification on a single step. In this paper, we propose \textbf{T}ail-risk-aware \textbf{A}gent \textbf{S}afety (TAS), a tail-risk-aware safety framework that models agent workflow safety as a sequence of dependent safety states and constrains the rare but extreme risks (i.e., tail risk) with a shared budget. TAS first models safety-state transitions sequentially along the workflow as an MDP to trace risk propagation, then measures a tail-sensitive risk to target rare but high-cost risks (i.e., tail risk), and constrains the risk with a shared budget to derive a step-wise safety strategy. By constraining tail risk rather than the overall average, TAS suppresses rare high-cost failures while preserving utility in normal cases. Experiments show that TAS achieves better performance compared with eleven baselines across three benchmarks.

cs.AI

新论文 · v1arXiv:2611.10410

SimDuoReal: Object-Centric Bimanual Policy for Emergent Dexterity

Qiyang Yan, Yulin Liu, Akash Sharma, Haozhi Qi, Guanya Shi, C. Karen Liu

原文摘要

Bimanual dexterous manipulation requires coordinated arm and hand control, yet specifying such coordination manually is challenging and often task-specific. In this work, we explore whether complex bimanual dexterity can emerge from a simple object-centric learning objective. We train reinforcement learning policies under a shared object-centric reward, varying their training environments rather than tuning separate rewards for each behavior. We find that the policies discover pre-grasp manipulation, grasping and regrasping, in-hand adjustment and handover to reach object-pose goals without hand-designed motion sequences. We then refine the experts on transition states and use DAgger and reinforcement learning to distill them into a single policy that composes these behaviors for diverse bimanual tasks. On real hardware, the same policy succeeds in 61 of 80 trials (76.3\%) across four tasks with unseen objects and no task-specific fine-tuning. In simulated barcode scanning, our policy achieves 85.0\% success, compared with 60.4\% for two single-arm policies without handover. These results suggest that this objective provides a simple and scalable interface for learning general bimanual dexterity, shifting the burden from manually designing robot coordination to specifying what should happen to the object.

cs.RO

版本更新 · v2arXiv:2605.23892

Good Token Hunting: A Hitchhiker's Guide to Token Selection for Visual Geometry Transformers

Shuhong Zheng, Michael Oechsle, Erik Sandström, Marie-Julie Rakotosaona, Federico Tombari, Igor Gilitschenski

原文摘要

Visual geometry transformers have become powerful architectures for multi-view 3D reconstruction, enabling joint prediction of multiple 3D attributes in a feed-forward manner. However, their computational cost grows quadratically with the input sequence length due to the global attention layers inside these models. This limits both their scalability and efficiency. In this work, we address this challenge with a simple yet general strategy: restricting the number of key/value tokens that each query interacts with during global attention. To achieve effective token selection, we introduce a two-stage framework. First, an inter-frame selection step operates at the frame level to identify frames that should be preserved. Second, an intra-frame selection step further discards more redundant tokens within the selected frames. Our analysis highlights the advantage of a diversity-based strategy for inter-frame selection, which ensures broad coverage of the scene. For intra-frame selection, we show that layer-aware sparsification is necessary, with the selection process guided by the entropy of the global attention pattern. Our approach offers a superior speed-accuracy trade-off compared to existing solutions. Extensive experiments show that it accelerates visual geometry transformers by over 85% for scenes with 500 images while maintaining, or even improving, baseline performance, which hints that how our token selection strategy can play a crucial role in future applications of visual geometry transformers. Our project website is available at https://zsh2000.github.io/good-token-hunting.github.io.

cs.CV · cs.AI · cs.GR · cs.LG · cs.RO

版本更新 · v5arXiv:2509.03738

Mechanistic Interpretability with Sparse Autoencoder Neural Operators

Bahareh Tolooshams, Ailsa Shen, Anima Anandkumar

原文摘要

We introduce sparse autoencoder neural operators (SAE-NOs), a new class of sparse autoencoders that operate in function spaces rather than fixed-dimensional Euclidean representations. We formalize the functional representation hypothesis, where data are explained through sparse compositions of structured functions. Unlike standard SAEs that represent concepts with scalar activations, SAE-NOs parameterize concepts as functions, enabling representations that capture not only a concept's presence, but also how and where it is expressed across the input domain. We achieve this through joint sparsity: concept sparsity selects active concepts, while domain sparsity governs where they are expressed. We instantiate this framework using Fourier neural operators (SAE-FNOs), parameterizing concepts as integral operators in the Fourier domain. This functional and spectral parameterization is particularly advantageous when data exhibit spatial structure across scales or when concepts are frequency-structured. We characterize SAE-FNO on vision data and demonstrate that it learns localized patterns, uses concepts more efficiently, and exhibits stable concept characteristics across sparsity levels. We further show that SAE-FNO adapts to changes in domain size and generalizes across discretizations, operating at resolutions beyond those seen during training, where standard SAEs fail. We also introduce lifting into SAEs and show theoretically and empirically that it acts as a preconditioner that accelerates optimization. Overall, our results show that moving from vector-valued to functional parameterizations, with concept and domain sparsity, extends SAEs from representing concept presence to modeling structured concept expression, highlighting the importance of parameterization.

cs.LG · cs.AI · eess.SP · stat.ML

新论文 · v1arXiv:2611.10406

Early Signals of Problematic AI Use: The Longitudinal Roles of Motivation and Emotional Engagement

Jina Suh, Constance Hadley, Koustuv Saha

原文摘要

Longitudinal evidence about the psychological impact of conversational AI use remains scarce. We report a survey study of 114 adults who reported their level of problematic AI use one year after reporting their motivations (for escape, social, entertainment, and instrumental reasons) for engaging with AI. We found high levels of AI motivation of all types predicted later problematic AI use, especially when participants turned to AI for managing their feelings and overcoming difficulties. Our findings imply that AI motivation could be used as an early warning signal of potential problematic use, especially if individuals engage with AI for emotional purposes over time.

cs.HC · cs.AI · cs.CY

新论文 · v1arXiv:2611.10405

Yuanbao-Reranker: Document-Grounded Synthesis and Preference Distillation for Text Reranking

Hongyuan Xu, Dezhi Ye, Jiahao Cao, Jie Liu, Haijin Liang, Haibo Shi, Jin Ma

原文摘要

High-precision text reranking is a critical component in modern search, recommendation, and agentic retrieval-augmented generation (RAG) systems. We present Yuanbao-Reranker, a family of general-purpose dense rerankers spanning 0.6B, 4B, and 8B parameter scales built upon the Qwen3 backbone. To construct diverse and reliable ranking supervision at scale, we pair answer-first and query-type-guided document-grounded synthesis with a three-stage curation funnel that verifies positive support, eliminates semantic redundancy, and enforces balanced source-by-length quotas. For each curated pair, we screen real retrieval pools with a teacher score margin to filter out potential false negatives and sample Hard, Medium, and Moderate negative tiers, yielding 2.75M $1{:}3$ ranking groups. We train the models by jointly optimizing pointwise relevance cross-entropy and a bounded minimum-gap hinge objective over both positive--negative and negative--negative pairs. This formulation transfers fine-grained teacher preferences across candidate tiers without over-constraining well-separated predictions, supporting both cross-scale and same-scale distillation followed by layerwise adaptive model merging. Across five comprehensive benchmarks spanning Chinese (CMTEB-R), multilingual (MMTEB-R), English (BEIR), long-document (MLDR), and code (MTEB-Code) retrieval, Yuanbao-Reranker consistently leads every parameter tier: the 0.6B variant reaches 70.96 Avg. to surpass 1B--4B active-parameter baselines, the 4B variant reaches 73.63 Avg. to outperform the 8B teacher (72.68), and the 8B variant achieves a state-of-the-art 74.21 Avg., with substantial gains generalizing across distinct first-stage retrievers.

cs.IR

新论文 · v1arXiv:2611.10400

Detecting and Explaining Robot Policy Failures from Sparse Internal Features

Riad Ahmed, Sujosh Nag, Moniruzzaman Akash, Momotaz Begum

原文摘要

Visuomotor policies can handle many robot manipulation tasks, but understanding their failures remains difficult. Unfamiliar conditions and small action errors can lead to failure, yet watching the robot alone tells us little about what went wrong inside the policy. We introduce MIND (Mechanistic Interpretability for Neural Diagnosis), a method that looks inside a trained policy to detect failures during execution and identify the internal concepts involved. MIND uses a sparse autoencoder to break down the policy's internal activations into features that can be studied individually. We give meaning to these features by linking their activity to the robot's position, movement, and surroundings, and by steering their activations to examine how they affect the robot's actions. From successful and failed executions, we learn how the features behave on the way to a failure. During execution, MIND flags deviations from these patterns and uses the meaningful features contributing most to the change to describe the problem. Experiments in simulation and on a physical robot show that changes in these internal features help detect and explain failures. By using the same features for detection and diagnosis, MIND connects failure signals to understandable changes inside the policy.

cs.RO

版本更新 · v2arXiv:2605.27788

Knowing When to Ask: Segment-Level Credit Assignment for LLM Tool Use

Abhijit Kumar, Zoey Wu, Mohit Suley

原文摘要

Humans know when to reach for help e.g. $347 \times 28$ warrants a calculator while $2+2$ does not. Language models do not. Prompt-based approaches can instruct a model when to invoke tools, but this scaffolding does not teach it to recognize the boundary of its own knowledge. RL approaches that assign a single outcome reward to the whole trajectory fare no better: trajectory-level credit cannot isolate which tool call in a successful episode actually helped, nor penalize unnecessary calls. We propose CARL (Competence-Aware Reinforcement Learning), which trains a critic on the model's own rollouts to learn where parametric knowledge suffices and where it needs external help. By decomposing each rollout at natural tool-use boundaries (e.g., code fence delimiters and context block transitions), CARL assigns independent credit to each segment from a single binary outcome, without external judges or step-level annotations. As a result, erroneous tool calls, incorrect extractions, and unnecessary calls each receive appropriately signed advantages. The trained critic captures the model's domain competence: it separates parametrically solvable from tool-dependent questions with AUC 0.93 at 7B. On five benchmarks spanning arithmetic, multi-hop factual QA, and numerical reasoning over financial tables, CARL improves average exact-match accuracy on the three search benchmarks by 6.7 points at 7B and 9.6 points at 3B over the best RL baseline, with the largest gain (+8.3 EM at 7B, +12.5 EM at 3B) on MuSiQue. At 7B, CARL has a 53% lower tool-use rate than Search-R1 PPO on parametrically answerable questions while remaining 9.8 EM points more accurate on them. Gains are largest at small scale: the 3B improvement is $1.4\times$ the 7B improvement, suggesting that knowing when to ask disproportionately benefits models with smaller parametric memory.

cs.LG · cs.CL

版本更新 · v2arXiv:2605.13716

SkillOps: Managing LLM Agent Skill Libraries as Self-Maintaining Software Ecosystems

Xinyuan Song, Hongji Pu, Liang Zhao

原文摘要

Large language model agents increasingly rely on skill libraries for multi-step tasks, yet these libraries can accumulate persistent defects as skills are added, reused, patched, and linked to changing dependencies. We call this failure mode skill technical debt: library-level defects that may not break a single skill locally but can harm future retrieval, composition, and execution. Existing skill-based agents mainly focus on task-time retrieval, planning, and repair, while library-time maintenance remains underexplored. We propose SkillOps, a method-agnostic plug-in framework for maintaining skill libraries. SkillOps represents each skill as a typed Skill Contract (P, O, A, V, F), organizes skills with a Hierarchical Skill Ecosystem Graph, and diagnoses library health across utility, compatibility, risk, and validation dimensions. Given a raw skill library, SkillOps produces a maintained library that can be used by existing retrieval or planning agents without changing their internal code. On ALFWorld, SkillOps achieves 79.5 percent task success as a standalone agent, outperforming the strongest baseline by 8.8 percentage points with no additional task-time large language model calls. As a plug-in layer, it improves retrieval-heavy baselines by 0.68 to 2.90 percentage points. The current rule-based maintenance implementation uses nearly zero library-time large language model calls or tokens, showing that skill-library maintenance can be added as a low-overhead architectural layer.

cs.SE · cs.MA

版本更新 · v2arXiv:2610.41191

When AI Investigators Cannot Trust Their Tools: Conflict-Aware Evidence Verification for Cyber Threat Investigation

Yuchen Zhang

原文摘要

LLM agents increasingly triage security alerts by combining the outputs of several detection tools, implicitly treating those tools as trustworthy. Unlike prior work on conflicts between a model's knowledge and its context, we study conflicts among the tools themselves: what happens when this machine-generated evidence is conflicting or wrong. On the corrected CIC-IDS2017 dataset, four heterogeneous detectors emit labels and confidences that we corrupt under a threat model varying the number of corrupted tools, the persistence of the corruption, and the confidence attached to wrong outputs. Across open-weight LLMs from 3B to 14B parameters, unverified investigators are fragile: one corrupted detector out of four lowers naive accuracy to .50-.72, three lower every unverified investigator to .13 or less, and honest disagreement alone makes them abstain on up to 63% of flows. We propose Conflict-Aware Evidence Verification (CAV), built from established components: an equal-weight vote that ignores reported confidences, an evidence-conflict gate, and an independent nearest-neighbour trust check that tests the leading candidates. With its prompt frozen before testing, CAV-LLM improves on the naive investigator in every corrupted condition ($p < 0.001$), reaching .84-.88 with three corrupted detectors against at most .08, while calling the LLM on 0.4% of clean traffic. A fixed verification rule is as accurate or better unless most tools are corrupted. When we make the independent channel itself noisy, CAV inherits its errors: it stays ahead when two or more tools are corrupted, but adding the channel as an extra voter is better when at most one is. Verification thus moves trust from the tools to the check, which must be measured and protected in turn.

cs.CR · cs.LG

新论文 · v1arXiv:2611.10395

The Limits of Information-Flow Control Enforcement for Agentic Systems

Jonathan Gregory, Kyle Domico, Kunyang Li, Jean-Charles Noirot Ferrand, Patrick McDaniel

原文摘要

Information-flow control (IFC) is a promising approach to enforce policy in agentic AI systems. However, IFC suffers from label creep, a phenomenon where monotonically increasing security labels propagate through a system and progressively limit its utility. Label creep in agentic AI systems can be mitigated using two major strategies: (1) improved system design and (2) label declassification. In this paper, we characterize how far system design alone can mitigate label creep in agentic systems. Specifically, we measure how system topology, scale, and agent composition affect utility loss and security enforcement cost proxies for conservative IFC enforcement in a stateful agentic system simulator. We develop a pipeline to systematically vary these features and measure their impacts on label propagation. Our results demonstrate strong associations between structural graph features and our IFC utility loss metric (up to a Spearman correlation $|ρ|$ of $0.96$) and show that system size and tool allocation substantially affect how labels propagate. These findings identify design choices to reduce IFC utility loss, but they also demonstrate that system design alone cannot fully prevent utility loss due to monotonic label propagation. This motivates the development of label declassification mechanisms that aid these design choices and diminish IFC utility cost.

cs.CR · cs.AI

新论文 · v1arXiv:2611.10393

Finite Time Analysis of Risk-Sensitive RL via Noisy Power Iteration

Waqar Mirza, Yashaswini Murthy, Laixi Shi, Eric Mazumdar, Adam Wierman

原文摘要

Estimating the principal eigenpair of a positive operator from stochastic evaluations gives rise to a noisy form of normalized power iteration: each step applies a sampled operator estimate and then renormalizes. We establish finite-time stochastic-approximation guarantees for this procedure, working in Hilbert's projective metric, which is well-suited to the multiplicative geometry of Perron--Frobenius eigenproblems. Our main application is risk-sensitive average-cost reinforcement learning (RL) with exponential utility. In this setting the Bellman equations are multiplicative, and value functions are characterized by nonlinear eigenvalue problems rather than additive fixed points. We cast both policy evaluation and control in this framework, obtaining risk-sensitive TD- and Q-learning algorithms that learn from a single Markovian trajectory. Under explicit positivity, mixing, and linear function approximation assumptions, we prove finite-time bounds on the recovered eigenvector and eigenvalue with $\tilde{O}(ε^{-2})$ sample complexity up to problem-dependent constants. The results cover both tabular and linear function approximation regimes, and through the duality between exponential utility and KL-robust control yield finite-time guarantees for KL-robust average-cost RL as well.

cs.LG

版本更新 · v2arXiv:2604.03787

Dimension-Free Sinkhorn for Optimal Transport: Bypassing Extremal Costs

Kun He

原文摘要

The Sinkhorn--Knopp (SK) algorithm is a cornerstone method for matrix scaling and entropically regularized optimal transport (EOT). Despite its empirical efficiency, existing bounds for achieving marginal accuracy $\varepsilon$ deteriorate severely in the presence of extreme entries, bottlenecked either by the global maximum regularized cost $η\|C\|_\infty$ (where $η$ is the regularization parameter and $C$ the cost matrix) or the matrix's minimum-to-maximum entry ratio $ν$. These bounds may be overly sensitive to rare extreme entries that do not dominate the overall matrix structure. We identify a broad structural regime in which the dependence on the magnitude of global extreme entries can be circumvented. For EOT, we introduce well-boundedness, a local mass property that captures the row- and column-wise bulk structure of the underlying matrix while maintaining robustness against scattered entrywise outliers. We prove that, governed by this structural condition, SK converges to the target transport plan for a problem of dimension $n$ in $O(\log n + \log(1/\varepsilon))$ iterations. This complexity bound is independent of the maximum regularized cost $η\|C\|_\infty$. A computationally negligible pre-scaling step eliminates this dimensional dependence, reducing the iteration complexity to a strictly dimension-free $O(\log(1/\varepsilon))$. Beyond EOT, we establish a sharp phase transition for general $(\boldsymbol{u},\boldsymbol{v})$-scaling governed by a critical matrix density threshold. We prove that when a matrix's density exceeds this threshold, the iteration complexity is strictly independent of $ν$. Conversely, when the density falls below this threshold, the dependence on $ν$ becomes unavoidable; in this sub-critical regime, we construct instances where SK requires $Ω(n/\varepsilon)$ iterations.

cs.DS · cs.LG

版本更新 · v3arXiv:2602.11674

Benchmark Health Index: A Systematic Framework for Benchmarking the Benchmarks of LLMs

Longyuan Zhu, Hairan Hua, Linlin Miao, Wei Wang, Bing Zhao

原文摘要

Large Language Models (LLMs) are advancing rapidly, yet the benchmarks used to measure this progress are becoming increasingly unreliable. Score inflation and selective reporting have eroded the authority of standard benchmarks, leaving the community uncertain about which evaluation results remain trustworthy. We introduce the Benchmark Health Index (BHI), a pure data-driven framework for auditing evaluation sets along three orthogonal and complementary axes: (1) Capability Discrimination, measuring how sharply a benchmark separates model performance beyond noise; (2) Anti-Saturation, estimating remaining headroom before ceiling effects erode resolution and thus the benchmark's expected longevity; and (3) Impact, quantifying influence across academic and industrial ecosystems via adoption breadth and practice-shaping power. By distilling 106 validated benchmarks from the technical reports of 91 representative models in 2025, we systematically characterize the evaluation landscape. BHI is the first framework to quantify benchmark health at a macro level, providing a principled basis for benchmark selection and enabling dynamic lifecycle management for next-generation evaluation protocols.

cs.AI

新论文 · v1arXiv:2611.10392

Tiresias: Towards Language-Conditioned World Models

Fengyi Wu, Yifei Dong, Yetong Sha, Lingdong Kong, Qiyu Hu, Bohan Xiong, Qi Dai, Alexander G. Hauptmann 等 9 位作者

原文摘要

World models enable planning through internal simulation, while natural language provides intuitive task specification. Yet, language-conditioned world modeling remains fragmented across domains with disparate action spaces, datasets, and evaluation protocols. We introduce LCWM, a shared task that maps an initial agent-view observation and a language instruction to an action sequence that fulfills the instruction. Its accompanying benchmark, LCWM-Bench, integrates 15 datasets across navigation, robotics, 3D gaming, and driving under a shared action interface and evaluation protocol, pairing 142,987 episodes with 428,961 instructions in three styles. We further propose Tiresias, an agent that plans via selective imagination: a language-conditioned diffusion policy generates candidate trajectories, a trajectory ranker shortlists them, and a video-pretrained world model imagines their visual consequences. A visual ranker evaluates these rollouts, complementing trajectory scores to select the final plan. Tiresias achieves competitive planning and imagination performance on LCWM-Bench. Ablations demonstrate the complementary benefits of trajectory ranking and visual verification, while experiments with alternative proposal methods show that the trained rankers remain effective beyond our diffusion policy. Together, LCWM-Bench and Tiresias provide a unified evaluation setting and an effective starting point for studying language-guided planning through world models.

cs.CV

版本更新 · v3arXiv:2602.05214

Disentangled Representation Learning via Flow Matching

Jinjin Chi, Taoping Liu, Mengtao Yin, Ximing Li, Yongcheng Jing, Jialie Shen, Leszek Rutkowski, Dacheng Tao

原文摘要

Disentangled representation learning aims to capture the underlying explanatory factors of observed data, enabling a principled understanding of the data-generating process. Recent advances in generative modeling have introduced new paradigms for learning such representations. However, existing diffusion-based methods encourage factor independence via inductive biases, yet frequently lack strong semantic alignment. In this work, we propose a flow matching-based framework for disentangled representation learning, which casts disentanglement as learning factor-conditioned flows in a compact latent space. To enforce explicit semantic alignment, we introduce a non-overlap (orthogonality) regularizer that suppresses cross-factor interference and reduces information leakage between factors. Extensive experiments across multiple datasets demonstrate consistent improvements over representative baselines, yielding higher disentanglement scores as well as improved controllability and sample fidelity.

cs.LG

新论文 · v1arXiv:2611.10390

A Computational Map of the Disaster Mental Health Research Landscape

Soheil Shapouri, Saber Soleymani

原文摘要

When disasters strike, they not only damage infrastructure, buildings, and the economy, but also affect the mental health of affected populations. Research on the mental health effects of disasters has expanded rapidly and has become an important, well-funded area of psychology, public health, and related social sciences. However, the overall structure of this research landscape has not been systematically characterized. Using large language models, we conducted a review of all articles related to terms 'disaster' and 'mental health' indexed in PubMed. 7,859 articles' metadata were retrieved using the NCBI API, and then Claude Sonnet was used to extract structured variables from the abstracts. The results indicated that earthquakes and epidemics were the most commonly studied disasters while many other disasters such as heat and cold waves have received less attention. PTSD, depression, and anxiety were well investigated but many other mental health issues (e.g., aggression, suicide) are relatively ignored. Despite enjoying relatively large sample sizes (median = 452, IQR = 142 - 1,568), the field rarely utilizes machine learning algorithms. Hypothesis significance testing, reviews, qualitative methods, and structural equation modeling are respectively most common methods while agent-based modeling, machine learning, casual inference, and NLP each appear in less than 1% of articles. Studies are limited to a small number of countries, particularly the United States, Japan, and China, while many disaster-prone regions remain substantially underrepresented. This LLM-assisted computational map provides a broad overview of the disaster mental health literature while highlighting important methodological gaps, with particular emphasis on the underutilization of computational methods that could help advance the field.

cs.AI · cs.CY · cs.DL

版本更新 · v2arXiv:2610.22100

When the next byte's type is fixed: why entropy patching skips computed values under tight budgets

Nicolás Vera Zúñiga

原文摘要

The Byte Latent Transformer (BLT) runs a large model once per patch and starts patches where a small model's next-byte entropy is high. Under tight budgets this rule skips computed values, such as the number after "= " in math. On BLT-1B at inference we trace the miss to the cue before a value: when the preceding bytes fix the next byte's type, the patcher's entropy is capped near log2 10 bits, close to a 10% budget's threshold. Entropy then starts a patch at 0% of values after a dollar sign and 19% after "= "; writing "is " instead of "= " raises the 19% to 66%. Version 1's two-byte lookup table, motivated by LZ novelty, acts as a learned detector of cues that tend to precede new values in its fitting text: on the solutions rewritten as prose it covers every value after a dollar sign and none after a plain space, and its margin over entropy falls from +42.1 to -8.9 points, partly because prose gives entropy more starts per value. In BLT-style models trained from scratch (1M-50M parameters, 128-byte context, 3 seeds per rule) and scored by teacher-forced exact match on answer spans, the table beats entropy at 50M by 36.8 points on computed results and 29.6 on final answers, which mostly copy the last computed result. A hand-written marker rule is not distinguishable from the table at 1M and 12M; its 50M control was not run. At 1M and 12M the table ran on a T4 GPU in fp16 and the baselines on an A40 in bf16; at 50M all ran on the A40. End to end, no 1M-50M model solves more than 2 of 100 GSM8K problems. A label-free probe of BLT-1B's patcher finds prose values, mostly after a dollar sign, but costs final answers. In six domains, no rule we tested beat entropy on exact match of whole words and numbers in more than one domain. The positive findings concern tight budgets on cue-formatted text. This version corrects version 1; the work is closed, with no venue intended.

cs.CL · cs.LG

新论文 · v1arXiv:2611.10389

From Task Affinities to Optimal Allocations: Probabilistic Graphical Modelling for Supervised Finetuning

Prateek Chanda, Saral Sureka, Krishnateja Killamsetty, Ganesh Ramakrishnan

原文摘要

Supervised fine-tuning over a large pool of heterogeneous tasks is often computationally expensive, leading practitioners to train language models on a fixed training budget of carefully chosen SFT tasks. Thus, achieving strong downstream performance hinges on selecting an effective subset of tasks, or equivalently, identifying an optimal task mixture under a fixed budget constraint. In practice, supervised finetuning mixtures are often determined using simple heuristics (e.g., uniform or size-proportional sampling) that ignore task interactions, which can hurt transfer and waste budget on redundant task sources. We introduce TASKMIXPGM, a framework for learning continuous task mixtures through energy-based modelling over SFT tasks. Each task forms a node in a Markov random field, with unary potentials capturing task utility and pairwise potentials modelling inter-task relationships via behavioral divergences between predictive distributions of single-task fine-tuned models. Optimizing this objective yields task mixtures that balance utility, coverage, and redundancy. We further show that the induced set function defined over the set of such tasks is weakly submodular under budget constraints, enabling approximation guarantees for discrete task selection. Empirically, we provide a fixed, benchmark-independent mixture that improves over heuristic mixing across multiple model families and different training budgets. TASKMIXPGM also uncovers interpretable structure in task interactions, revealing complementarities and redundancies that inform effective SFT mixtures.

cs.LG