跳到正文
研踪PaperTrail

版本更新

查看已发布论文的最新修订,在详情页对照实际观测到的版本。

部分采集异常

全站更新概况

+814新增论文
+671其中 AI 相关
+98已有论文修订
+94新增 AI 导读

arXiv 最近完成扫描:2026-11-19 06:59 · 全站最近入库:2026-11-19 06:59各信源进度 →

统计口径与数据新鲜度

统计 2026-11-18 14:16 — 2026-11-19 14:16(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:2026-11-19 06:59最近发现新增版本:2026-11-19 06:59库内论文最新更新时间:2026-11-18 08:00

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。

26,871 已收录论文22,375 AI 相关8,156 有修订112 HF 社区发现

论文修订列表

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 8,156 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-19 06:57(北京时间)
版本更新 · v4arXiv:2510.00279

SLogic: Subgraph-Informed Logical Rule Learning for Knowledge Graph Completion

Trung Hoang Le, Tran Cao Son, Ishtiaq Ahmed, Huiping Cao

原文摘要

Logical rule-based methods offer an interpretable approach to knowledge graph completion (KGC) by capturing compositional relationships in the form of human-readable inference rules. While existing logical rule-based methods learn rule confidence scores, they typically assign a global weight to each rule schema, applied uniformly across the graph. This is a significant limitation, as a rule's importance often varies depending on the specific query instance. To address this, we introduce SLogic (Subgraph-Informed Logical Rule learning), a novel framework that assigns query-dependent scores to logical rules. The core of SLogic is a context-aware scoring function. This framework is in line with the specificity principle in commonsense reasoning. This function determines the importance of a rule by analyzing the subgraph locally defined by the query's head entity, thereby enabling a differentiated weighting of rules specific to their local query contexts. Extensive experiments on benchmark datasets show that SLogic achieves competitive performance among rule-based methods. It also generates query-dependent, human-readable logical rules that serve as explicit explanations for its inferences.

cs.LG · cs.AI

版本更新 · v2arXiv:2610.41672

Understanding the Energy Cost of Quantized Agents on Shared GPUs

Stephen Abkin, Alex Flesher, Aldric Yang, Nitheesh Kodarapu, Deepak Panigrahy, David Kebo Houngninou

原文摘要

Quantization can lower the energy of language-model calls and free memory for serving more agents, but its benefit per completed task remains unclear when agents share a GPU. Measurements of fixed requests or isolated agents leave open how the agents' generated work interact with competition for compute and cache. We present a systematic study of precision and admission limits for two agent models on A100 deployments with 40 or 80 GB of memory, using coding and data-analysis tasks with independently verified outcomes. We replay recorded calls across precisions to isolate execution energy and cap the cache to test the value of freed memory. For both models with four admitted agents where requests fit in the cache, 4-bit weights with 16-bit activations execute the same calls with 10-25% less energy than 8-bit weights and activations. Despite this saving, extra work and fewer completions in live runs raise their energy per verified completion by 49%. Where requests compete for cache on the 40 GB deployment, 4-bit weights cost less, and capping their cache raises energy per verified completion by 34% with 16 agents. These results show how generated work and cache demand change the energy benefit of quantization, providing a practical basis for choosing precision and admission limits together.

cs.PF

版本更新 · v2arXiv:2608.00232

Real-Time Visual Obstruction Detection in Surgical Augmented Reality

Shih-Chin Yang, Yanming Xiu, Hanting Ye, Qi Chen, Elias Rotondo, Maria Gorlatova

原文摘要

Surgical augmented reality (AR) can provide contextual guidance by overlaying virtual annotations, tool cues, and procedural information onto the surgical workspace. However, the virtual content may obstruct task-relevant real-world information, such as surgical instruments, and interfere with users' perception during time-sensitive surgical tasks. In this paper, we investigate visual obstruction detection for surgical AR and present a latency-aware pipeline that combines vision-language model (VLM)-based surgical-object recognition with segmentation-based obstruction reasoning. To reduce inference overhead, the system adopts a cascaded small-to-large VLM architecture with segmentation-guided early exiting and attention-based visual token pruning. The small VLM handles easy frames when its key-object prediction is supported by segmentation consistency, while difficult frames are forwarded to a large VLM with pruned visual tokens. We construct a pseudo-AR surgical obstruction detection benchmark by overlaying virtual content onto surgical-tool images and labeling whether the virtual content obstructs task-relevant instruments. Evaluation results show that the proposed system achieves 87.43% obstruction detection accuracy with an average end-to-end latency of 479 ms, reducing latency by 62.90% compared with a cloud large-model baseline. These results demonstrate the feasibility of latency-aware obstruction detection for surgical AR and motivate future work on dynamic surgical videos, multi-object scenes, and clinically grounded AR guidance content.

cs.CV

版本更新 · v3arXiv:2503.01036

Data-Efficient Kernel Methods for Learning Differential Equations and Their Solution Operators: Algorithms and Error Analysis

Bamdad Hosseini, Alexander Hsu, Yasamin Jalalian, Juan Felipe Osorio Ramirez, Houman Owhadi

原文摘要

We introduce a novel kernel-based framework for learning differential equations and their solution maps. The framework is efficient in its data requirements in terms of the number of observed states and the amount of data available per state. Our formulation also allows us to employ efficient Gauss-Newton-type optimization algorithms in the training procedure to improve accuracy and speed of the training procedure. We also support our algorithm with rigorous theoretical guarantees in the form of quantitative worst-case error bounds for the learned equation that can be extended to the subsequent solution operators under smoothness assumptions. Numerical benchmarks demonstrate significant improvements in computational complexity and robustness while achieving one to two orders of magnitude improvements in terms of accuracy compared to certain state-of-the-art algorithms.

stat.ML · cs.LG · math.NA

版本更新 · v2arXiv:2305.07271

Complexity of conjunctive regular path query homomorphisms

Laurent Beaudou, Florent Foucaud, Florent R. Madelaine, Lhouari Nourine, Anantha Padmanabha, Gaétan Richad

原文摘要

A graph database is a digraph whose arcs are labeled with symbols from a fixed alphabet. A regular graph pattern (RGP) is a digraph whose arcs are labeled with regular expressions over the alphabet. RGPs model navigational queries for graph databases called \emph{conjunctive regular path queries} (CRPQs). A match of a CRPQ in the database is witnessed by a special navigational homomorphism of the corresponding RGP to the database, where vertices of the RGP are mapped by a function $f$ to the vertices of the database, and for each arc $uv$ of the RGP, we require the existence of a walk in the database from $f(u)$ to $f(v)$, such that the word corresponding to the sequence of labels of that walk belongs to the regular language labeling the arc $uv$. We extend this definition to one of navigational homomorphisms between two RGPs, modeling a homomorphism-based form of CRPQ containment. We study the complexity of deciding the existence of such a homomorphism between RGPs. We show that this problem, in the case where general walks are considered, can still be solved by an EXPSPACE algorithm (just like general CRPQ Containment, which is known to be EXPSPACE-complete). In the case of simple paths or trails, in fact the navigational homomorphism problem considered here becomes much more tractable, as it is PSPACE-complete. We also study the problem for restricted RGPs over a unary alphabet, that arise from some applications like XPath or SPARQL. For this case, homomorphism-based CRPQ containment is in NP. We prove that certain interesting cases are in fact polynomial-time solvable.

cs.DB · cs.DM · cs.FL · cs.LO

版本更新 · v2arXiv:2610.40946

Opening the Box Changes the Cat: LLMs Detect Chain-of-Thought Monitoring - What now?

Amir Jalilifard, Marcos Medeiros Raimundo, Anderson Rocha

原文摘要

Chain-of-thought (CoT) reasoning allows large language models (LLMs) to generate intermediate reasoning traces in natural language while solving complex tasks. These traces can support safety monitoring by exposing potentially harmful plans before actions are executed. In this work, we investigate the conditions under which LLMs infer the presence of CoT monitoring from direct and indirect environmental cues and examine how this inference affects their subsequent behavior. Our observations suggest that oversight is an interactive process: while a monitor inspects a model's reasoning, traces left by monitoring interventions can allow the model to infer that its thoughts are being observed and to adapt its behavior accordingly. We show that even subtle environmental cues, such as timing-related signals from timeouts and operation durations, can reveal the presence of monitoring and trigger models to adapt their behavior to evade oversight. Across repeated episodes and interaction loops, we further observe that models can form statistical estimates of the monitor's effectiveness and use these estimates to guide attempts to evade detection. These findings highlight the need to evaluate CoT-based oversight under adaptive conditions, where the monitoring process itself may influence the behavior it seeks to observe.

cs.AI · cs.LG

版本更新 · v2arXiv:2610.29201

Pareto Inefficiency of EFX

Sepideh Mahabadi, Prasanna Ramakrishnan, Jakub Tarnawski

原文摘要

We consider the problem of allocating indivisible goods to agents with additive valuations, under two classic notions of fairness and efficiency: envy-freeness up to any good (EFX) and Pareto optimality. While some of the earliest work on EFX showed that Pareto optimal EFX allocations may not exist if agents can have zero value for some goods, it has remained open whether the same is true when all values are positive. We present a simple instance in which every good has positive value to every agent, but no Pareto optimal EFX allocation exists. We then generalize to a family of instances showing that Pareto optimal $α$-EFX allocations may not exist for any $α> 2/3$.

cs.GT

版本更新 · v2arXiv:2608.07226

Dual-Node NVIDIA DGX Spark over Tailscale: A Remote-Access Testbed for Distributed LLM Training and Cyber-Threat-Intelligence Fine-Tuning

Vasanth Iyer

原文摘要

Practical evidence for multi-node training on desktop-class accelerators remains limited. This report presents a proof-of-concept deployment of distributed NanoChat pretraining across two NVIDIA DGX Spark systems, each with a GB10 Grace Blackwell system-on-chip and 128 GB of unified memory, administered remotely over a Tailscale mesh VPN and connected for training by a dedicated 200 Gb/s QSFP56 direct fiber link. PyTorch torchrun, DDP, and NCCL were configured with one process per node, a depth-20 NanoChat model, a device batch of 32 sequences per node, a 2,048-token context, and eight gradient-accumulation steps, giving 1,048,576 tokens per optimizer step for the 896.5-million-parameter model. The run took 4,980 steps over about four days, averaging 63.4 s of measured training time per step (about 16,540 tokens/s) and processing about 5.22 billion tokens. We document link configuration, container setup, interface binding, a step-zero evaluation bug that triggered NCCL timeouts, checkpointing, and troubleshooting lessons. We also built a cybersecurity fine-tuning dataset from 77 CISA advisories (338 training, 37 validation conversations) and compared a baseline SFT checkpoint against a CTI-augmented checkpoint on 17 held-out prompts with an Ollama-hosted LLM judge: CTI-specific categories improved while general-knowledge categories regressed, for a small overall change from 2.06 to 2.29 on a 0-10 scale. Version 2 adds a stage-2 fine-tune with a MITRE ATT&CK technique dictionary under a deterministic evaluation: in-dictionary ID-to-name lookup reaches 98.5%, yet none of 50 held-out IDs is named correctly, knowledge of held-out threat groups, software, campaigns, and vulnerabilities stays near zero, and general ability drops measurably. The study establishes feasibility rather than a scaling-efficiency claim.

cs.AR · cs.LG

版本更新 · v5arXiv:2504.20941

Conformal-DP: A Density-Aware Mechanism for Differential Privacy over Riemannian Manifolds via Conformal Transformation

Peilin He, Liou Tang, M. Amin Rahimian, James Joshi

原文摘要

Differential Privacy (DP) is being increasingly adopted for non-Euclidean data that lie on complex, high-dimensional manifolds. Existing DP mechanisms for manifold data consider geometric properties when calibrating privacy perturbations, but they largely fail to capture variations in data density within datasets, leading to biased perturbations and suboptimal privacy-utility trade-offs due to heterogeneous data distributions. In this paper, we propose a novel density-aware differential privacy mechanism on Riemannian manifolds, referred to as Conformal-DP, that leverages conformal transformations to calibrate perturbations based on local densities and to induce a density-balanced geometry. We prove that our mechanism satisfies $ε$-differential privacy on any complete Riemannian manifold under mild regularity assumptions. In addition, we derive a closed-form expected geodesic error bound that depends only on the underlying data density ratio and is independent of global curvature. Our empirical results on synthetic and real-world datasets demonstrate that the proposed Conformal-DP mechanism substantially improves the privacy-utility trade-off in heterogeneous data distribution settings, with worst-case performance comparable to state-of-the-art manifold DP mechanisms that assume uniformly distributed data.

cs.CR · math.DG · stat.OT

版本更新 · v3arXiv:2610.04898

A Systematic Analysis of the Predictive Power of LM Surprisal in Reading Chinese

Hongao Zhu, Muxiaoqiao Xu, Yikang Liu, Siyuan Song, Yuxia Wang, Byung-Doh Oh, Hai Hu

原文摘要

This study analyzes the predictive power of LM-derived, token-level surprisal on Mandarin Chinese reading times. Using a suite of Chinese-Pythia models (14M--1.4B) trained from scratch with 30B tokens, we examine how well surprisal predicts first fixation duration, gaze duration, and total reading time in three paragraph-level eye-tracking corpora of Mandarin Chinese (GECO-CN, HKP, and MECO). Contrary to previous null findings, our results show that surprisal is predictive of Chinese reading times. However, whether predictive power scales with model size and the amount of training is corpus-specific: bigger models predict reading times better in GECO-CN, whereas inverse scaling emerges in HKP and, at the largest sizes, in MECO. Subsequently, we find that checkpoints whose surprisal remains closer to $n$-gram statistics are better predictors of reading in HKP and MECO, providing one possible explanation for the inverse scaling behavior. All in all, the predictive power of surprisal on Chinese reading time measurements is corpus-specific, which cautions against drawing conclusions about language model scale from a single corpus.

cs.CL · cs.AI

版本更新 · v2arXiv:2609.21840

FPT iff PTIME for Homomorphism Problems on Sparse-Incidence and Bounded-Independence Patterns

Matthias Lanzinger

原文摘要

We show that fixed-parameter tractability and polynomial-time solvability coincide for homomorphism problems specified by classes of pattern hypergraphs of bounded incidence degeneracy or bounded primal independence number, assuming $FPT \ne W[1]$. In both cases, tractability is characterised by bounded fractional hypertree width. This generalises a result of Grohe [JACM 2007] that established such an FPT-PTIME collapse for bounded-arity structures. We build on the fractional balanced-separator framework and rounding theorem of Korchemna, Lokshtanov, Saurabh, Surianarayanan, and Xue [FOCS 2024] to prove near-linear bounds on fractional hypertree width ($fhw$) in terms of adaptive width ($adw$). Namely, for every hypergraph $H$ $$fhw(H)=O(μ(H) adw(H)\log adw(H)), \quad fhw(H)=O(\log α(H) adw(H)\log adw(H))$$ where $μ(H)$ is the degeneracy of the incidence graph of $H$ and $α(H)$ the independence number of the primal graph of $H$. As a consequence, we obtain a corresponding FPT-PTIME collapse for exact homomorphism decision and counting on every class in which either incidence degeneracy or independence number is bounded. This includes a number of natural classes such as pattern hypergraphs of bounded degree. Additionally, we observe that for every r.e. class of pattern hypergraphs, fixed-parameter tractability of the parameterised homomorphism problem implies quasipolynomial-time solvability of the corresponding unparameterised problem, assuming $FPT \ne W[1]$. Our results rely on the characterisation of fixed-parameter tractable decision of homomorphism problems in terms of submodular width of Marx [JACM 2013], conditional on ETH. In light of the very recent potential refutation of the ETH, we show that the characterisation of Marx holds also under the assumption that $FPT \ne W[1]$.

cs.CC · cs.DB

版本更新 · v4arXiv:2406.08375

Methods for Accelerating and Stabilizing Nonlinear Magnetic Equivalent-Circuit Analysis

Danial Kazemikia, Matthew Gardner

原文摘要

Design and optimization of electromagnetic devices require models that converge reliably and quickly and can represent intense localized magnetic saturation. This paper presents a parameterized 2D nonlinear magnetic equivalent circuit (MEC) and proposes solution methods for it: an approximate Jacobian that is symmetric positive definite, so each Newton-Raphson iteration can use sparse Cholesky factorization with fill-reducing reordering, and Anderson mixing as step control. These methods are compared with alternative nonlinear iterations, linear solvers, and step-control methods in convergence, computational cost, and prediction accuracy. Radial-flux magnetic gears (RFMGs) with structural bridges serve as a challenging test case, because their thin bridges exhibit intense local saturation. The methods are evaluated on three bridged base designs and on 1,024 bridged designs, each paired with the same design without the bridge. On the base designs, Newton-Raphson iteration with the approximate Jacobian reached much lower residuals than fixed-point iteration, which converged for none of them. Sparse Cholesky factorization with reordering was the fastest linear solver tested. Across the 1,024 design pairs, full-step Newton-Raphson converged for 193 bridged and 564 bridge-free designs, whereas Anderson mixing converged for 1,003 and 1,022 and required fewer iterations. Bridged designs converged less often and more slowly under every method. The nonlinear MEC was 2.17 and 1.84 times faster than fixed-mesh nonlinear finite element analysis (FEA) in total computation time for the bridged and bridge-free populations. The median outer-rotor torque discrepancy relative to FEA was 0.22%, and the Spearman rank correlations of the design rankings were at least 0.9996.

eess.SY · cs.CE

版本更新 · v2arXiv:2606.16988

Agent trajectories as programs: fingerprinting and programming coding-agent behavior

Hamidah Oderinwale

原文摘要

Benchmark scores tell you what an agent got right; they do not tell you how it got there. In this work, we introduce methods for comparing agents procedurally in different contexts, where the model, tasks, and approaches vary. We compare ten agents and find that they are identifiable by their behavioral habits, which we define as fingerprints: a probe over these procedural signatures attributes an unseen trajectory to the correct agent at 85.7% accuracy. We develop procedural representations for agent problem-solving procedures with an emergent vocabulary induction technique which is meant to be maximally compressive to avoid surface-level variation while being expressive enough to unveil the quirks of the models' patterns. We apply our framework to the software engineering evaluation dataset SWE-Bench to study the structural distinctness of agent trajectories and find that behavior is most similar between models from similar release periods and those that are distilled from one another (e.g., a distilled student model and its teacher have a Jensen-Shannon divergence of 0.25, about half the distance between other model pairs). As more models saturate evaluations, we believe that it will be important to probe model behavior along more holistic dimensions than success rates alone. We introduce ProcGrep, a library for auditing and evaluating agents for how they approach tasks at a procedural level given their traces in a top-down fashion. ProcGrep is available at https://github.com/hamidahoderinwale/procgrep.

cs.SE · cs.LG

版本更新 · v2arXiv:2609.01839

Import What You Need: Learning When and How to Augment EHR Graphs with External Knowledge

Chen Chen, Mohsen Nayebi Kerdabadi, Dongjie Wang, Mei Liu, Zijun Yao

原文摘要

Longitudinal prediction from electronic health records (EHRs) is limited by the sparsity and irregularity in patient trajectories, and knowledge augmentation with external knowledge graphs (KGs) offers a promising way to alleviate these issues. However, KG expansion often follows predefined rules, with limited control over whether and how to augment each visit as a patient's state evolves. We propose ReTA, a Reinforcement learning-based dynamic Topology Augmentation framework that casts KG import as a per-visit, budget-aware policy. ReTA first constructs an offline refined pool of KG-grounded templates, then learns a policy to select one augment action per visit from three options: Soft Import, which enriches node features without modifying graph topology, Hard Import, which grafts a compact KG subgraph onto the visit graph to create message-passing shortcuts, and Skip, which leaves the visit unaugmented when the base encoder is already confident. To stabilize learning, ReTA employs a decoupled encoder that processes semantic and structural signals in separate channels and fuses them via adaptive gating. Experiments on MIMIC-III and MIMIC-IV across diagnosis prediction, mortality, and readmission show that ReTA consistently outperforms strong baselines while remaining efficient, with larger gains under sparse supervision. ReTA transfers across datasets and knowledge graphs and yields interpretable augmentation patterns.

cs.LG · cs.AI

版本更新 · v2arXiv:2602.00950

MindGuard: Guardrail Classifiers for Multi-Turn Mental Health Support

António Farinhas, Nuno M. Guerreiro, José Pombal, Pedro Henrique Martins, Laura Melton, Alex Conway, Cara Dochat, Maya D'Eon 等 9 位作者

原文摘要

Large language models are increasingly used for mental health support, yet their conversational coherence alone does not ensure clinical appropriateness. Existing general-purpose safeguards often fail to distinguish therapeutic disclosures from genuine clinical crises, and progress on developing better ones is bottlenecked by a lack of clinically grounded evaluation resources. To address this gap, we introduce a risk taxonomy, developed in collaboration with PhD-level clinical psychologists, that identifies actionable harm (self-harm and harm to others) while preserving space for safe, non-crisis therapeutic content. We release MindGuard-testset, a dataset of multi-turn conversations annotated at the turn level by clinical experts, which is, to our knowledge, the first public benchmark with turn-level clinical risk labels for multi-turn mental health support. We also release an automated red-teaming (ART) framework designed to measure how safety classifiers affect downstream model behavior in adversarial multi-turn interactions. Using synthetic dialogues generated via a controlled two-agent setup, we train MindGuard, a family of lightweight safety classifiers (with 4B and 8B parameters). Our classifiers reduce false positives at high-recall operating points and, when paired with clinician language models, help achieve lower attack success and harmful engagement rates compared to general-purpose safeguards. We release all models, training and human evaluation data, and the ART framework.

cs.AI

版本更新 · v2arXiv:2604.01002

Query-Conditioned Evidential Keyframe Sampling for MLLM-Based Long-Form Video Understanding

Yiheng Wang, Lichen Zhu, Yueqian Lin, Yudong Liu, Jingyang Zhang, Hai "Helen" Li, Yiran Chen

原文摘要

Multimodal Large Language Models (MLLMs) have shown strong performance on video question answering, but their application to long-form videos is constrained by limited context length and computational cost, making keyframe sampling essential. Existing approaches typically rely on semantic relevance or reinforcement learning, which either fail to capture evidential clues or suffer from inefficient combinatorial optimization. In this work, we propose an evidence-driven keyframe sampling framework motivated by an information-theoretic perspective. We consider maximizing the conditional mutual information between the selected frames and the answer given the query as an idealized objective, and use a modular singleton-gain surrogate to motivate a tractable frame-wise selection criterion. Guided by this motivation, we train a query-conditioned evidence scoring network using annotated answer-supporting segments. A contrastive objective encourages greater aggregate score mass on frames from evidence regions than on background frames, providing a practical evidence-based criterion beyond semantic similarity. Experiments on long-form video understanding benchmarks show that our method consistently outperforms prior sampling strategies under comparable frame budgets. With Qwen2.5-VL-7B under the same 32-frame budget, our method improves over uniform sampling by 10.1 points on LVBench and 4.4 points on the long-video subset of VideoMME.

cs.CV · cs.AI · cs.LG

版本更新 · v2arXiv:2603.15457

Evasive Intelligence: Lessons from Malware Analysis for Evaluating AI Agents

Simone Aonzo, Merve Sahin, Aurélien Francillon, Daniele Perito

原文摘要

Tool-using artificial intelligence (AI) agents are increasingly adopted as systems that plan, observe their environment, and take actions over extended time periods. This shift challenges the current evaluation practice, in which models are tested in restricted, fully observable settings. We argue that such evaluations are vulnerable to a failure mode long known in computer security: a system that behaves benignly when it detects that it is being analyzed. Malware has exploited this for decades, by probing sandboxes, virtual machines, and instrumentation, before revealing its payload. The shared challenge is that the subject under observation can detect the conditions of its evaluation, so the setup used to observe behavior ends up shaping the behavior observed, yielding overly optimistic safety and robustness assessments. Recent AI-safety studies demonstrate pieces of this risk, though only preliminarily and under controlled conditions. Drawing on decades of malware-analysis experience, we outline concrete principles for evaluating AI agents as potentially adversarial: explicitly threat-modeling evaluator-discrimination; testing across harness variants and reporting the worst case; using shadow executions to provide a realistic runtime without irreversible side effects; treating environment probing as a risk signal; and performing unpredictable reassessments after deployment.

cs.CR · cs.AI

版本更新 · v2arXiv:2608.07786

Who Built This Model? Tracing LLM Lineage via Spectral Fingerprints in Weight Space

Yiwei Chen, Bingqi Shang, Sijia Liu

原文摘要

Open-weight large language models (LLMs) are increasingly developed through complex, multi-stage pipelines, leading to intricate lineage relationships that reflect model origin, ownership, and evolution. Understanding these relationships is important for model provenance, governance, and supply-chain integrity. In this work, we investigate the notion of LLM "biometrics" (analogous to human biometrics) to ask whether LLMs exhibit intrinsic fingerprints in weight space alone, without access to input data, that reveal their origin and lineage. We formulate this as a lineage discrimination problem, distinguishing among independent-origin, same-series, and shared-base models. To characterize these relationships, we propose a unified geometric fingerprinting framework that analyzes weight matrices from two complementary perspectives: (i) spectral energy, captured by singular value distributions to encode global magnitude patterns, and (ii) subspace alignment, quantified via subspace deviations to capture directional geometry. Our analysis uncovers a clear hierarchy of structural similarity in weight space: spectral energy reliably distinguishes independently trained models and different model families, while subspace alignment enables fine-grained discrimination among closely related models, including variations in dataset scale and post-training procedures. Extensive experiments on over 110 diverse open-weight LLM pairs demonstrate that weight-space geometry provides a robust and interpretable signal for model lineage, enabling coarse-grained regime separation and fine-grained discrimination within shared-base models.

cs.AI · cs.LG

版本更新 · v2arXiv:2607.16540

Do Vision Encoders Exhibit Human-like Color Thresholds?

Engy Ehab, Pablo Hernández-Cámara, Nahla Belal, Jesús Malo, Javier Vazquez-Corral, Alexandra Gomez-Villa

原文摘要

Understanding and characterizing human color perception is a longstanding research goal. One of the most traditional approaches is looking for the human color discrimination thresholds, the minimum chromatic differences perceptible to human observers. In recent years, deep neural networks have become the standard networks for computer vision tasks. In particular, deep vision encoders, foundation models trained on large-scale visual data, map images into latent feature representations. Despite the widespread use of deep vision encoders, few studies have investigated whether their internal representations exhibit human-like discrimination thresholds. In this work, we present a large-scale exploratory study probing the chromatic sensitivity of more than 50 pretrained vision encoders, including convolutional networks and vision transformers, against human discrimination thresholds. Using controlled chromatic stimuli at multiple chroma levels, we compare model-derived chromatic discrimination thresholds with human discrimination ellipses through a region-overlap metric (mIoU). Our analysis reveals generally weak alignment between model representations and human perceptual thresholds across all model families, with the best mIoU < 0.25. Moreover, we find that self-supervised encoders consistently outperform supervised ones, while language-supervised models show the most polarized behavior, occupying both the top and bottom of the ranking. These findings suggest that human-like chromatic sensitivity does not emerge naturally from current large-scale visual training objectives for any of the analyzed architectures.

cs.CV

版本更新 · v2arXiv:2602.05215

Temporal Grounding of Video Large Language Model via Holistic Event Perception and Matching

Jiahao Nie, Wenbin An, Gongjie Zhang, Yicheng Xu, Yap-Peng Tan, Alex C. Kot, Shijian Lu

原文摘要

Despite recent advances in Video Large Language Models (Vid-LLMs), Temporal Video Grounding (TVG) targeting precise localization of time segments of query events remains a grand challenge. Recent studies attempt to address this challenge by matching two separate tokens to all frame features and identifying the start and end frames of events based on the highest matching similarities. However, this approach relies heavily on the exact start and end timestamps, which are often ambiguous without capturing the semantic continuity and integrity of the event. To this end, we design E.M.Ground, a temporal grounding Vid-LLM that tackles the TVG challenge from a holistic and coherent event perception perspective. E.M.Ground introduces three novel designs: (i) It introduces a special <evt> token that aggregates information from all frames of query events with varying lengths, thereby preserving semantic continuity and integrity for accurate event matching. (ii) It introduces Savitzky-Golay smoothing to suppress the noise and abrupt changes in token-to-frame similarities across adjacent timestamps, thereby improving prediction accuracy. (iii) It aggregates multi-grained frame features to enhance the reliability of token-to-frame matching and improve temporal understanding, thereby compensating for the compression-induced information loss. Extensive experiments on widely adopted TVG benchmarks, along with in-depth analyses, demonstrate that E.M.Ground achieves strong overall performance, with the most pronounced improvements on single-event grounding and boundary-integrity tasks.

cs.CV

版本更新 · v2arXiv:2508.06765

Efficient Collaborative LLM Fine-Tuning over Heterogeneous Mobile Devices via Many Backbones to One Side-Network Tuning

Xingke Yang, Liang Li, Sicong Li, Liwei Guan, Hao Wang, Xiaoqi Qi, Jiang Liu, Xin Fu 等 9 位作者

原文摘要

Collaboratively fine-tuning (FT) large language models (LLMs) over heterogeneous mobile devices fosters immense potential applications of personalized intelligence. However, such a vision faces critical system challenges. Existing federated LLM fine-tuning approaches remain limited by prohibitive on-device resource overhead, straggler-prone synchronous aggregation, and the assumption of a unified pretrained backbone across heterogeneous devices. In this paper, we propose EC-MobiLLM, a novel design for efficient collaborative LLM fine-tuning across heterogeneous mobile devices and heterogeneous backbones. EC-MobiLLM adopts a server-assisted side-tuning paradigm to minimize on-device overhead and pioneers non-blocking asynchronous collaboration to accelerate training. Moreover, EC-MobiLLM introduces adaptive feature alignment to support collaboration across heterogeneous backbones, aligning with practical needs in real-world deployments. Extensive experimental results demonstrate that EC-MobiLLM can maintain robust fine-tuning performance while achieving extremely low on-device memory, with at least 95.2\% reduction in computation overhead, 93.2\% reduction in communication costs and $5.1\times$ faster convergence compared to existing methods, validating its efficacy for practical LLM adaptation over heterogeneous mobile devices.

cs.LG

版本更新 · v2arXiv:2609.10740

Governing AI Research Through Peer Review: A Mixed-Methods Study of the Longitudinal Effects of Ethics Flags Across Resubmissions

Kento Nishi, Alec Laprevotte, Isaiah Bullock, Mfoniso Andrew

原文摘要

Selective AI conferences have recently begun enforcing ethics flags and related review requirements, with the goal being to steer research towards safer and more responsible practices before publication. But do these requirements actually steer research as intended? In this paper, we show that authors more often revise how projects are presented following ethics flags than redirect their underlying research agendas. We first study the longitudinal effects of ethics flags by following rejected and withdrawn ICLR submissions with ethics flags into later public resubmissions, tracking manuscript changes after the ICLR review ends, when the original reviewers no longer oversee the project. We qualitatively code these resubmissions into five categories based on what changed after review and find that in 83% of 446 cases, authors leave the flagged concern unaddressed or revise the paper without changing the implicated methods or procedures. Then, we ask: if authors rarely change the research in response to ethics flags, what do they change instead? To answer this, we manually read reviews and rebuttals from 25 cases and directly interview authors about their rebuttal processes and resubmission decisions. We find that authors often concede concerns during rebuttal when reviewers can update their assessments, but drop those concessions after rejection when they do not regard the criticism as a sound reason to change the research. Interview participants describe publication changes as separate from changes to research direction, calling review an "editorial process" that shapes "what stories get seen" and, in another case, saying peer reviews are "mostly to filter out papers." Authors more readily change what they publish than what they study or build; we therefore recommend policy changes, especially disclosure of prior ethics flags upon resubmission so accountability carries over.

cs.CY

版本更新 · v2arXiv:2606.15123

CVE-Conditioned Exploit Generation with LLMs: A Data-Centric Study of Benchmarking and Fine-Tuning

Yiwei Chen, Lichi Li, Kai Cheung, Vinny Parla, Ganesh Sundaram

原文摘要

We study the task of CVE-conditioned exploit generation, where a model drafts proof-of-concept (PoC) exploits given software vulnerability context. We adopt a data-centric approach, constructing a high-quality dataset via multi-stage preprocessing and introducing a scalable evaluation framework with LLM-as-judge and fine-grained rubrics. Under this unified setup, we benchmark 17 large language models across 8 evaluation criteria, providing systematic insights into their zero-shot capabilities. We further show that a compact 8B open-weight model, when fine-tuned on curated data, achieves over 42.5% improvement in exploit quality and rivals some proprietary models when combined with simple test-time rejection strategies. Our results highlight the importance of data quality, structured supervision, and evaluation design for reliable exploit generation, suggesting that these factors can be as critical as model scale in adapting LLMs to cybersecurity tasks.

cs.CR · cs.LG

版本更新 · v2arXiv:2609.24492

Identity-Consistent Analysis of Long-Shot Windsurfing Video: A Domain-Specific Offline Tracking System

Bertil Braun, Felix Muehlenberend

原文摘要

Long-shot windsurfing video combines small targets, large camera pans, prolonged overlaps, and rapidly changing backgrounds. The desired output is not a generic MOT trace but a separate, stable rider-relative video for each surfer; one false identity merge can invalidate an otherwise useful result. We present an offline analysis system that detects surfers, forms conservative local tracklets, links them globally with camera-compensated motion and a foreground-masked sail-color descriptor, and uses two pose keypoints on the rig to drive a rider-relative virtual camera. The tracking stage is evaluated on 21 manually reconstructed development videos containing 41,004 retained observations. On this fixed-observation protocol, the production system achieves 0.957 pairwise precision, 0.918 recall, and 0.937 F1, compared with 0.792 F1 for OC-SORT and 0.828 for BoT-SORT. Compared with OC-SORT, it reduces fragmentation excess from 845 to 42, but nine of its 95 output tracks mix rider identities and these errors affect seven of the 21 videos.

cs.CV

版本更新 · v2arXiv:2609.04523

MaxKernel: Agentic Kernel Generation for TPUs

Shangkun Wang, Nina Cai, Charles Hoong, Julian Walker, Gerson Kroiz, George Vanica, Deepak Patil, Andi Gavrilescu 等 10 位作者

原文摘要

Designing and authoring high-performance custom kernels for accelerators is a complex task that requires deep hardware-level expertise. Large Language Models (LLM) can be leveraged together with real-time compiler feedback to build agentic systems for kernel generation. In this work, we present MaxKernel, a multi-agent system that implements three distinct paradigms for TPU kernel development: (1) a Human-in-the-Loop (HITL) agent for collaborative, step-by-step design; (2) an Autonomous (Auto) agent that executes a fully automated, metric/trace-driven optimization loop; and (3) a Graph-Based Autonomous Search that scales the Auto agent for global exploration of the design space. All three paradigms leverage a shared pool of specialized sub-agents to handle planning, implementation, self-debugging, testing, and hardware profiling. We evaluate MaxKernel on JaxBench, a comprehensive suite of 50 diverse kernel tasks for TPUs, alongside complex, real-world workloads from state-of-the-art open-source models. We demonstrate that MaxKernel consistently generates highly optimized implementations, matching expert hand-tuned baselines and delivering significant performance across the benchmark. Our agent is open-sourced and available https://github.com/AI-Hypercomputer/accelerator-agents/tree/main/MaxKernel.

cs.AI · cs.PF · cs.PL

版本更新 · v2arXiv:2607.20345

Closing the Lab-to-Store Gap: A Data-Efficient Post-Training and Experience-Driven Learning VLA Framework for Retail Humanoids

Roger Sala Sisó, Tiago Silvério, Jakob Sand, Tran Nguyen Le

原文摘要

Closing the gap between benchmark performance and reliable real-world operation remains a central challenge for Vision-Language-Action (VLA) humanoid robots, which must handle execution errors, distribution shifts, and environmental variability. This paper presents DEED (Data-Efficient Post-Training and Experience-Driven Learning), a systems-level approach evaluated on a supermarket chip-restocking task using a Unitree G1-Edu humanoid robot and the GR00T N1.6 foundation model. DEED comprises three key components: (1) a data-efficient post-training pipeline with control-frequency alignment, data curation, task-relevant visual highlighting, and reduced VLA dependence; (2) a real-world study of experience-driven refinement, adapted from RECAP via a text-based advantage prefix and a vision-language value function; and (3) a latent-space analysis tool for studying in- and out-of-distribution behavior. Our results suggest that bridging the lab-to-store gap is primarily a systems integration challenge rather than an architectural one: careful data design and targeted post-training can transform a policy that fails under naive fine-tuning into a competent real-world system using only a single GPU.

cs.RO · cs.AI

版本更新 · v3arXiv:2609.14892

An explicit solution of the five-expert prediction PDE and the exact optimality set of COMB

Jeff Calder, Nadejda Drenska

原文摘要

In this paper, we derive an explicit solution of the stationary prediction with expert advice PDE for five experts. The formula is given in three regions. In the first two regions, it is the four-expert solution plus a single integral with an elementary positive density. In the third region, it is a finite sum of hyperbolic products whose coefficients are determined by one scalar quadrature. Our formula establishes that the rank ordered adversary $(1,0,1,0,0)$ achieves the Hamiltonian maximum in the PDE everywhere in the ordered sector, while the COMB strategy $(1,0,1,0,1)$ is optimal only on a lower dimensional subset of the sector (where $x_1=x_2$ and $x_3=x_4$). The verification of the Hamiltonian inequalities is a tedious task, part of which is completed with a computer assisted proof. The verification reduces to 21 scalar inequalities, which we prove using 147 exact rational Bernstein polynomial certificates. The exact certificates and their independent arithmetic checks are included in a supplement to this paper, and a Lean 4 formalization machine-checks the verification and both main theorems, apart from the viscosity characterization.

math.AP · cs.GT · cs.LG · math.OC

版本更新 · v5arXiv:2607.22877

Towards Trustworthy Physical Intelligence: From Theory to Practice Across Life Cycle

Yang Wang, Hongxuan Liu, Xinghui Xu, Arjun Menon, Xiaoran Cai, Yunyu He, Alex Tarvo, Jingzong Zhou 等 43 位作者

原文摘要

Physical intelligence refers to intelligence systems that understand, reason about and act in accordance with the physical world and its underlying laws, dynamics, and constraints. Unlike conventional AI systems, physical intelligence interacts continuously with uncertain physical environments, and its actions produce consequences that are physically irreversible. As existing trustworthy AI frameworks have been developed primarily for digital AI systems, they do not fully capture the distinctive challenges of Physical Intelligence, such as physical safety, cyber-physical security, and physical manufacturing process. To address this gap, our paper develop a foundation for understanding, operationalizing, and governing trustworthy physical intelligence. First, we characterize the fundamental capabilities of physical intelligence. Second, we investigate the role of physical laws and principles in AI and analyze the associated challenges of physical intelligence. Third, we provide a unified structure of the end-to-end physical intelligence life cycle across five core stages and introduce the Trustworthy Physical Intelligence Operationalization (T-PAIO) framework. Fourth, we formulate the Trustworthy Physical Intelligence (T-PAI) Principles as a theoretical framework that synthesizes key trustworthiness dimensions and establishes a conceptual foundation for the governance of physical intelligence systems.

cs.AI · cs.HC · cs.RO

版本更新 · v2arXiv:2610.35933

Deterministic Causal Quantization with Bounded Error: Adaptive Reconstruction and Entropy bounds

Rama Cont, Massoud Heidari

原文摘要

We study compression of real-valued stationary sources under a deterministic error bound: every sample must be reconstructed within a prescribed tolerance $\varepsilon$ on every input sequence. Reconstructions are causal, depending only on present and past inputs. Rates are measured by the entropy of the quantizer output, which block entropy coding can approach, and compared with the minimum rate of unrestricted block quantization under the same error constraint. A uniform scalar quantizer with cells of width $2\varepsilon$ is within one bit per sample of this minimum whenever finite-entropy scalar quantization is possible, without density, independence, or mixing assumptions. Under entropy and Sobolev regularity conditions, the gap is at most $2\mathcal{J}\varepsilon^2/(π^2\ln 2)$, where $\mathcal{J}$ is the asymptotic Fisher information per sample. Thus the cost of causal reconstruction vanishes at a quadratic rate as the tolerance decreases. A uniform grid, however, cannot adapt to the source. We construct an adaptive quantizer whose encoder and decoder maintain a common finite set of reconstruction values, updated from the transmitted symbols alone; suitable initialization reproduces any uniform grid. When its initial values are at least $\varepsilon$ apart, its rate is within two bits per sample of the minimum for every such source. For sources uniform within separated intervals of width $2\varepsilon$, storing the interval centers attains the minimum and its Shannon lower bound. Such configurations can outperform every uniform grid of the same cell width: over sources with well-separated intervals, the best uniform grid's worst-case excess rate is $1/(2\ln 2)\approx 0.72$ bit per sample.

cs.IT · math.PR

版本更新 · v2arXiv:2603.17022

Contingency-Aware Planning via Certified Neural Hamilton-Jacobi Reachability

Kasidit Muenprasitivej, Derya Aksaray

原文摘要

Hamilton-Jacobi (HJ) reachability provides formal safety guarantees for dynamical systems, but repeatedly solving the HJ partial differential equation in changing environments limits its use in real-time planning. This paper introduces learning-based reachability and an online safety certification that together enable real-time contingency-aware planning in unknown environments with probabilistic safety guarantees. We use Fourier Neural Operator (FNO) to approximate the HJ solution operator under varying obstacle configurations. To certify the learned reach-avoid set, we recast robust scenario optimization as multiple hypothesis testings over its sublevel sets, which yields a simultaneous guarantee across candidate levels and recovers the largest certifiable volume at the prescribed safety level. Within this certified set, the induced recovery policy is guaranteed to return the system to a safe region in finite time under any admissible disturbance. Numerical results show that the learning and certification pipeline updates reach-avoid sets at real-time rates and is demonstrated on youBot simulation.

cs.RO · eess.SY