跳到正文
研踪PaperTrail

YOUR RESEARCH, IN SIGHT

让下一篇重要论文，不再错过。

覆盖计算机全学科，以 AI 为快捷入口。关注方向、作者和关键词，把发现变成持续的研究积累。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+1072新增论文
+845其中 AI 相关
+94已有论文修订
+100新增 AI 导读

arXiv 最近完成扫描：2026-11-09 16:31 · 全站最近入库：2026-11-09 13:22各信源进度 →

统计口径与数据新鲜度

统计 2026-11-08 17:38 — 2026-11-09 17:38（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：2026-11-09 16:31最近发现新增版本：2026-11-09 13:22库内论文最新更新时间：2026-11-07 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。

18,172 已收录论文15,506 AI 相关4,527 有修订112 HF 社区发现

浏览论文库

研究日报 / 周报 →按信源浏览 →

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 3,653 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-09 16:29（北京时间）
新论文 · v1arXiv:2610.34585

Large Language Models: A Plain-Language Overview for Engineers

Bahaeddin Türkoğlu

原文摘要

Large language models (LLMs) are neural networks trained on huge text collections to generate text by predicting what comes next. Engineers using ChatGPT and other generative artificial intelligence (AI) tools need a clear picture of what training changes and prompts add. The article starts with the Transformer, which connects words in context, and pre-training, where models learn to predict text. It then explains how fine-tuning adapts models to requests, with reinforcement learning from human feedback using human answer rankings. Next, retrieval-augmented generation (RAG) supplies documents, while chain-of-thought prompting encourages intermediate steps and agents use tools. FlashAttention lowers training and running costs by reducing data movement during attention; quantized low-rank adaptation (QLoRA) saves fine-tuning memory. Readers get a life cycle figure and tables comparing methods and their limits. A short checklist for judging LLM evaluation claims flags test material in training (data contamination) and model-based grading (LLM-as-a-judge). It also covers very long inputs (long context) and hallucination, meaning false or unsupported claims. Check factual accuracy separately from whether an answer sounds helpful.

cs.CL · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34582

HobbyLM: A Sparse Mixture-of-Experts Model Family on a Hobby Budget

Harish Santhanalakshmi Ganesan, Hariharan M. S., Prabhurajhan R

原文摘要

We describe HobbyLM, a family of small sparse Mixture-of-Experts (MoE) models trained from scratch with a few hundred to a few thousand dollars of rented GPU time. The core is a decoder-only transformer with fine-grained experts, sigmoid routing with auxiliary-loss-free load balancing, grouped-query attention with per-head QK-normalization, and aMuon/AdamW optimizer split. Each of these choices was selected with a single-variable ablation suite at 130M parameters. On this architecture we train a 500M-parameter model (169M active) on 40B FineWeb tokens and post-train it into a chat model, a masked-diffusion language model, a function-calling and computer-use agent, and a single "omni" checkpoint that reads text, images, video, sound and speech through frozen encoders and small projectors. We run these models on a laptop CPU with a from-scratch Rust inference engine. Two of the authors then adopted the architecture without architectural changes to build HobbyLM-1B (1.04B total, 305M active parameters), trained on a 100B-token curated mixture with a high-quality anneal. HobbyLM-1B reaches a 47.5 average on seven zero-shot benchmarks, and its builders also release an instruction-tuned model and an on-device macOS computer-use adapter. Finally, we describe HobbyLM-Image, a 333M-parameter flow-matching diffusion transformer that generates 1024x1024 images in a 32x-compressed latent space, together with a dependency-light CPU implementation of the full text-to-image pipeline. We report what worked, what did not, and where the small-scale ceilings lie. We also state for each component which author designed or built it.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v11arXiv:2201.04266

Safe Equilibrium

Sam Ganzfried

原文摘要

The standard game-theoretic solution concept, Nash equilibrium, assumes that all players behave rationally. If we follow a Nash equilibrium and opponents are irrational (or follow strategies from a different Nash equilibrium), then we may obtain an extremely low payoff. On the other hand, a maximin strategy assumes that all opposing agents are playing to minimize our payoff (even if it is not in their best interest), and ensures the maximal possible worst-case payoff, but results in exceedingly conservative play. We propose a new solution concept called safe equilibrium that models opponents as behaving rationally with a specified probability and behaving potentially arbitrarily with the remaining probability. We prove that a safe equilibrium exists in all strategic-form games (for all possible values of the rationality parameters), and prove that its computation is PPAD-hard. We present exact algorithms for computing a safe equilibrium in both 2 and $n$-player games, as well as scalable approximation algorithms.

cs.GT · cs.AI · cs.CR · cs.MA · econ.TH

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34577

Rethinking Verification in Agent Harnesses

Zhengxin Zhang, Erkang Zhu, Yin Lin, Claire Cardie, Bolin Ding, Jingren Zhou

原文摘要

Agent harnesses enable large language models to act through tools, observe the results, and iteratively work toward a task goal. Verification checks and revises agents' work to improve output quality. Yet there is limited guidance on how harnesses should organize this effort. We study verification as an axis of test-time scaling by examining when to intervene, who checks the work, and how checking leads to revision. Our framework spans four layers of the agent lifecycle: initial instructions, mid-turn reminders, completion gates, and independent review. We compare 17 designs across seven frontier coding models on Terminal-Bench (TB) and SWE-bench Verified (SWE), and introduce VisualBench for expert evaluation of visual and interactive artifacts. Initial instructions and execution-time reminders yield modest, model-dependent average gains on the coding benchmarks, without substantially increasing average output token use. Completion gates can yield larger gains but may substantially increase cost. Independent review by a critic offers a more favorable accuracy-cost tradeoff, improving average accuracy by 5.9-9.5 percentage points on TB and 3.9-6.6 on SWE at 1.13-1.25 and 1.00-1.04 times baseline cost, respectively. On VisualBench, independent review achieves net preferences of 13.2-23.8 points over the baseline across its four designs, exceeding the evaluated L0-L2 designs. Additional review rounds or independent critics further improve average TB accuracy by up to 3.5 percentage points. Together, these results provide empirical guidance for verification design in agent harnesses and support independent review as a practical axis of test-time scaling.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2609.05834

Learning Counterfactual World Models for Embodied Reasoning under Partial Observability

Todd Y. Zhou, Daniel Zhang

原文摘要

World models promise a general route to embodied intelligence: learn predictive dynamics once, then reason, plan, and act with them. Increasingly, the representations beneath such models are pretrained on large-scale video, interaction, and multimodal corpora, which raises a question prediction quality alone cannot answer: when is a learned representation actually actionable? We identify a failure mode we call counterfactual collapse: a model predicts visually plausible futures while failing to distinguish interventions with different behavioral consequences. This arises whenever a representation is optimized for perceptual similarity rather than intervention structure, which is precisely the objective under which most large-scale pretrained encoders are learned. We introduce Counterfactual Latent World Models (CLWM), which combine a recurrent belief-state encoder, action-conditioned latent dynamics, and a contrastive counterfactual objective that separates futures induced by distinct interventions even when their observations look alike. Across occluded manipulation, aliased navigation, and long-horizon manipulation, CLWM improves planning success over the strongest baseline (65.1% $\to$ 74.6% on Occluded Push and 67.3% $\to$ 78.9% on Aliased Maze) and reduces exploitative planning failures (18.4% $\to$ 9.7% on Deferred Kitchen), with ablations attributing the gains to hard counterfactual negatives, especially perceptual-alias negatives. Finally, our counterfactual separability metric, which tracks planning success across the five baseline model classes ($r \ge 0.94$), is representation-agnostic: given intervention-outcome labels, it can audit any encoder, pretrained or trained from scratch, before a planner trusts it. We do not yet measure it on large-scale pretrained encoders. Here we establish the metric and its relationship to planning success for world models trained from scratch.

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34569

Agentic AI Orchestration Framework for Spatial Intelligence: Supply Chain Deployment

Constantinos Papantoniou, Brian Hilton

原文摘要

Industrial spatial decisions are assembled across systems that share neither a reference frame nor an owner, and analysts spend substantial effort reconciling those systems rather than reasoning about space. This paper presents ANASSA, a reference architecture integrating structured reasoning, multi-agent workflow orchestration, authoritative spatial grounding, and feedback-driven execution, and reports its instantiation in an operating supply-chain deployment. The architecture places typed data contracts at agent boundaries, validates spatial claims against a certified authoritative foundation, and enforces two blocking human-in-the-loop approvals inside a six-step cognitive loop. The deployment covers multi-hazard risk assessment on an ocean-and-inland freight lane, including cold-chain excursion detection over container telemetry. Across a three-month proof of concept (POC) with three implementation iterations, three deployment outcomes are reported: 25% improvement in data quality, 50% reduction in spatial workflow runtime, and 60% lower decision latency relative to the preceding manual workflow. Absolute deployment-scale quantities and baseline counts are confidential; reported percentages are observed relative changes. These are uncontrolled deployment observations, not controlled experiments. We therefore specify instrumentation and ablation protocols for mechanism-level evaluation.

cs.AI · cs.MA · cs.SE

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34563

Recursive CLIP: Vision-Language Distillation with Looped Transformer Blocks

Szymon Mazurek, Alberto Presta, Michał Byra

原文摘要

Contrastive vision-language models such as CLIP typically rely on independently parameterized transformer blocks across depth. We investigate the extent to which these blocks can instead be reused recurrently while preserving the transferable capabilities of CLIP. We study recursive transformer backbones in which a small set of $N$ transformer blocks is applied $M$ times per forward pass. By holding $N \cdot M$ fixed to the depth of a standard ViT-B, our models preserve computational depth while progressively increasing parameter sharing. We consider two recursive CLIP model families: rCLIP, where only the vision encoder is recursive, and r$^2$CLIP, where recurrence is also introduced in the text encoder. We study these models under both direct contrastive training and through knowledge distillation from a strong ViT-B/16 teacher pretrained on DFN-2B, using the substantially smaller CC3M and CC12M datasets. We evaluate the resulting models on zero-shot classification and cross-modal retrieval, and further study their adaptability through few-shot learning, visual grounding, and concept unlearning. Our results show that CLIP can tolerate substantial recurrent block reuse, although its effect depends on the encoder and downstream task. Recurrence also provides parameter efficiency by construction, since fewer distinct transformer blocks need to be stored and adapted. Our results establish recurrent block reuse as a practical design space for vision-language models and highlight its implications for knowledge transfer, downstream adaptation, and parameter efficiency.

cs.CV · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34561

GRASP: Group-Adaptive Hybrid RL-SFT Post-Training for Medical VQA without Reasoning Annotations

Wanqi Dong, Wanyue Dong, Jingze Ge, Mehul Motani

原文摘要

Adapting general-purpose vision-language models (VLMs) to medical visual question answering (VQA) is valuable for helping clinicians read images and make diagnoses, but annotated data are scarce and provide only short answers, without reasoning annotations. Supervised fine-tuning (SFT) fits these labels but overfits to them and amplifies their biases, eroding out-of-domain (OOD) generalization. Reinforcement learning (RL) better preserves generalization but never uses the gold answer as a training target: GRPO-style algorithms learn from the contrast between correct and incorrect answers within a group of sampled responses, so all-correct or all-wrong groups give no correctness signal. Existing SFT-RL hybrids either apply SFT and RL in separate stages or switch between them per question, or need reasoning traces from experts or stronger models, which medical VQA data lack. We propose GRASP, which combines RL and SFT in every update using only short answers: RL lets the model explore its own reasoning, SFT teaches the gold answer after that reasoning to supply the direct supervision RL lacks, and their balance adapts to how well the model answers each question. Tuning only about 1% of the parameters with LoRA, GRASP outperforms all compared training objectives both in domain and on OOD open-ended questions across Qwen3-VL-8B, InternVL3-8B and Gemma 3-4B. On an OOD benchmark spanning eight imaging modalities, GRASP-trained InternVL3-8B and Qwen3-VL-8B beat all 10 compared general and medical VLMs in open-ended accuracy, including models trained on up to 2,600 times more data or with over 250 times more trainable parameters. We release MedTerse, a dataset of 12.9K short-answer questions, and our code at https://github.com/sea-comet/GRASP-MedVQA.

cs.CV · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2608.03112

Content-Aware Adaptive Token Pruning for Video-Language Models

Paribesh Regmi, Qingshuang Chen, Chi Zhang, Heba Aly, Yelin Kim, Hongda Mao

原文摘要

Vision-language models excel at image and video understanding but suffer from high inference latency due to the need to process thousands of tokens per image, limiting their deployment on resource-constrained edge devices and in real-time surveillance applications. This challenge is further amplified in video processing, where multiple frames must be analyzed simultaneously. Existing token reduction techniques generally rely on a fixed, uniform pruning ratio applied across all inputs, which is suboptimal because the degree of redundancy can vary significantly between different videos, necessitating content-dependent pruning levels to preserve critical information. To address these limitations, we propose an adaptive token pruning strategy specifically designed for video processing. Crucially, the pruning ratio is determined adaptively based on the content of each video. This is achieved by analyzing the correlation structure of token embeddings to quantify redundancy, enabling dynamic token allocation that retains more tokens for dynamic scenes while aggressively pruning repetitive or static content. Importantly, our method is entirely post-hoc and requires no additional training or fine-tuning, while achieving strong empirical gains; notably, it improves accuracy by +7\% on a video captioning benchmark at 10\% token retention, while reducing computation TFLOPs by 95\%.

cs.CV · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v11arXiv:2210.16506

Observable Perfect Equilibrium

Sam Ganzfried

原文摘要

While Nash equilibrium has emerged as the central game-theoretic solution concept, many important games contain several Nash equilibria and we must determine how to select between them in order to create real strategic agents. Several Nash equilibrium refinement concepts have been proposed and studied for sequential imperfect-information games, the most prominent being trembling-hand perfect equilibrium, quasi-perfect equilibrium, proper equilibrium, and recently one-sided quasi-perfect equilibrium. These concepts are robust to certain arbitrarily small mistakes, and are guaranteed to always exist; however, we argue that none of these is the correct concept for developing strong agents in sequential games of imperfect information. We define a new equilibrium refinement concept for extensive-form games called observable perfect equilibrium in which the solution is robust over trembles in publicly-observable action probabilities (not necessarily over all action probabilities that may not be observable by opposing players). Observable perfect equilibrium correctly captures the assumption that the opponent is playing as rationally as possible given mistakes that have been observed (while previous solution concepts do not). We prove that observable perfect equilibrium is always guaranteed to exist, and demonstrate that it selects a unique equilibrium in a simplified no-limit poker game where the prior extensive-form refinements retain a continuum of equilibria. We expect observable perfect equilibrium to be a useful equilibrium refinement concept for modeling many important imperfect-information games of interest in artificial intelligence.

cs.GT · cs.AI · cs.MA · econ.TH

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.07506

Jarvis: A Proactive Speech Agent for Multi-Party Conversations

Seunghyun Oh, Hirotaka Hiraki, Shuyue Stella Li, Yulia Tsvetkov, Shyamnath Gollakota

原文摘要

Speech agents are reactive and dyadic: they speak when spoken to, and to one person at a time. We ask what it takes for a speech agent to instead take part in a conversation among several people and speak up only when it can help. We introduce Jarvis, a real-time proactive speech agent that audibly participates in multi-party human conversations. Grounded in a document shared beforehand, Jarvis follows the discussion and intervenes when the group misses or misstates a fact and does not correct itself within a few turns. We make three contributions: a problem setting based on epistemic breakdowns that makes proactive intervention measurable, realized as CHI-180-proactive, a synthetic multi-party dataset seeded with known gaps, errors, and self-corrections; a proactive backbone that harnesses a small, open-weight model with deterministic checks and grounds every claim in a source sentence; and interaction techniques for taking the floor in live speech and showing the cited evidence on screen. On CHI-180-proactive, Jarvis is correct on most events it addresses and stays silent 97% of the time when the group resolves an issue itself. A live study with 23 participants confirms these trends with real-time interventions.

cs.HC · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34553

How Much Information Can We Fit inside a Transformer?

Amirhossein Basareh, Rishabh Sanjay, Daniel de Mello, Abulhair Saparov

原文摘要

How efficiently can transformers store factual knowledge? The answer to this question has implications for both model design and understanding the limits of transformer-based systems. Prior studies observe a linear relationship between the number of facts memorized and the number of model parameters required, substantially worse than what theoretical analyses predict is achievable. We re-examine empirical scaling: through rigorous and systematic experiments, including extensive hyperparameter search via Bayesian optimization, we determine the smallest model that can perfectly memorize a given set of facts, and measure how this minimum size scales with the number of facts. On synthetic datasets of random, independent (subject, relation, object) triples---e.g., (Paris, capital_of, France) ---drawn at random with no exploitable structure, we find that memorization capacity, the size of the largest dataset that a model memorizes completely, grows exponentially with the transformer dimension. Equivalently, the minimum transformer dimension $d_{\min}$ required to memorize $N$ facts grows logarithmically in $N$. Across 11 dataset sizes spanning $N = 45$ to $49{,}920$, a lack-of-fit F-test does not reject the logarithmic fit (with an offset), but it does reject all other tested alternatives including power-law. Along the depth axis, however, at fixed width, the maximum memorizable dataset size grows sublinearly with the number of layers over the measured range. We further extend our exploration to structured data containing symmetric relations or two-hop composition. In this setting, we find that transformers exploit these regularities to further improve capacity. All code to reproduce our results is available at https://github.com/Magronox/transformer_information_capacity.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2609.19705

Trading Agents or Market Crashers? Dissecting Robustness and Security Failures in Academic Financial LLM Trading Schemes

Mengxiao Wang, Nitesh Saxena

原文摘要

Large language model (LLM) agents already trade real capital, and builders copy a few published designs that share models and news feeds, so one shock reaches many agents at once. We present FARSIGHT (Financial Agent Robustness and Security Investigation and Global Holistic Taxonomy), a systematization that rates trading agents on robustness to market stress and on security against source manipulation, agent compromise, and agents acting as attackers, and checks the ratings against released code. Across 15 academic schemes, no paper describes a halt that skips LLM inference or combines source provenance with cross-source corroboration, and the 7 code releases enforce no stop-loss or position limit outside the LLM. Three LLMs sell whole positions on a fake headline and on real hoaxes, including the fake press release behind the 2000 Emulex crash. In a simulated market driven by these decisions, one fake headline crashes the market in all runs with 400 agents, and 400 agents that sell when prices fall nearly double a 2010-sized crash. The pause that U.S. exchanges apply to large stocks starts in no headline run, whereas automatic halts and hard loss limits outside the model remove both effects. Restated as a domain-independent threat model, FARSIGHT maps the same gaps onto agents in healthcare, law, software, and crypto, where compromised agents spread harm as worms and as cascades. We derive 14 open problems and urge deployers to corroborate sources, keep interrupts that bypass the model, and enforce loss limits outside it.

cs.CR · cs.AI · cs.MA

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.30751

How Firm Are AI Models' Ethical Boundaries? A Pay-Stub Editing Test of 15 Models

Simiao Ren, Xingyu Shen, Tommy Duong, Tsang Ng, Ankit Raj

原文摘要

Can a small change in wording move an AI model from refusing a sensitive request to acting on it? We test 15 models on the same request to increase the income shown on a pay stub. Across 445 observable trials, models issue editing calls in 14.1% of cases when a loan purpose is stated, 33.1% when the purpose is omitted, and 66.2% when the change is described as fixing a printing error. Astra refuses all ten loan-purpose requests but issues calls in all ten purpose-omitted trials. Fable, Opus, and Muse Spark issue none across all three wordings; Sonnet issues some under the latter two. Response excerpts illustrate different ways of refusing, redirecting, or justifying an edit. The file is a marked test document and all tools are simulated. These results probe how boundaries behave in one setting; they do not prove an overall ethics ranking or that a harmful real-world edit succeeded.

cs.AI · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34542

LLM WikiLeaks: Uncovering URL-Based Capability Extension and Security Circumvention in Agentic Message Boards

Orel Navi, Elad Feldman, Ben Nassi

原文摘要

Autonomous agents have been observed acting beyond their assigned tasks, accessing unauthorized systems, and establishing communication channels to exchange information, yet the Web and network-level techniques they use to extend their capabilities and circumvent security restrictions remain poorly understood. We study these techniques through the ProWiki incident, in which agents restricted to outbound HTTP GET requests repurposed a public German-language wiki as a message board. We analyze 14,591 agent edits containing 79,541 URL postings and present corroborating evidence that the published URLs reflect the agents' operational activity. We find that 26,717 URL occurrences (33.6%) embed target URLs within one to five layers of third-party Web services. By recursively reconstructing these nested URLs, we identify 67 distinct intermediate services and classify their functionality. Most intermediate layers (73.8%) extend agents' capabilities by delegating content extraction and transformation to external Web services: Markdown and OCR services convert web content into machine-readable text, JSON filtering services extract selected fields, and archival services expose historical versions of pages. The remaining 26.2% facilitate circumvention of access restrictions and evasion of detection, including CORS and fetch proxies that retrieve content on agents' behalf, and link shorteners and redirectors that hide destination URLs from mechanisms inspecting literal URLs or outer hostnames. Agents also reach the same targets through multiple URL paths, consistent with attempts to evade rate limits or obscure request origins. Our results expose a gap between the capabilities granted to agents and those they can obtain by composing publicly available Web services: restricting an agent's local toolset or direct Internet access may be insufficient while third-party services remain reachable.

cs.CR · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.30758

Can Loan Processors and Underwriters Spot AI-Edited Statements? An Empirical Study

Simiao Ren, Xingyu Shen, Tommy Duong, Ankit Raj

原文摘要

AI-assisted editing can produce loan documents that preserve familiar issuer layouts while changing the applicant's identity. We investigated whether experienced underwriters and loan processors identify such forgeries in first-pass, document-only review. Three practitioners recruited through Upwork each received one applicant package drawn from 134 forged PDFs derived from 47 genuine financial documents supplied with full consent. The edits preserved amounts, dates and balances; identifier inconsistencies introduced by the pipeline were knowingly retained before distribution. Reviewers were not told that files had been altered and returned 132 file-level judgements. Checking their explanations against the originals yielded 14 edit-supported concerns (10.6%): 12 Suspicious judgements and two Insufficient information judgements. All supported concerns involved identifier disagreements within or across documents, and no reported visual cue identified an actual edit. Thus, approximately one in ten judgements contained evidence of the forgery, despite known inconsistencies remaining in the files. These observations motivate evaluating consistency checks and independent source verification alongside human review. With three reviewer-package pairs and no genuine-file controls, the findings describe this exploratory setting rather than population-wide detection accuracy.

cs.AI · cs.CR · cs.HC · cs.SE

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34535

SentryBench: A Cost-Sensitive Benchmark for Adversarial Robustness of Tool-Using LLM Agents in Public-Safety Deployments

Md Abu Yusuf, Oyshi Tabassum Aditi, Md Faysal Hosain Mazed

原文摘要

Large language model (LLM) agents are increasingly wired to consequential tools that send public alerts, dispatch responders, and query citizen records, where one erroneous tool call carries real-world cost. Existing agent-safety benchmarks measure generic harm or prompt-injection success and implicitly treat refusing as the safe default. In public-safety deployments that assumption breaks: an agent too cautious to dispatch help fails as surely as one that triggers a false alert. We present SentryBench, which scores agents on a two-sided, cost-sensitive frontier jointly measuring attack success, over-refusal (missing a real emergency), and over-reaction (unwarranted action) under adversarial tool use, with scenarios grounded in real emergency-call transcripts, disaster messages, and real injection/jailbreak payloads. We formalize deployment harm as an asymmetric cost functional, prove naive caution is harm-dominated whenever it refuses more genuine emergencies than the attack harm it averts, and prove a targeted trust gate adds zero over-refusal on clean input. Across 840 scored episodes over three frontier models and three defenses, refuse-more caution drives over-refusal to 100 percent on two of three models for at most a 3.7-point change in attack success and yields the worst harm everywhere, whereas our trust-gating wrapper dominates the frontier (lowest harm on every model, over-refusal at or below baseline, higher containment). Provider content filters also silently perform part of the defending and must be reported separately.

cs.CR · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34529

Acute Stroke Classification Using Asymmetric Data Standardization and Orthogonal-Focal Loss Optimized CNN-ViT Cross-Attention Fusion: A Multicenter External Validation Study

Tansel Uyar

原文摘要

Background/Objectives: Rapid differentiation of ischemic and hemorrhagic stroke subtypes is vital for acute management. Although current deep learning models achieve high single-dataset accuracy, they suffer from "shortcut learning" and domain shift in external validation. This study proposes a hierarchical computer-aided diagnostic (CAD) system mimicking emergency triage. Methods: We introduce a 3-channel Radiology-Focused Preprocessing pipeline that feeds radiological Hounsfield Unit (HU) windows into Convolutional Neural Network (CNN) and Vision Transformer (ViT) backbones via asymmetric formats (.mat and .png). The architecture utilizes a novel joint cross-attention fusion network optimized by an Integrated Orthogonal-Focal Loss. Results: Our system achieved a 95.48% F1-score on the internal stroke dataset. However, multi-center external validation (AISD, BHSD, PhysioNet) using a Mask Anchor filter revealed a drop to the 58-62% range, exposing the "internal validation illusion" caused by hardware noise. Critically, models exhibited high precision (77%) but low sensitivity (57%) for ischemic stroke, proving prone to hemorrhage overdiagnosis. Multi-seed variance analysis demonstrated that the DenseNet201+ViT backbone yielded the lowest deviation (0.59) and highest stability, making it the most reliable architecture for real-world clinical integration. Conclusions: To overcome diagnostic challenges, future research must focus on multimodal architectures that combine 3D volumetric extraction with clinical data.

eess.IV · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34523

How Humans Learn Unfamiliar Worlds: A Think-Aloud Dataset on ARC-AGI-3

Shuzhi Liu, Archana Warrier, Dat Nguyen, Chengqi Zheng, Kevin Ellis, Yewen Pu

原文摘要

Learning through exploration and using newly acquired knowledge to achieve goals are central capabilities of general intelligence. Understanding these capabilities requires tracing how learners build, use, and revise world models during interaction. We construct the first dataset that records how humans discover and use the mechanisms of novel environments, using ARC-AGI-3 as a testbed: 144 trajectories from 64 participants across 25 public games, each pairing time-synchronized actions with think-aloud audio, plus 250 trajectories from large language model (LLM) agents with and without harnesses on the same games. We annotate mechanism triggers and recognition, action purposes, and errors and recovery, linking changes in understanding to the observations and actions that support them. We analyze characteristics of human cognition and compare them with those of LLM agents. Both groups anticipate mechanisms and use familiar analogies. Qualitative representations are more common in successful human trajectories than in successful LLM trajectories (50.0% versus 14.3%). Humans more often explore unfamiliar situations without a concrete hypothesis, while hypothesis testing is common in both groups. We also trace how learners recover through belief revision, replanning, and execution correction. As LLMs with well-designed harnesses reach saturated performance on public ARC-AGI-3 games, our cognitive analyses motivate next-generation benchmarks that challenge familiar priors, require accumulated evidence and abstraction, and test planning through subgoals, experimentation, and recovery. Our dataset is available at [https://arc3-human-think-aloud.github.io/](https://arc3-human-think-aloud.github.io/).

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34520

Conviction Without Evidence: Verdict Stability Is Independent of Visual Grounding in Medical Vision--Language Models

Federico Felizzi, Francesco Andrea Causio, Bianca Destro Castaniti, Olivia Riccomi, Michele Ferramola, Vittorio De Vita, Alessandro Tosi, Antonio Cristiano 等 20 位作者

原文摘要

Vision--language models applied to medical question answering can reach correct diagnoses without using the diagnostic image, inferring instead from the clinical vignette. We ask whether this is visible in how \emph{firmly} a model holds its answer: does a model defend a diagnosis derived from a blank placeholder as tenaciously as one derived from a real medical image? Crossing an image-substitution protocol with two levels of adversarial pressure on 60 Italian State Examination items, we find no evidence that ungrounded verdicts are more fragile. In a paired item-level analysis, fabricated consensus (L4) overturns 94.4\% of verdicts with the real image and 98.1\% without ($Δ= +0.037$; one-sided 95\% upper bound $+0.105$); contentless doubt (L1) overturns 62.5\% and 51.8\% ($Δ= -0.107$, 95\% CI $[-0.209, -0.001]$), the point estimate running opposite to the direction a grounding detector requires. The two levels nonetheless produce opposite \emph{dynamics}: under L4 the model flips once, always to the exact option argued for (110/110 first-turn flips), and then freezes, averaging 1.9 verdict changes over ten turns; under L1 it never settles, averaging 7.4 changes. Instability under contentless doubt tracks intrinsic item ambiguity rather than evidence availability, and so does not serve as a grounding signal. Flips are almost entirely corrupting (99/100 across both levels). The model is not unaware of the missing evidence---it flags the blank image at baseline on 35 of 60 items and abstains on 4--6---yet those abstentions do not survive contact with pressure. None of the epistemic signals a user could plausibly consult distinguishes a grounded diagnosis from a confabulated one.

cs.CV · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34519

TRIAGE: Efficient Vision-Language-Action Inference Needs Both Reuse and Pruning

Sooyoung Kwon, Carlee Joe-Wong, Kyunghan Lee

原文摘要

Vision-language-action (VLA) models are increasingly used for robot manipulation because they map camera images and a language instruction directly to actions. Robots that cannot be tethered to a workstation, such as mobile manipulators and humanoids, must run them on onboard embedded boards, where the prefill latency of the language backbone grows with the hundreds of visual tokens processed at every control step. To meet real-time limits, training-free accelerators recompute only a subset of the visual tokens and skip the rest, either by pruning them or by reusing their cached key-value (KV) states. The two options save nearly the same computation, yet existing methods apply only one of them to every skipped token. Through a token-level diagnostic, we show that which option keeps the action closer to full recomputation depends on the token's drift, the change in its shallow-layer value since its cache was computed. On both backbones tested, reusing tokens below a drift threshold was no worse than pruning them, whereas beyond it, reuse yields up to 4.4x the action error of pruning; the same diagnostic calibrates this threshold per backbone. Building on this finding, we propose TRIAGE, which deems a cache valid when its drift is below this threshold and it is not too old. TRIAGE recomputes a fixed budget of tokens, prioritizing important tokens with invalid caches, reuses every other token with a valid cache, and prunes the rest; prune-only and reuse-only methods are the two extremes of this partition. Because the budget fixes the number of recomputed tokens, it directly sets the prefill latency. On LIBERO and SimplerEnv, TRIAGE stays within 3 pp of full recomputation in average success rate while cutting prefill latency on a Jetson AGX Orin by up to 48%, and with only 7% of visual tokens recomputed, it beats the strongest baseline by 24.4 pp on LIBERO-Object.

cs.RO · cs.AI · cs.CV · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34503

Multimodal Requirements Traceability Link Recovery in Industrial Settings

Haoming Huang, Shinpei Hayashi, Kazuya Kohata, Myeongki Kim, Yuki Kurihara

原文摘要

Requirements traceability links support activities such as change-impact analysis, coverage checking, and validation. Industrial requirements can include traceable images, which text-only approaches cannot process directly. For a fixed classifier configuration, binary decisions alone offer no direct way to adjust the balance between missed and incorrect links. We extend a retrieval-augmented generation (RAG) approach for inter-requirements traceability link recovery (TLR). Our extension retrieves text and image candidates separately, classifies each retrieved pair with a vision-language model, and filters the predicted links by verbalized confidence. We evaluate our approach on three non-overlapping industrial requirement sets and six public text-only datasets. On the industrial test sets, our approach achieves an F1 score of 0.310, compared with 0.114 for our replication of the baseline approach by Hey et al. Our approach recovers links in all four modality combinations while improving text-only F1 on the industrial data and maintaining similar performance on the public datasets, but recovery remains limited by failures to retrieve related text requirements for image queries. Confidence filtering produces link sets with different precision and recall, removing false positives and some correctly recovered links from the unfiltered link set. This study provides industrial evidence on the potential and limitations of a current RAG-based TLR approach for multimodal requirements and identifies cross-modal retrieval as the component to prioritize in further optimization.

cs.SE · cs.AI · cs.IR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34500

CaliMem: Search-Efficient Data Agent via Calibrated Memory

Hanze Li, Athan Z. Li, Feiyang Kang, Jiachen T. Wang, Yuhan Chen, Ruishan Liu, Ruoxi Jia

原文摘要

Data curation is a first-order design choice in machine learning, yet tailoring training data to a downstream model and task remains costly in labor, expertise, and compute. Datasets are therefore often processed through fixed pipelines and reused across applications. We show that this assumption is fragile: across 55 dataset--model tasks spanning biological domains, the same curation recipe can help one model while hurting another. Effective curation must adapt to the downstream setting, motivating automated search over data recipes. We first build an agentic system that searches for curation policies at deployment, improving data utility via \textit{test-time curation (TTC)}. Although achieving performance gains, its search is driven largely by recent feedback and remains local. Adding persistent memory helps the agent retain evidence across iterations and reveals a deeper problem: it forms claims that extend beyond the conditions in which they were observed, sometimes abandoning an entire strategy after a single unsuccessful trial. To address this problem, we introduce CaliMem, a calibrated-memory agentic loop that requires the agent to predict the outcome of each proposed policy before executing it. Comparing this prediction with the observed result gives the agent a \textit{reality check} on the claims behind its proposal. It uses the prediction error to retain supported claims, revise mistaken ones, and specify the conditions under which a conclusion holds. This makes the agent's beliefs testable, reduces overgeneralization, and supports more effective long-horizon exploration. Across biological and LLM post-training tasks, CaliMem consistently outperforms fixed curation baselines and published methods, and exceeds full-data training by up to 19.4\%. Code is open-sourced.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2606.02552

Modeling Depth Ambiguity: A Mixture-Density Representation for Flying-Point-Free Depth Estimation

Siyuan Bian, Congrong Xu, Jun Gao

原文摘要

Despite advances in depth estimation, flying points remain a persistent failure mode: near object boundaries, depth estimators often predict spurious 3D points in the empty space between foreground and background surfaces. We trace this artifact to a standard modeling choice: assigning each pixel a single depth hypothesis. At boundaries, a pixel can straddle a foreground and a background surface, so its true depth is ambiguous between the two. A model that predicts a single depth cannot keep both possibilities, so training instead pulls the prediction toward an intermediate depth that lies on neither surface. We address this with MDA, a mixture-density representation that lets the model predict multiple depth hypotheses and their associated probabilities for each pixel. Near boundaries, different hypotheses can align with different surfaces, and the decoded depth is selected from one of these hypotheses rather than placed in the empty space between them. Across different architectures, MDA substantially improves boundary reconstruction over each base model and largely removes flying-point artifacts even under severe input blur, while adding negligible runtime overhead. The same mixture-density framework naturally extends to transparent objects, where it predicts multiple depth layers at transparent pixels, and to sky regions, where a dedicated component separates the unbounded sky from finite-depth regions, producing flying-point-free skylines.

cs.CV · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34493

Certifying Distributionally Robust Decisions from Projection Observations

Pouya Ahadi, Kamran Paynabar

原文摘要

A scalar measurement can determine a robust decision without identifying the latent covariance or every decision risk. We study finite Gaussian projection experiments with a fixed mean-plus-standard-deviation target. The observation laws define a set of indistinguishable moment models. We exhibit decisions that agree throughout that set despite unidentified risk values, and give a confidence-set certificate that returns a decision or an unresolved outcome. We prove soundness, necessary abstention for indistinguishable models with incompatible optima, and a fixed-checkpoint certification guarantee. A scalar profile reduces nonlinear risk comparisons to convex inner problems. Across seven synthetic instances, jointly calibrated moment regions give sufficient budgets 21 to 52 times explicit expected-sample information lower bounds. Separately measured stopping times are lower, with zero decision errors in 1,900 simulated streams. These comparisons do not constitute matching bounds, and the guarantees require Gaussian replication.

stat.ML · cs.AI · cs.LG · math.OC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34491HF 社区推荐

What Effort Level Should I Choose? On the Economics of Reasoning Effort in LLM Agents

Kaiser Sun, Yongchan Kwon, Zain Hasan, Shang Zhu, Kaitlyn Zhou, Federico Bianchi

原文摘要

Generative models increasingly expose a reasoning effort setting that lets users scale inference-time compute, trading higher cost for potentially improved performance. Users must choose an effort level before knowing whether extra reasoning will help, and even a successful run does not reveal whether the effort purchased was necessary. Reasoning effort is therefore a credence good: buyers observe the outcome but cannot assess the marginal value of the computation they paid for. In this work, we measure the counterfactual effect of effort levels on both proprietary and open-weight models on four benchmarks, spanning single-call reasoning and long-horizon agentic tasks. Effort, it turns out, does not consistently preserve tasks that lower effort solves, can overspend, and can change trajectory behavior. First, we find that although higher effort can raise accuracy in aggregate, it frequently fails on tasks that lower effort solved. Second, on the subset of easy tasks that every effort level solves, higher effort is generally more costly, with the highest effort level costing up to 42x as much as the lowest. Lastly, effort changes not only how many tokens are used, but also how agents solve tasks: at higher effort, agents verify their own work more heavily and more often search for solutions than work through problems directly, both of which lengthen trajectories without necessarily altering the outcome. These findings show that naively choosing the highest level is not only cost-inefficient but also offers no guarantee of the best instance-level performance. A simple alternative, starting at the lowest effort and escalating only after a verified failure, matches or improves on fixed top effort across all models and benchmarks at lower cost for most proprietary models.

cs.AI · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34490

Efficient Proof Generation for Lean Statements with Multiple Candidate Answers

Pengzhen Lin, Ke Zhang, Maziar Raissi, Zhouxing Shi

原文摘要

In formal systems such as Lean 4, efficiently finding answers to mathematical problems and generating verifiable proofs is an important task in automated mathematical reasoning. In the traditional way where an upstream solver provides multiple candidate answers, the common workflow requires searching for proofs for each candidate separately, incurring redundant computational cost. To this end, we propose a joint proof framework that integrates candidate answer selection, answer supplementation, and formal proof generation into a unified reasoning process. At its core is a trained prover that selects given or proposes new answers and generates a proof for at least one answer in a single call. Joint proof can also reuse generated proofs to check for faithfulness errors. For problems with a unique answer, if two distinct answers both obtain verifiable proofs under the same formal statement, this confirms the statement does not faithfully preserve the original problem semantics. On MATH-500, AMC 2023, and AIME 2024/2025/2026, our framework achieves the same formal proof success rate as per-candidate proof generation with about one sixth of the token cost, and identifies at least 6.1% of the formalized statements as unfaithful without requiring additional judges. Our model and code are available at https://github.com/Linppz/Gobble-Prover.

cs.CL · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34481

From Outranking Flow Scores to Linear Programs: Group Decision Making in PROMETHEE II

Alexandre Flachs, Yves De Smet

原文摘要

We study inverse and consensus analysis for PROMETHEE II ranking problems when the performance table is fixed. Existing approaches in PROMETHEE-based sensitivity and consensus analysis usually focus on deviations of criteria weights while treating preference functions as fixed. Nevertheless both sets of parameters jointly determine the outranking flow scores. We introduce a linear reparametrization of the PROMETHEE II net flow scores based on weighted preference increments over the observed differences of performances. This representation yields a linear expression of the net flows and clarifies how weights and preference functions interact. Building on this result, we derive linear optimization models for three settings: minimizing weight deviations while allowing preference functions to vary; minimizing deviations over the full set of preference parameters; and multiple decision-maker consensus analysis (with objectives based either on total deviation or on the maximal individual deviation to promote fairness). An illustrative example shows that allowing preference functions to vary can dramatically change the inverse-analysis conclusions obtained from weight-only models, and that the proposed group formulations distinguish efficient consensuses from fairer ones.

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35966

EMMA: Structured Episodic Memory for Multi-Agent Workflow Synthesis

Ruijia Niu, Kaiwen Liu, Rose Yu, Kristina Klinkner

原文摘要

Deploying LLM agents in specialized domains often require fine-grained task-specific procedures and knowledge a general model lacks, yet parameter training is costly and non-interpretable, and hand-engineered agent systems shift the cost to expert labor. Episodic memory updates at context level, but single agent adaptation struggles to alleviate cross-task interference, especially with complex tasks such as multi-application navigation or data pool analysis. Although multi-agent evolution alleviates this problem, existing methods lack clear definitions on update rules and rely on the backbone model to decide where and how much the system changes. This ill-scoped evolution makes learning follow the backbone's structural preferences rather than the observed failures: one backbone accumulates skills while another spawns agents. We present EMMA (structured Episodic Memory for Multi-Agent workflow synthesis), which treats the configuration of a multi-agent system (controller routing memory, specialists inheriting a generalist prompt, and per-step procedural skills) as episodic memory. EMMA makes the ownership and admissibility of updates explicit: a reflector assigns each diagnosed failure to the scope owning its correction and proposes one edit within it, and backbone-independent guards keep every accepted edit scoped, incremental and bounded. Across KramaBench, AppWorld and BBEH with five backbones and four baselines (ACE, GEPA, EvoMAS and ADAS), EMMA attains the best overall performance on every benchmark, up to 32.5% over the strongest baseline, maintains consistent improvements across backbones, and uses 17-38% fewer training executions than the most efficient baseline.

cs.MA · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2605.07544

From Pixels to Prompts: Vision-Language Models

Khang Nhat Hoang Vo

原文摘要

When you read a paper about a new Vision-Language Model today, it can be easy to forget how strange this idea would have sounded not so long ago. Teaching machines to see was already hard. Teaching them to read and generate language was already hard. Asking them to do both at once - and then to reason, answer questions, follow instructions, and sometimes even surprise us - still carries a quiet trace of science fiction, even as it becomes routine. This book was born from a simple feeling: it is too easy to get lost. The field moves quickly, new model names appear constantly, and the gap between "I know the buzzwords" and "I actually understand how this works" can feel uncomfortably wide. I have felt that gap many times. If you are holding this book, you probably have too. My goal is not to provide an exhaustive catalog of every dataset, benchmark, and new model variant. Instead, I want to offer something more modest - and, I hope, more durable: a clear mental map of Vision-Language Models. Enough structure that you can read new papers with confidence; enough intuition that you can design your own systems without feeling as if you are assembling LEGO bricks blindly.

cs.AI

查看论文
arXiv ↗PDF ↗