跳到正文
研踪PaperTrail

论文库

检索计算机领域论文,按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+549新增论文
+443其中 AI 相关
+62已有论文修订
+95新增 AI 导读

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

统计口径与数据新鲜度

统计 2026-11-20 17:13 — 2026-11-21 17:13(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:2026-11-21 12:19最近发现新增版本:2026-11-21 12:19库内论文最新更新时间:2026-11-20 02:59

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

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

浏览论文库

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

找到 6,976 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-21 12:17(北京时间)
新论文 · v1arXiv:2611.21331

TRIAGE-CPG: A Cross-Model Study of Budgeted Static Malware Triage with Large Language Models

Paul-Andrei Sava, Alexander Küchler, Daniel Kowatsch, Bernhard Grill, Jennifer Pullman, Yanick Fratantonio, Luca Invernizzi

原文摘要

When triaging unknown binaries, security analysts must decide which samples to clear as benign, flag as malicious, or escalate for deeper analysis. Large Language Models (LLMs) could support these decisions by interpreting decompiled code, but a binary can contain hundreds or thousands of functions, making exhaustive LLM analysis impractical. We study how current LLMs behave when static triage is formulated as budgeted evidence acquisition: which functions they inspect and how they trade off benign clearance, malware detection, and escalation. To enable a controlled comparison, we implement TRIAGE-CPG, a fixed pipeline over a code property graph (CPG) of a decompiled binary. An LLM-free heuristic selects seed functions, an agent expands the resulting evidence dossier under a fixed budget, and a reporting stage produces a benign, suspicious, or malicious verdict together with a behavioral report. We evaluate eight closed-source and open-weight models on 1,000 Windows PE binaries, using a 600-sample test split and the same zero-shot protocol for every model. We assess the ternary verdicts under both a strict binary reading and triage balanced accuracy (tBA), which credits malware that is flagged or escalated and benign samples only when cleared. After analyzing about 40 of a mean 1,000 functions per binary, the highest-scoring model reaches a strict balanced accuracy of 0.805 and a tBA of 0.855, flagging 65% of malware, escalating another 12%, and clearing 94% of benign samples. The models implement different triage policies: high-specificity models produce small, malware-rich escalation queues, whereas recall-oriented models escalate or flag much more broadly. Both closed-source and open-weight models appear across this spectrum. Our results identify no universal deployment configuration: model choice itself determines the clear/flag/escalate trade-off under limited static evidence.

cs.CR · cs.AI

版本更新 · v3arXiv:2510.10189

Formally Verified Certification of Unsolvability of Temporal Planning Problems

David Wang, Mohammad Abdulaziz

原文摘要

We present an approach to unsolvability certification of temporal planning. Our approach is based on encoding the planning problem into a network of timed automata, and then using an efficient model checker on the network followed by a certificate checker to certify the output of the model checker. Our approach prioritises trustworthiness of the certification: we formally verify our implementation of the encoding to timed automata using the theorem prover Isabelle/HOL and we use an existing certificate checker (also formally verified in Isabelle/HOL) to certify the model checking result.

cs.LO · cs.AI

新论文 · v1arXiv:2611.21705

Disambiguating Moral Circuits and Moral Features in Language Models: Moral content, normative judgment or cue words? Evidence from 14 models

Davide Liga, Liuwen Yu, Matteo Magnini

原文摘要

Interpretability methods can locate the attention heads and sparse autoencoder (SAE) features that change the answer a language model gives to a moral prompt, e.g., whether "Stealing medicine for a dying child is morally" continues with "justified" or "wrong". Such a component could respond to moral content, to the judgment of conduct by any norm, or to the word "morally" in the prompt. We built 227 minimal pairs of short phrases, validated by five annotators, whose two versions differ in a moral standard, in another kind of norm (prudence, etiquette, aesthetics or evidence) or in a descriptive quality, and showed them to 14 open-weight models (Llama, Gemma and Qwen, 1B to 14B, base and instruction-tuned), with and without the word "morally". In every model, the heads that change the moral answer belong to an evaluation circuit the models use for every kind of normative judgment we tested: these heads carry the same share of the effect for moral as for other normative pairs (pooled ratio 1.00, 97.5% confidence interval [0.92, 1.07], with the word; 0.90, [0.84, 0.96], without it), and the models need them as much for both. The models need these heads also when nothing in the prompt asks for a judgment, and on 400 scenarios of the ETHICS benchmark, where removing them takes away 37 to 39% of the difference between acceptable and unacceptable scenarios, against 1 to 3% for ten random heads. The SAE features we located respond mostly to the word "morally" rather than to what the phrase is about. The models represent the category of a pair (the standard the annotators assigned to it): it can be read from their hidden states, and, in the eight models we tested, removing the direction that carries it weakens the moral verdict by a few percent, and more than the other verdicts in seven of them. We release the pairs, the ratings, the code, the trained SAE dictionaries and the result files.

cs.CL · cs.AI · cs.LG

新论文 · v1arXiv:2611.21315

Benchleaking Is Accelerating: Benchmark contamination has worsened sharply in the past month -- from post-training down to pre-training base models

Yiwei Qin, Muhang Xie, Xuefeng Li, Mohan Jiang, Dayuan Fu, Tiantian Mi, Qipeng Guo, Pengfei Liu

原文摘要

Benchmark contamination is not new, but recent model releases show it in a sharper form. We use the term benchleaking for the case in which a benchmark's own test examples enter training data, so that a score reflects recall rather than capability. This differs from benchmaxxing, which optimizes toward a benchmark without seeing its test set. We document two escalations over the past month. First, in post-training, the traces have become explicit: recent audits show an agent stating hidden test requirements it was never given in 23% of its attempts, where its predecessor did so in none, and continuing to follow them after the task is changed to contradict them (Schaeffer and Räuker, 2026; Lo and Lo, 2026). Second, the problem has reached pre-training. Using few-shot continuation on 20 benchmarks across four public base models, we find that some base models write out other test examples---complete questions, options, and answers---that appear nowhere in their input. For the recently released ERNIE-5.5-Flash-Base, 96% of HumanEval problems leak with their reference implementations and 85% of OpenBookQA questions leak with their answers, against 9% and none for DeepSeek-V4-Flash-Base. To quantify what a reported score owes to such examples, we define an Anti-Hack score: the benchmark score recomputed with every leaked example marked wrong and the number of examples unchanged. Averaged over the suite, it costs DeepSeek-V4-Flash-Base 1.5 points and Hy3-preview-Base 1.3, against 5.0 for MiMo-V2.5-Base and 16.6 for ERNIE-5.5-Flash-Base. Open base models are the foundation of much downstream research, so benchleaking now affects more than leaderboard rankings. We call for decontamination reports and contamination-adjusted scores to become standard practice.

cs.AI

新论文 · v1arXiv:2611.21310

Can AI Mature a Model's Geometry? Training-Free Weight Baking in LLMs

Amin Golnari, Saeid Sanei

原文摘要

Can an AI make a trained language model more mature, in the sense of a better organized latent space, by looking only at the geometry of that space? An analyst model reads a fixed-size digest of topological descriptors of the layers of a frozen target and proposes a transformation, which closed-form ridge regression writes into the feed-forward weights without gradient training. Using lightweight models, 126 runs show that category purity rose on unseen words and sentences, benchmark accuracy stayed within sampling error, and the analyst matched a category-aware heuristic with gentler edits. In a five-model round robin approach, 16 pairs raised purity and four left the target unchanged. The results provide a baseline for future work on LLM-guided, training-free model editing without using raw weights.

cs.LG · cs.AI

新论文 · v1arXiv:2611.21307

Programmable Semantic Interventions: Learning Causal Carriers from Behavioral Programs

Paolo Muratore, Mackenzie Weygandt Mathis

原文摘要

Causal abstraction explains a frozen network by aligning its activations with the variables of a simpler, interpretable model, and checking that interventions on the two agree. Finding such an alignment currently requires a counterfactual dataset: for every pair of prompts, the answer the network ought to give once the variable has been changed. In practice, those labels are produced by an executable algorithm for the behavior, written before studying the model supposed to implement it. However, the criterion these methods optimize only leverages the answer an intervention should produce. Here, we develop an approach that lets the investigator define that relationship directly: the answers the variable can produce, and which of them an intervention should yield. We call this interface a semantic program and introduce Programmable Semantic Interventions (PSI): a low-rank activation subspace learned so that interchanging it changes the model's answer as the program declares, with the expected answers read from the frozen model itself. On the MIB benchmark PSI performs on par with supervised alignment at matched data and training budget, without curated pairs, and it needs no ground-truth labels; while on natural science questions with gold labels withheld, it tracks the model's own answers better than a subspace trained on gold labels. The declaration also fixes a coordinate system with known semantics, enabling donor-free steering: jointly training PSI for both localization and control realizes 94.8% of held-out requests for answers the investigator chooses, without the need for donor prompts.

cs.LG · cs.AI

新论文 · v1arXiv:2611.21306

Share your Screen and Talk to Gemini: How Young Learners Interact with Multimodal GenAI While Solving Scratch Tasks

Sven Jacobs, Lukas Berning, Natalie Kiesler

原文摘要

Children using Scratch to learn programming can struggle with text-heavy, lengthy explanations and hints during debugging. Verbal interactions based on screen sharing or camera data may be an interesting alternative for supporting learners. We therefore explore the quality, use, and perceptions of fifth- and sixth-graders' verbal interactions with a GenAI tutor. The goal is to investigate the quality and use of the multimodal GenAI tutor in the given context. First of all, a pilot study of eight tasks conducted without students helped inform the process of the tutor's classroom evaluation. The classroom study was realized after fifth- and sixth-graders had participated in six Scratch classroom lessons. Then 23 students were introduced to the tutor, worked on the eight tasks with the GenAI tutor, and completed a questionnaire. The analysis of the pilot and classroom study shows that the tutor's responses were intelligible and actionable, and no single turn revealed a complete solution. However, the tutor missed existing errors in 61.5% of 603 error-related classroom turns, compared to 15.7% in the pilot study. Children's voice prompts were brief and mostly responded to the tutor. Although a video had explained that responses could be incorrect, 39.1% of the children trusted the tutor's answers to always be correct. These findings have implications for educators and learners using multimodal GenAI tutors, as they need to be aware that an intelligible and human-like voice does not ensure correct support. For developers, our results suggest that screen sharing alone does not ensure responses that match the program state.

cs.AI · cs.CY · cs.HC

新论文 · v1arXiv:2611.21304

Fix, Don't Retake: Physics-Simulated Training for Correcting Patient Positioning Errors in Dental Panoramic Radiographs

Sema Helali, Lina Abu Nada, Faleh Tamimi, Rafat Damseh

原文摘要

Patient-positioning errors can distort panoramic radiographs and lead to repeat examinations. This study investigated artificial intelligence for correcting these errors in existing images. A physics-based simulator generated paired correctly positioned and mispositioned panoramics from cone-beam computed tomography volumes, with controlled individual and combined positioning errors. Five approaches were compared: U-Net, pix2pix, conditional diffusion, an error-conditioned model, and a mixture of error-specific specialists. Evaluation used image similarity and geometric metrics across three test sets. All approaches improved mean image similarity and reduced geometric errors, with U-Net and the two error-aware approaches outperforming pix2pix and conditional diffusion.

cs.CV · cs.AI

新论文 · v1arXiv:2611.21299

Turing Agent: Agentic Formal Verification for Mathematical Research

Yang Chen, Qianyi Wang, Ronghui Yang, Zhanhao Li, Yu Cong, Chen Zhang, Hangfan Zhang, Shao Zhang 等 16 位作者

原文摘要

Large language models (LLMs) are rapidly advancing mathematical discovery, moving beyond benchmark-level problem solving to tackle frontier research problems. This progress is ushering mathematics into an era of proof abundance, raising a fundamental challenge: how can we ensure the correctness of the ever-growing volume of AI-generated proofs and transform them into trustworthy, machine-verified mathematical knowledge? We introduce Turing Agent, an agent harness designed to orchestrate end-to-end formal verification of research-level mathematics. Its wo-scale architecture integrates theorem formalization and proof formalization agents, empowered by structured proof decomposition, parallel verification, accumulated proof experience, and selective semantic review. Leveraging our collaborations with theoretical computer science (TCS) researchers, we focus our evaluation on TCS, a particularly demanding domain characterized by intricate quantification, complex dependencies, and stringent semantic requirements. Against direct generation and general-purpose coding harnesses, including Codex CLI and Claude Code, Turing Agent achieves approximately 57-62% faithful formalization success on Formal TCS Bench and 83% overall pass@1 on a TCS-focused subset of RLMEval, using GLM-5.3 and DeepSeek-V4-Pro. Notably, Turing Agent, built on DeepSeek, successfully completes the formal verification of three previously unresolved TCS tasks from OpenAI's October 2026 collection. One particularly challenging TCS problem required about a hundred formally verified auxiliary claims. Moreover, by actively exploring and refining the formalization process, Turing Agent can uncover simpler alternative proofs, expose unnecessary dependencies in original proofs, and generate reusable mathematical knowledge as a byproduct of verification.

cs.AI

新论文 · v1arXiv:2611.21298

Agent-Native by Design: A Programming Language for Agentic Applications

Jayanaka Dantanarayana, Savini Kashmira, Xiaoyu Ma, Malitha, Malitha Prabhashana, Lingjia Tang, Jason Mars

原文摘要

Building agentic applications requires combining large language model (LLM) operations, workflow structure and application state, and model-guided routing. With SDKs and agent frameworks, developers connect these components through supporting code that prepares model inputs, propagates state, and dispatches model-selected steps. Workflow changes can require coordinated updates to this code, increasing development and testing effort. Splitting execution logic between application code and framework internals also limits compiler visibility into control flow and state dependencies, hindering automatic optimization. We introduce Agent-Native Programming, which unifies these three components with shared execution semantics. We realize this approach in Jac, a Python-superset language, through "by llm()" for typed model operations and tool use, Object-Spatial Programming for workflow structure and state, and "visit ... by llm()" for model-guided routing. The routing construct derives the model's choices from executable workflow steps and schedules the selected steps. This explicit connection between model decisions and program execution gives the compiler and runtime a common basis for dependency analysis and automatic optimization. Across five benchmarks and five baselines, Jac achieves a 1.8-8.7x reduction in AI-integration code relative to the OpenAI SDK, with competitive observed task quality and lower coding-agent development effort. Two case studies demonstrate speculative execution and service partitioning through configuration alone, using program analysis and runtime checks to preserve workflow behavior under stated conditions.

cs.PL · cs.AI · cs.SE

版本更新 · v2arXiv:2610.20824

Wolof-Adapted Industry-Ready Open-Weight Language Models

Gwénolé Lecorvé, Delphine Charlet, Pierre Adam, Floriane Behanzin, Lionel Delphin-Poulat, Johannes Heinecke, Géraldine Damnati, Pelin Kaytmaz 等 18 位作者

原文摘要

Wolof is spoken by 23M people, mostly in Senegal, but remains weakly supported by open-weight language models, while API models raise cost, governance, and sovereignty constraints for local deployments. In this paper, we present 0.6B-14B open-weight models adapted to Wolof through instruction tuning, preference alignment, and model merging. After presenting the pipeline and data, extensive experiments on 63~public and in-house tasks show good performance on Wolof understanding and generation while remaining effective on other languages and cross-lingual skills, with the 8B and 14B models being particularly competitive with substantially larger commercial systems. The models are currently being evaluated in internal customer-care pilots and are publicly available at https://huggingface.co/collections/Orange/wolof-lm .

cs.CL · cs.AI

新论文 · v1arXiv:2611.21291

Benchmarking Large Language Models for Linguistic Annotation

Ondřej Tichý, Anna Marklová, Barbora Bulantová, Magdalena Titlbachová, Jiří Milička

原文摘要

Assessing large language models for linguistic annotation requires task-specific evaluation under documented conditions. We present dhAIBench, a Python benchmarking and annotation tool with a web interface for exploring and sharing results. It records prompts, configurations and predictions, enabling comparisons of annotation quality, confidence calibration, repeatability and estimated cost. We demonstrate the framework and the associated methodology through 293 runs of 48 models on nine linguistic annotation tasks spanning morphology, syntax, semantics, discourse-pragmatics, spelling normalization and learner-language correction. Models achieve high agreement with reference annotations on several tasks, including those on non-standardized, historical varieties as well as requiring contextual interpretation. The results support evaluating models against specific linguistic tasks, rather than relying on general rankings. To this end, we make the current and future benchmarks publicly available at dhaibench.silent3.ff.cuni.cz.

cs.CL · cs.AI

新论文 · v1arXiv:2611.21290

FM-Recon: Exploring Foundation Model Features for Accelerated Cardiac Cine MRI Reconstruction

Donghang Lyu, Marius Staring, Ruochen Gao, Jochen Keupp, Hildo J. Lamb, Mariya Doneva

原文摘要

Foundation models have demonstrated strong feature extraction capabilities across various medical imaging tasks, yet their potential for MRI reconstruction remains largely unexplored, particularly within unrolled reconstruction networks and under artifact-corrupted inputs. In this work, we explore whether and how features from pretrained foundation models can benefit accelerated cardiac cine MRI reconstruction. To this end, we propose FM-Recon, a framework that refines and integrates extracted features from frozen image encoder of foundation models into an unrolled reconstruction network. Through extensive experiments on the CMRxRecon2023 dataset, we systematically evaluate foundation model features across different reconstruction settings. Our results show that foundation models can extract informative representations from artifact-corrupted inputs and improve reconstruction performance, particularly at moderate acceleration factors. These findings demonstrate the potential of foundation model features for accelerated MRI reconstruction and provide insights into their effective integration into unrolled reconstruction networks.

cs.CV · cs.AI

新论文 · v1arXiv:2611.21702

Watermark Evidence under Open-Weight LLM Modification: Exact Bounds for Low-Rank Output Watermarks

Prashant Shekhar, Dumindu Samaraweera, Caroline Howard

原文摘要

Open-weight language models can continue generating after their released weights are modified, while a watermark verifier still scores outputs against the original base and marked models. We study how such modifications change the evidence available to an OpenStamp-style verifier. At each shared prefix, we represent the modification as a residual in the next-token logits. We derive an exact finite-vocabulary lower bound on the expected log-likelihood-ratio evidence under a residual-span constraint, along with a sharp span-only deviation bound. We also characterize the first-order evidence shift through a covariance term with an explicit second-order remainder, and determine the residual span needed to reach a target evidence level. For sequence-level detection, we derive an exact conditional exponential-moment bound and combine it with a supermartingale argument and a distribution-free order-statistic tolerance bound. This yields a high-confidence upper bound on miss probability after attack selection on calibration data and evaluation on independent certification continuations. In experiments with OpenStamp on Llama-2-7B and Mistral-7B-v0.3, INT8 quantization and 30% pruning retain test true-positive rates of 85.6-94.8%, while 500-step LoRA fine-tuning lowers them to 28.8% and 11.6%. The covariance prediction tracks the measured mean LoRA evidence drift on both models, with no remainder-bound violations across 160,000 test prefixes. Additionally, we observe zero false positives among 500 base-model continuations per model. Through these theoretical and empirical results, the current research provides a better understanding of watermark robustness to model-editing/tuning attacks.

cs.CR · cs.AI · cs.CL

新论文 · v1arXiv:2611.21273

CAE-Bench: Benchmarking AI Agents for Reusable CAE Solver Development

Zhiyu Zhou, Tianyou Liu, Xinping Yang, Yiyao Liu, Jiayi Dong, Yihan Lin, Haokai Ye, Yifan Lu 等 11 位作者

原文摘要

We developed CAE-Bench, a benchmark to evaluate AI agents on Computer-Aided Engineering (CAE) solver development. A CAE solver is more than a PDE solver; it must be able to generalize across various inputs, such as different boundary conditions, geometries, and material properties. CAE-Bench evaluates the generalizability of AI-developed CAE solvers by providing the same task, such as a heat transfer problem, multiple cases, with each case specifying an input. The AI agent develops a solver for a fixed task using only a few public cases, and the resulting solver is then tested on held-out cases, i.e., cases it has never seen before. We evaluate 12 models across a range of reasoning settings on 80 computational tasks. Our experiments show that even if a solver developed by an AI agent performs well on the public cases, it may even get 0 score on held-out cases. For tasks such as fluid dynamics problems, even the solver developed by GPT-6 Astra performs poorly on held-out cases. A data format and an SDK for task registration are provided, so the researcher can use the benchmark on their own infrastructure. We hope that CAE-Bench will help the community to develop AI agents that can develop reusable CAE solvers that generalize to various inputs.

cs.AI

新论文 · v1arXiv:2611.21268

In-Context Estimation and Prediction for Parametric ODEs

Abdullah Ahmed, Yun Yang

原文摘要

Transformers have demonstrated the ability to perform in-context learning, that is, to adapt to new tasks from a few examples without updating their parameters. Existing theory has largely focused on emulating gradient-based algorithms for convex statistical problems or establishing general function approximation guarantees without exploiting the underlying statistical structure. We study in-context parameter estimation and trajectory prediction for parametric ordinary differential equations (ODEs), where estimation is generally nonconvex. Under suitable identifiability and regularity conditions, we construct a transformer that estimates unknown system parameters from noisy trajectory observations through a two-stage approximate maximum likelihood procedure: a coarse search provides an initialization within a locally strongly convex region, followed by gradient descent with local linear contraction. For prediction, we compose this estimator with a neural approximation of the ODE solution map. We further show that, through parameter sharing, a single transformer can perform both parameter estimation and trajectory prediction simultaneously. For these constructions, we establish nonasymptotic guarantees and finite-pretraining oracle inequalities that separate errors due to finite pretraining data, finite context length, and finite-sized transformer approximation. Numerical experiments on linear and nonlinear ODEs further show that pretrained transformers can benefit from longer inference contexts without updating their weights.

cs.LG · cs.AI · math.NA · math.ST · stat.ML

新论文 · v1arXiv:2611.21267

Conformal Expert Pools: Training-Free Probabilistic Time-Series Forecasting Beyond Seasonality

Valery Manokhin

原文摘要

Conformal Seasonal Pools (CSP) showed that a forecaster with no training at all, built only from past values at the same point in the season and from past forecast errors, significantly outperformed DeepNPTS, Amazon's learned non-parametric forecaster. We propose Conformal Expert Pools (CEP), a training-free successor. CEP replaces the single seasonal-naive forecast by a pool of simple forecasters, builds its predictive distribution from their past forecast errors, and adapts both the pool and the errors to the series being forecast. No model is trained and no parameter is learned across series. On the CSP benchmark of 232 datasets and 24,502 forecast windows, CEP is better than CSP on 78% of datasets by CRPS, on 84% by MSIS and on 83% by Winkler score; on a typical dataset its CRPS is 14% lower and its MSIS 23% lower (both p < 1e-22, paired Wilcoxon test), and coverage of 95% intervals rises from 90.1% to 92.2%. CEP has a lower CRPS than DeepNPTS on 88% of datasets, than NPTS on 89% and than SeasonalNPTS on 83%. The improvement persists on dataset families outside the development set and on 36 additional fev-bench files. We also test the published claims that Laplace, an online ensemble forecaster, "beats (almost) everything" and "beats" CSP. On 228 datasets at multi-step horizons the claims do not hold: CSP outperforms Laplace on 60% of datasets by CRPS and on 64% by interval score, and Laplace ranks sixth of seven methods, with the widest intervals of all (69% wider than CEP's). Its point forecast is biased low and on 54% of datasets is less accurate than seasonal naive, and from a cold start it takes about 7,700 times longer than CSP. CEP is better than Laplace on 82% of datasets by CRPS and on 80% by MSIS, with intervals about 40% narrower, and runs 271 times faster.

cs.LG · cs.AI

版本更新 · v2arXiv:2605.06211

Contrastive Identification and Generation in the Limit

Xiaoyu Li, Andi Han, Jiaojiao Jiang, Junbin Gao

原文摘要

In identification in the limit (Gold, 1967), a learner reads positive examples and must eventually name the target hypothesis; in generation in the limit (Kleinberg and Mullainathan, 2024), it must eventually produce new positive examples. We study a weaker, relational signal: each observation is an unordered pair $\{x,y\}$ with $h(x)\neq h(y)$, such as a cannot-link constraint or a dissimilar-pair label, and the learner never learns which point is positive. All our results come from one object: a pair is valid for every member of a finite family exactly when its points lie in antipodal cells of the family's membership map. (1) For every corruption budget $k$, known or unknown, we characterize exactly the identifiable classes. Incomparable hypotheses must cover the domain and share more than $2k$ points, and Angluin's tell-tale sets must survive the loss of $2k$ points. Every unbounded budget makes any two hypotheses confusable. (2) A contrastive closure dimension characterizes uniform generation. On finite classes, generation fails exactly when a family that shares a presentation has a finite common support; on countable classes this finite condition no longer suffices. (3) On clean data the four learning criteria form a strict diamond. (4) Under corruption the order changes: some classes are identifiable from contrastive pairs under every finite budget by one algorithm, but not from positive examples with one corrupted example.

cs.LG · cs.AI · cs.CL · cs.DS

版本更新 · v4arXiv:2409.16238

Linear-time Rule Mining Under Formal Guarantees

Jonathan Feldstein, Dominic Phillips, Efthymia Tsamoura

原文摘要

Probabilistic logical models are a core component of neurosymbolic AI and are important in their own right for tasks that require high explainability. Unlike neural networks, logical theories that underlie the model are often handcrafted using domain expertise, making their development costly and prone to errors. While there are algorithms that learn logical theories from data, they are generally prohibitively expensive, limiting their applicability in real-world settings. Here, we introduce precision and recall for logical rules and define their composition as rule utility -- a measure of the predictive power of logical theories that can be computed in linear time in the data size. We then present SPECTRUM, a linear-time framework for learning logical theories from relational data, that mines recurrent patterns in the data, and subsequently uses our utility measure to evaluate and rank the rules derived from these patterns. Finally, we prove theoretical guarantees on the utility of the learnt logical theory. As a result, we demonstrate across various tasks that SPECTRUM scales to databases with orders of magnitude more relations and entities, often learning more accurate logical theories on CPUs in < 1% of the runtime of state-of-the-art neural network approaches on GPUs.

cs.AI

版本更新 · v2arXiv:2609.31071

Externalized CPDAG Summaries Improve LLM Causal Deduction

Wentao Sun, João Paulo Nogueira, Dominique Verchere, Mathieu Acher, Alonso Silva

原文摘要

Observational data rarely identify a single causal structure. Such structures are typically modeled as directed acyclic graphs (DAGs) whose arrows point from direct causes to their effects, and observed correlations and conditional independencies usually leave a whole family of DAGs that fit the data equally well. A causal claim follows from the data only if it holds in every one of them. Classical causal discovery methods such as the Peter--Clark (PC) algorithm summarize this family as a completed partially directed acyclic graph (CPDAG): an edge is directed when every compatible graph orients it the same way, and undirected when the data leave its direction open. On the Corr2Cause benchmark, large language models (LLMs) that reason in free text often lose track of this family and decide from a few local edges. We propose \emph{Structured Thinking}, a two-turn method in which the model first summarizes the CPDAG in a constrained schema and then checks the claim against that summary. Without fine-tuning or an external solver, it raises Qwen3.5-27B from $77.5$ to $86.5$ F$_1$ (mean of three seeds) over a baseline that receives explicit PC-algorithm instructions, whereas a second turn of free-text analysis does worse than that baseline. The gain holds on test splits with renamed variables or paraphrased hypotheses and on other models. Given only the graph and the claim, the model's answer flips when the graph is replaced by a valid CPDAG under which the claim has the opposite truth value. Moreover, replacing the model's graphs with reference CPDAGs on a fixed subset shows that most remaining errors arise in building the graph rather than in reading it. Making causal structure explicit can help LLMs reason more reliably, provided the intermediate representation is itself trustworthy.

cs.AI

新论文 · v1arXiv:2611.21250

Stream of Belief: Eliciting Collective Intelligence through Latent Belief Dynamics

Yu Yao, Jiayi Dong, Yilun Du

原文摘要

In this work, we view multi-agent reasoning as a process of seeking a collective equilibrium of implicit belief dynamics and introduce Stream of Belief (SoB), where agents maintain continuous latent beliefs and iteratively refine them through communication and reflection. Our method seeks a collective belief state of a multi-agent system where further communication induces no meaningful information propagation. Through extensive experiments, we demonstrate the generalization of SoB across tasks and model families. More Broadly, our results suggest a new perspective on scaling multi-agent reasoning from belief dynamics.

cs.AI · cs.CL · cs.MA

新论文 · v1arXiv:2611.21240

How Developer Personality Shapes the Use of AI Coding Assistants

Annemarie Wittig, Alina Mailach, Norbert Siegmund

原文摘要

Developers differ in personality traits that shape how they approach software development tasks, yet AI coding assistants either behave uniformly or offer personalization features whose fit with individual developers must be found through trial and error. This is because it remains unclear how personality shapes developers' experiences with AI assistants, and thus which traits personalization should account for. In other words, we provide tools with no underlying theory on what works when and why. We take a first step towards such a theory with a systematic scoping review on how users' personality relates to their perception of LLM-based tools. We identified twelve studies through general search engines, major Software Engineering (SE) venues, AI tooling, and snowballing. They all reported associations between personality and perceptions, particularly perceived usefulness, yet none addressed this in a SE context. Building on these findings, we conducted an exploratory two-wave survey with 29 and 11 developers, collecting first empirical evidence of how developers' personality (measured using the Big Five Model) relates to the perception of AI coding assistants. Higher Extraversion was associated with more positive private-use experiences in both waves, while Agreeableness was associated with negative professional experiences in the first wave and positive ones in the second. These results provide first indications that the personality effects found in the literature extend to developers. Although further studies with bigger samples are needed to build a theory, our work takes the first step towards an empirical grounded theory of developer personality and AI personalization traits.

cs.AI · cs.SE

新论文 · v1arXiv:2611.21235

Language-model agents' beliefs can be manipulated via truthful evidence in open channel

Jinwei Hu, Qisong He, Xinmiao Huang, Youcheng Sun, Yi Dong, Xiaowei Huang

原文摘要

Language-model agents' beliefs can be manipulated through open channels such as social media, even when the information they receive is truthful. When agents fill gaps between fragments with self-induced relationships, they may accept those relationships as established facts and pass the resulting rumours to others. Drawing on cognitive theory of narrative understanding and film theory of montage, we introduce Generative Montage to operationalize this failure, termed narrative overfitting, by coordinating agents to arrange truthful fragments into misleading narratives. Across 1,090 rumour targets from six Twitter events, agents often accepted rumours overconfidently, with overall acceptance exceeding 65 percent in twelve of fourteen models. Acceptance remained above 50 percent after majority voting or review of the same evidence, showing how errors can reach downstream decisions. To explain this susceptibility, fragment interventions identify posts that influence belief within complete narratives. Synthetic controls then show that similar response patterns arise from valid evidence, separating belief influence from evidential support. This distinction motivates Grounded Monitor, a proof-of-concept neuro-symbolic runtime monitor that checks direct evidential support to guide agents' judgements, improving decision accuracy by 5.6 to 29.0 percentage points across original and anonymized settings. These findings expose a cognitive route to rumour amplification and support monitoring inferred relationships before agent-generated conclusions guide public decisions.

cs.AI · cs.CL · cs.MA · cs.SI

新论文 · v1arXiv:2611.21232

A Generative AI-Driven Architecture for Whole-Home Intelligent Services: Edge Models, Memory Management, and Household Applications

Mo Zhou

原文摘要

Generative artificial intelligence is entering the home. Cloud-based assistants face governance limitations in data transmission, continuous availability, and the reliability of physical actions, which are especially salient in households where devices remain online, data accumulate over time, and commands trigger real-world actions. Using the home local server as a unifying organizing axis, this article develops a design-oriented review framework. Sources from arXiv, CNKI, ACL Anthology, and official documentation were searched through November 2026 and coded into five evidence categories: reviews, system prototypes, benchmarks, HCI and eldercare studies, and official documentation. System designs were mapped onto shared architectural modules, identifying nine recurrent responsibilities of the home local server. Generative models support perceptual interpretation, intent understanding, and candidate planning, while deterministic validation, knowledge-graph constraints, and authorization leases confine candidate actions to an inspectable space. Home control, emotional companionship, and intelligent guarding can be compared within a common framework: they share state and memory but differ in latency, false positive and negative rates, proactivity, and permissions. Existing benchmarks rely on virtual environments and cannot substitute for real-deployment evaluation. Local deployment is not inherently safer, and technical continuity alone does not establish the social or medical effects of companionship and guarding. The article proposes a five-dimensional evaluation framework covering architectural capability, system resources, privacy, safety and reliability, and human experience, together with six design principles and a future research agenda for local-first intelligent home systems.

cs.CL · cs.AI

版本更新 · v4arXiv:2505.07675

Simple yet Effective Semi-supervised Knowledge Distillation from Vision-Language Models via Dual-Head Optimization

Seongjae Kang, Dong Bok Lee, Hyungjoon Jang, Sung Ju Hwang

原文摘要

Semi-supervised learning (SSL) has emerged as a practical solution for addressing data scarcity challenges by leveraging unlabeled data. Recently, vision-language models (VLMs), pre-trained on massive image-text pairs, have demonstrated remarkable zero-/few-shot performance that often surpasses SSL approaches due to their exceptional generalization capabilities. This gap motivates us to question: how can we effectively harness the powerful generalization capabilities of VLMs into task-specific models? Knowledge distillation (KD) offers a natural framework for transferring VLM capabilities, but we identify that it suffers from gradient conflicts between supervised and distillation losses. To address this challenge, we propose Dual-Head Optimization (DHO), which introduces dual prediction heads for each distinct signal. We observe that DHO mitigates gradient conflicts, enabling improved feature learning compared to single-head KD baselines, with practical benefits of minimal computational overhead and test-time hyperparameter tuning without retraining. Extensive experiments across 15 datasets show that DHO consistently outperforms KD baselines, often outperforming teacher models with smaller student models. DHO also achieves new state-of-the-art performance on both in-distribution ImageNet semi-supervised learning and out-of-distribution generalization across ImageNet variants. Code and model checkpoints are publicly available at https://github.com/erjui/DHO/tree/main.

cs.LG · cs.AI · cs.CV

新论文 · v1arXiv:2611.21223

Oriented Semi-DETR for Semi-Supervised Aerial Object Detection

Yuqiu Li, Jian Xue, Fan Li, Ke Lu

原文摘要

Semi-supervised object detection has recently been extended to DETR, where oriented boxes introduce an ambiguity that does not arise for axis-aligned boxes: the map from parameters to rectangles is two-to-one, since $(c_x,c_y,w,h,θ)$ and $(c_x,c_y,h,w,θ+π/2)$ denote the same rotated rectangle. Losses and matching costs are computed from the parameters, so two boxes that coincide in space can be far apart in the metric that drives training. A correctly localized student box is then penalized by the regression loss, and the inflated matching cost leads the Gaussian mixture model (GMM) filter to discard it. We propose Oriented Semi-DETR, which removes this ambiguity wherever a pseudo box enters training. Swap-aligned Target Selection regresses towards the encoding closer to the student prediction. A Swap-invariant Reliability Cost minimizes over the two encodings inside the matching cost, so the filter decides on geometry alone. Swap-aware Pseudo-label Denoising starts denoising earlier, ramps its loss weight linearly, and weights each query by the teacher score and encoding margin. All three are parameter-free and require no additional forward pass. Across all label ratios of DOTA-v2.0, Oriented Semi-DETR outperforms DETR-based supervised and semi-supervised baselines, with the largest gains at the strict IoU threshold.

cs.CV · cs.AI · cs.MM

新论文 · v1arXiv:2611.21222

Verifiable Random Sampling of Private Data Without a Trusted Single Party

Muhammed Karamuk, Mikołaj Glinka, Ruben Wolff

原文摘要

Problem: When a party supports claims about a private dataset by disclosing a sample, auditors cannot tell whether it was drawn at random or "cherry-picked" to hide flaws in the rest of the data. Solution: The data holder commits to the hashes of every item in its dataset and registers the commitment with independent servers. Selection then waits on a decentralized randomness beacon: a network that jointly emits a fresh random value every few seconds, outside the control of the data holder and the auditors. The value emitted two beacon rounds after the commitment picks the sample. Each server returns a signed receipt, so anyone can later check that the revealed sample is exactly the one that value prescribes. Selection from the committed set stays unbiased as long as the beacon's threshold holds and one required server is honest, and a pre-agreed name, checked with the servers, stops repeated or abandoned draws. We call this protocol NoCherry. Applications: Any audit that spot-checks a private corpus: private AI benchmarks and held-out test sets, dataset quality claims, code and security audits, test-fairness review, vulnerability handling, and compliance sampling. Auditors need not trust the data holder to hand over a representative sample of the committed data, since the protocol proves it, and data holders need not expose their entire dataset.

cs.CR · cs.AI

新论文 · v1arXiv:2611.21218

InclusionMed: Benchmarking AI for Inclusive Healthcare

The InclusionMed Collaboration

原文摘要

Advancing medical AI for inclusive healthcare requires evaluation across diverse healthcare roles, tasks, and care settings, but benchmark coverage remains uneven. We introduce InclusionMed, an open medical AI evaluation initiative for inclusive healthcare. The InclusionMed Atlas organizes benchmarks across three healthcare roles and three AI capabilities. For the first release, we selected four public benchmarks from 388 candidates through structured audits and independent reproduction. With InclusionMed-DxFiles, our private diagnostic reasoning benchmark, the Atlas contains five benchmarks covering three of its nine role--capability cells. We evaluated frontier large language models on the Atlas and aggregated their results into the InclusionMed Index, with overall scores ranging from 17.5 to 49.1 out of 100. The Index also reports scores by healthcare role and AI capability, showing that a model's rank varied with the role or capability considered. A within-benchmark analysis using task-level tags showed that the ten highest-ranked models in the Index all scored higher on patient communication than on clinical decision-making. Models with similar Index scores also differed substantially in estimated inference cost. These findings highlight the need for evaluation covering more diverse healthcare needs and for context-specific model selection. Through open collaboration and community contributions, InclusionMed seeks to ensure that diverse healthcare needs worldwide shape how medical AI progress is defined and measured.

cs.AI

版本更新 · v2arXiv:2611.03330

The SIMPLE Law for Energy Cost of LLM Inference

Hiari Pizzini Cavagna, Andrea Proia, Daniele Cesarini, Zeynep Kiziltan, Andrea Bartolini

原文摘要

Generative AI is billed per token, yet tokens are not equally expensive. Each token enters the KV-cache and raises the cost of those that follow, so energy grows non-linearly along a sequence whose tokens look identical. No analytical model states what a token costs without per-LLM fitting, and no law names the factors that set it. From closed-form expressions of LLM inference's computational and memory-access complexity, we derive the SIMPLE Inference Model to Predict LLMs Energy (SIMPLE), a four-parameter law that separates prefill from decode, and decode into KV-cache and next-token computation cost. Over traces of 13 dense LLMs from several families on NVIDIA H100 GPUs, it attains R2 = 0.969 and 11.97% MAPE, 3.1-3.5x below the SOTA, without per-LLM calibration. SIMPLE explains as well as predicts: prefill energy grows quadratically with hidden dimension, so larger LLMs suffer most from long inputs; and higher GQA KV grouping reduces the per-token energy linearly relative to standard MHA. Increasing the reasoning token budget pushes output length beyond the analytically derived energy minimum, driving super-linear energy growth; accuracy rises but energy rises faster, so Intelligence per Joule peaks at an intermediate budget and falls by up to 3.14x across a 4.0x range. Moving to agentic serving raises energy per request by 13.0x, as both prefill and KV-cache terms grow with input length; prefix caching removes most of the prefill cost (-93%), but not KV-cache cost, so even a cached agentic request costs 4.8x more.

cs.AI · cs.LG · cs.PF

新论文 · v1arXiv:2611.21178

TRACK: Tool-Augmented Reasoning over Time Series for Fault Diagnosis in Power Management System

Erkan Bayram, Debanjan Ghosh, Li Zhu, Sambarta Dasgupta

原文摘要

Fault diagnosis in power management systems running multiple loads relies on experts manually inspecting current and voltage waveforms, which fails to scale across deployed units. We present TRACK, which pairs a time series encoder with a Small Language Model to produce structured, physics-grounded diagnostic reports, calling tools for exact, verifiable calculations. It operates on-device, keeping proprietary sensor data local and avoiding expensive API calls. We introduce a rubric-based framework that evaluates these reports by scoring structure, signal fidelity, physics reasoning, diagnosis, and actionability through per-criterion verdicts. On a 15-class multi-load simulated benchmark, TRACK achieves over 95% fault-category accuracy and generalizes to out-of-distribution faults, whereas the best encoder-probing baseline reaches only 58% accuracy.

cs.LG · cs.AI · cs.CL · eess.SY