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

+0新增论文
+0其中 AI 相关
+0已有论文修订
+91新增 AI 导读

arXiv 最近完成扫描:2026-11-22 21:02 · 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 2026-11-22 00:02 — 2026-11-23 00:02(北京时间) · 点击数字查看对应简报;AI 相关包含在新增论文内。

最近完成检查:2026-11-22 21:02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 23,345 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-22 21:00(北京时间)
新论文 · v1arXiv:2611.25590

Measuring Hallucination in Large Language Models for Cyber Threat Intelligence: An Exploratory Study

Seungho Song, Donggyu Lee, Jaehan Kim, Jiyoung Choi, Jinwoo Kim

原文摘要

Large language models (LLMs) are increasingly used in cyber threat intelligence (CTI) tasks, such as summarization and question answering. However, they may produce hallucinations, false or misleading content not grounded in the source text, which can mislead analysts and corrupt downstream security decisions. This paper presents an exploratory measurement study of hallucinations in LLM-based CTI. We introduce HalluVision, a reproducible framework that generates LLM outputs, extracts domain-specific entities using a fine-tuned named entity recognition model, and evaluates factual consistency using multiple similarity metrics. Using 4,940 real-world cybersecurity articles, we measure hallucinations in outputs from two commercial LLMs and one open-weight LLM. Measurements are validated against 100 human-annotated outputs and two established faithfulness detectors. Results show that semantic similarity aligns with human judgments, whereas entity-based metrics exhibit limited agreement. Differences between models are concentrated in question answering, while entity extraction quality limits what entity-level metrics can reveal.

cs.AI · cs.CR

新论文 · v1arXiv:2611.25588

Rubric Semantics or Numerical Priors? Auditing Scale-Direction Bias in LLM Judges

Chanwoo Kim, Doyun Kim, Minsoo Song, Sihyeon Lee, Chanjun Park

原文摘要

Recent benchmark evaluations increasingly rely on Large Language Models (LLMs) as automatic judges. This paradigm assumes LLMs evaluate based on the rubric's semantic meaning rather than its numeric scale. We question this assumption by introducing semantic faithfulness: the principle that semantically equivalent rubrics must yield consistent scores. To quantify this, we propose a scale-reversal probe that inverts score anchors without altering evaluation criteria. Furthermore, we introduce floor normalization to calibrate reversal disagreement against intrinsic model variability, enabling fair cross-model comparisons. Experiments across five meta-evaluation benchmarks with seven diverse LLM judges spanning open-weight and proprietary models reveal widespread violations of semantic faithfulness, showing that raw reversal agreement alone is misleading. Finally, we demonstrate that bidirectional anchor averaging reduces dependence on any single anchor direction without compromising alignment with human judgments. Our findings emphasize that trustworthy LLM judges must reliably follow rubric semantics, not just correlate with human ratings.

cs.CL

新论文 · v1arXiv:2611.25587

Cognitive-Horizon Action Chunking: Memory-Guided Adaptive Reasoning for Embodied Navigation

Weile Gong, Zijian Lu, Jiahao Zheng, Bo Xu, Yiping Zuo, Jiajia Guo, Xin He, Weijun Wang 等 9 位作者

原文摘要

Multimodal foundation models are advancing embodied navigation, but frequent inference incurs substantial computational cost. We present Cognitive-Horizon Action Chunking (CHAC), a training-free framework for memory-guided adaptive reasoning with a frozen navigation policy. Retained visual context, completion cues, and controller feedback determine the cognitive horizon, or how many cached waypoints the robot executes before refreshing its prediction. At the policy's stop request, local visitation history guides a bounded correction with at most one extra movement and no additional model calls. CHAC reduces vision-language model calls by 33.4% on HM3D-OVON and 64.7% on R2R-CE, with success-rate decreases of 0.63 and 10.82 percentage points, respectively. Success rate and success weighted by path length improve per model call on both benchmarks. Across four physical navigation tasks with 40 runs per method, CHAC achieves 18 successes versus 14 for native execution, reducing model calls by 54.9% and GPU inference time by 49.7%.

cs.RO

新论文 · v1arXiv:2611.25586

Learning Networked Dynamical Systems via Neural ODE-based Graph Structure Inference

Peiyan Sun, Jinming Yang, Gujia Gao, Yaohui Jin, Yanyan Xu

原文摘要

Inferring the latent interaction graph from multivariate time series is fundamental to understanding complex dynamical systems in neuroscience, biology, and physics. Despite recent progress in neural relational inference, most methods rely on discrete-time modeling, which degrades under sparse and irregular observations and tends to entangle multi-hop dependencies into spurious direct edges. We show that a continuous-time formulation naturally decouples inference from any fixed observation schedule, allowing gradient signals to be integrated continuously between observations for both dynamics and structure learning. We propose NOGI (Neural ODE-based Graph Inference), a framework that couples a deterministic graph learner with an MPNN-based Neural ODE. The two modules are trained end-to-end via the adjoint method. Across four simulated dynamical systems at two scales, NOGI achieves the best structure recovery in 30 out of 32 evaluated cases with highly competitive forecasting performance, remaining robust under observation degradation. On the realistic Netsim fMRI benchmark, it attains an average AUPRC of 0.90 (AUROC 0.96). These results suggest that explicitly decoupling relational interactions from self-dynamics within a continuous-time vector field provides a powerful mechanism for graph structure learning under realistic observation conditions.

cs.AI

新论文 · v1arXiv:2611.25585

AI-Driven Transformation of Digital Platform Work: A Task-Based Model for Measuring Labour Disruption and Vulnerability, with Early Evidence from Online Freelancing in Kenya

Farhan Latif

原文摘要

Assessments of AI's impact on labour often rely on conventional occupational categories or broad market indicators. This dominant trend, however, obscures how tasks within the same occupation vary in practice across work arrangements. This limitation is particularly important in online freelancing, where workers are typically remote and devote substantial time performing unpaid tasks. This article introduces a task-based framework that distinguishes AI-driven disruption, defined as the extent of change in how work is performed, from livelihood vulnerability associated with that change. The framework is subsequently applied to assess the potential for automation, substitution, and enhancement across four categories of platform work: writing, virtual assistance, transcription, and web and software development. The model estimates high disruption across all four categories, while livelihood vulnerability scores are lower and vary by occupation. Potential efficiency gains on unpaid tasks partly offset potential wage impact associated with paid work in the model. The article compares these estimates with evidence from a survey of 52 online freelancers and 20 follow-up interviews in Nairobi, Kenya. Workers report changes in how they perform tasks and intensifying competition, while their accounts show how adapting their skills and using AI strategically shape their concerns about job loss. The findings show how task-based modelling can provide a baseline for assessing AI-driven changes in platform work and how workers' experiences differ from projections and help explain how these changes may threaten their livelihoods.

cs.CY

版本更新 · v2arXiv:2502.07807

CP-Guard+: A New Paradigm for Malicious Agent Detection and Defense in Collaborative Perception

Senkang Hu, Yihang Tao, Zihan Fang, Guowen Xu, Hangcheng Cao, Yiqin Deng, Sam Kwong, Yuguang Fang

原文摘要

Collaborative perception (CP) is a promising method for safe connected and autonomous driving, which enables multiple vehicles to share sensing information to enhance perception performance. However, compared with single-vehicle perception, the openness of a CP system makes it more vulnerable to malicious attacks that can inject malicious information to mislead the perception of an ego vehicle, resulting in severe risks for safe driving. To mitigate such vulnerability, we first propose a new paradigm for malicious agent detection that effectively identifies malicious agents at the feature level without requiring verification of final perception results, significantly reducing computational overhead. Building on this paradigm, we introduce CP-GuardBench, the first comprehensive dataset provided to train and evaluate various malicious agent detection methods for CP systems. Furthermore, we develop a robust defense method called CP-Guard+, which enhances the margin between the representations of benign and malicious features through a carefully designed Dual-Centered Contrastive Loss (DCCLoss). Finally, we conduct extensive experiments on both CP-GuardBench and V2X-Sim, and demonstrate the superiority of CP-Guard+.

cs.CR · cs.AI · cs.CV · cs.LG

新论文 · v1arXiv:2611.25584

An Explainable and Scalable Hierarchical Tensor Framework for Integrating Multi-Omics and Digital Pathomics

Ahmad Hussein, Ali Anaissi, Mukesh Prasad, Ali Braytee

原文摘要

Background: Precision oncology requires integration of heterogeneous molecular and pathology data, yet many fusion models offer limited biologically grounded interpretability and struggle with missing modalities. Methods: We present Hierarchical Interpretable Tensor Fusion (HIT-Fuse), integrating six modalities: somatic copy number variation, DNA methylation, RNA sequencing, microRNA expression, RPPA protein abundance, and vision transformer-derived pathomics across four TCGA cohorts (glioma, lung adenocarcinoma, breast, and kidney cancer). Modality-specific autoencoders compress high-dimensional inputs, followed by two-stage CP tensor decomposition. Missing latent modalities are reconstructed using clinical covariates and observed omics to expand cohorts and retrain downstream models. A mathematical traceback mechanism enables global and local xAI attributions. Results: HIT-Fuse identifies distinct latent subgroups with significant survival stratification in glioma, breast, and kidney cohorts, both with complete and reconstructed modalities. Cohort expansion yields cohort-specific gains, including a 56.4% relative increase in weighted F1 for glioma. Biological validation against DepMap CRISPR screens, miRTarBase and RPPA annotations supports decoded drivers across all four cohorts. Essential gene rates range from 6.2% in glioma to 7.6% in lung cancer; example drivers include ANAPC11, SNRPD1, PLK1 and RPL14. Of decoded miRNAs, 71-92% have miRTarBase regulatory support, while 77-84% of selected genes are covered in DepMap. Conclusions: HIT-Fuse combines hierarchical tensor fusion, missing-modality reconstruction and traceable attributions to support interpretable multi-omics and digital pathology-based cancer prognosis across heterogeneous cohorts.

q-bio.GN · cs.AI

新论文 · v1arXiv:2611.25583

DriftGuard: Online Memory-Drift Detection via Causal Attribution for LLM Agents

Licheng Pan, Yuzhou Chen, Zhixuan Chu, Yichen Sun, Han Zheng, Zhiyu Wang, Hao Wang, Yan Wang 等 12 位作者

原文摘要

Large language model (LLM) agents rely on persistent memory to plan tool calls and complete multi-step tasks, which exposes them to memory drift: an adversarial or misleading memory alters the execution path and final outcome. In this work, we investigate online memory-drift detection, which estimates each retrieved memory's causal contribution to potential drift at serving time. We identify two fundamental challenges. (1) Content-level features are insufficient for causal memory-drift attribution, as perplexity or retrieval relevance alone cannot determine which retrieved memory would change agent behavior if neutralized; (2) Counterfactual attribution is prohibitively expensive for online detection, as it requires replaying downstream agent and tool execution for each candidate memory, exceeding real-time serving budgets. To address these challenges, we propose DriftGuard, a two-stage framework that decouples offline causal supervision from online estimation. Specifically, DriftGuard first constructs causal supervision through repeated paired counterfactual replays, continuing the original and single-memory neutralized trajectory suffixes from a shared pre-action state. The resulting annotations are aggregated by component-wise majority voting into a memory drift score (MDS) that captures action change, attack removal, violation removal, and task recovery. Guided by this supervision, DriftGuard then learns an online risk model that estimates MDS from pre-action information and flags drift-inducing memories via its maximum component risk, without online counterfactual rollouts. Extensive experiments demonstrate that DriftGuard accurately identifies drift-inducing memories and enables efficient online defense. Code is available at https://anonymous.4open.science/r/DriftGuard-7F0B.

cs.AI

新论文 · v1arXiv:2611.25580

FlashX: Scaling Robotic Experience Reuse for Embodied Agents with CBA-Based NAND Flash

Yuseon Choi, Sanghyeon Park, Eunjong Lee, Hanseo Lee, Sangjin Kim

原文摘要

Embodied agents are emerging as a promising paradigm for complex robotic tasks, leveraging Vision-Language-Action (VLA) and World Action Models (WAMs) to generate actions through Diffusion Transformer (DiT)-based iterative denoising. However, existing models often struggle to generalize to out-of-distribution (OOD) environments. We observe that retrieving successful experiences and using them to warm-start denoising enables agents to better adapt to unfamiliar situations while reducing computational overhead. This creates a new opportunity to improve task accuracy and inference efficiency simultaneously, in contrast to conventional acceleration techniques that trade accuracy for computation. However, realizing this opportunity requires scaling experience banks, introducing substantial storage and search overhead that challenges the capacity, bandwidth, and power constraints of edge systems. We propose FlashX, a CMOS Directly Bonded to Array (CBA)-based NAND flash architecture for scalable experience reuse in embodied agents. FlashX exploits CBA's high internal bandwidth for parallel pSLC key search, employing interleaved Compute-in-Flash (CIF) engines to sustain sensing throughput while retaining dense TLC latent storage. Proprioception-aware clustering further reduces search overhead and enables incremental experience acquisition without global reclustering. Evaluations across VLA and WAM workloads demonstrate up to 12.24\%p and 7.43\%p improvements in task success under dense and compute-efficient configurations, respectively. FlashX achieves 2.57$\times$ higher search throughput than the strongest prior NAND search architecture under pSLC sensing, with a further 9.7$\times$ search reduction through clustering. This delivers 1.29--3.95$\times$ end-to-end speedup and 1.12--2.34$\times$ energy efficiency improvement over SSD-based retrieval.

cs.AR · cs.AI · cs.RO

版本更新 · v2arXiv:2611.20448

MMIntentFlow: An Intent-Driven Real-Time Multimodal Search System with Image-Text Queries at Xiaohongshu

Dongjia Zhao, Yuchen Ma, Chunsheng Shi, Ziqiu Gui, Chao Yuan, Jianing Zhu, Zhaoxing Zhang, Yahui Liu 等 12 位作者

原文摘要

Multimodal search with image-text queries helps users find notes in user-generated content (UGC) communities and serves as a key search tool for multimodal search agents. For such queries, query-note relevance depends on the user's retrieval intent and cannot be determined by visual or topical similarity alone. Simply reranking visual search results cannot fully address this challenge, as notes relevant to the user's intent may be missing from the initial candidate set. This motivates explicit intent modeling to align candidate retrieval and reranking under consistent relevance criteria. We present MMIntentFlow, an intent-driven real-time multimodal search system deployed at Xiaohongshu over an index on the order of ten billion entries. MMIntentFlow represents retrieval objectives with a hierarchical intent taxonomy and uses intent-specific relevance criteria to guide data construction, representation learning, and reranking. An offline agentic data workflow acquires candidates and verifies their relevance under these criteria to produce training supervision. The embedding model combines intent-aware contrastive learning, graded ranking, and distribution distillation, while a compact reranker assesses query-candidate evidence conditioned on predicted hierarchical intents. Evaluations demonstrate that MMIntentFlow delivers substantial gains in search engagement while supporting real-time search at industrial scale. In a user-level A/B test on the Xiaohongshu platform, MMIntentFlow outperforms the existing production system based on visual search followed by reranking, achieving relative gains of 28.17% in Top-4 query click-through rate, 18.96% in query click-through rate, and 12.56% in average time spent on search results pages per user. Production monitoring records an end-to-end P50 latency of 1.07 s.

cs.IR

新论文 · v1arXiv:2611.25578

QFT-Based Hybrid PointNet-Quantum Neural Networks for LiDAR Point-Cloud Classification

Anurag Kulkarni, Nouhaila Innan, Muhammad Shafique

原文摘要

Hybrid quantum-classical learning architectures provide a promising framework for integrating quantum feature transformations into complex data-processing tasks. Among these transformations, the Quantum Fourier Transform (QFT) offers a structured approach to feature processing, motivating an investigation of how its circuit configuration influences predictive performance, quantum resource requirements, and noise robustness. We present a hybrid PointNet-quantum neural network architecture for LiDAR point-cloud classification, combining classical feature extraction, amplitude encoding, and a variational quantum circuit. Five Fourier-processing configurations are systematically evaluated: No-QFT, Full-QFT-After, QFT-no-swap, IQFT-After, and Trainable-QFT. Experiments on KITTI, nuScenes, and PandaSet across three random seeds assess classification performance, class-level behavior, circuit complexity, and sensitivity to the IBM FakeManilaV2 noise model. Under ideal simulation, IQFT-After achieves the highest weighted F1-score on KITTI (0.857 $\pm$ 0.033), compared with 0.838 $\pm$ 0.023 for No-QFT. Removing the final SWAP gates in the QFT circuit reduces circuit complexity while preserving the Pauli-Z measurement features up to a fixed permutation that can be absorbed by the classification head. Under quantum noise, the circuit ranking changes, with IQFT-After exhibiting greater performance degradation and No-QFT and Trainable-QFT showing stronger robustness. Per-class analysis further identifies dataset-dependent differences in minority-class performance. These results demonstrate that Fourier-processing configuration influences both classification behavior and noise sensitivity, highlighting the role of circuit structure and hardware-aware evaluation in hybrid quantum learning for three-dimensional point-cloud data.

cs.CV · quant-ph

版本更新 · v2arXiv:2609.31568

DeepEdu-F: A Framework for Efficient and Scalable Agentic LLMs in Vietnamese Education

Quang Nguyen, Hieu Nguyen, Toan Pham, Hien Hoang, Cong Tran, Nam Vu

原文摘要

Cloud AI assistants send student data to foreign servers, conflicting with data-sovereignty laws such as Vietnam's Decree 53, and are not grounded in the national curriculum, so they often hallucinate on local content. Self-hosting an open model keeps data on-premise but faces two costs that weight quantization does not remove. An AI tutoring session revisits the same long material every turn, so once prefix caching removes the prefill cost, decoding over a growing KV Cache dominates response time; and grounding the tutor in local knowledge by fine-tuning is too costly to repeat as that knowledge changes. We present DeepEdu-F, a framework for the Vietnamese AI tutor DeepEdu with two components. LastSet selects the critical KV Cache entries once per turn from the question's attention, decodes the whole turn from that set, and keeps the full cache so later questions can select again. REPAIR (Reliable Evidence-based Playbook Adaptation and Insight Revision) curates a verified playbook from past interactions instead of updating weights. LastSet decodes up to 3.1x faster than dense decoding at a 256k-token context, faster than four sparse baselines, with accuracy close to full attention. Compared with ACE (ADD), REPAIR raises final accuracy from 70.0% to 80.5% on Formula and from 53.6% to 59.7% on FiNER, while mean task/scenario goal completion across AppWorld's normal and challenge splits increases from 9.28% to 16.28%.

cs.AI

新论文 · v1arXiv:2611.25577

Design and Evaluation of Epira: An Automated Interview System for Case Investigation and Contact Tracing

Ishita Asnani, Shaghayegh Fazliani, Zachary Ryan, Kriyoshig Prasai, Chirag Kumar, Indra Gonzalez Ojeda, Marc Lipsitch, Madeleine Udell

原文摘要

Objective: Case investigation and contact tracing are labor-intensive. Web forms struggle when respondents need clarification before disclosing contacts, and open-ended chatbots may deviate from protocols. We developed and evaluated Epira, a protocol-controlled large language model (LLM) interviewer for multi-disease case investigation and contact tracing. Materials and Methods: Epira pairs LLM dialogue with a rule-based controller enforcing required questions, safety handoffs, and a structured case record, and adapts to new diseases by questionnaire upload. Local (Qwen2.5) and cloud (GPT-5.4) backends were tested. Study 1 used 108 simulated cases with ground-truth fields and contacts for measles and gonorrhea under Epira and open-ended conditions (864 sessions). Study 2 enrolled 72 volunteers in a 2$\times$2 disease-by-model interview assessing completion, usability, acceptability, and trust. Results: Across Study 1, Epira with a local LLM captured 88.9% of required fields and 60.1% of ground-truth contacts. Open-ended interviewers recovered 31.1-41.9 percentage points fewer required fields than the same models in Epira. In Study 2, completion was 100%; mean adapted System Usability Scale (SUS) was 82.95 $\pm$ 16.17 on a 0-100 scale and mean trust was 3.85 $\pm$ 0.77 on a 1-5 scale. Discussion: Protocol control drove completeness more than model choice. Local and cloud deployments produced similar records, supporting privacy-preserving local use with human review of routed interviews and missed contacts. Conclusion: Epira recovers required fields and contacts under diverse disclosure conditions, works with local and cloud LLMs, and received favorable usability and trust ratings.

cs.LG · cs.HC

新论文 · v1arXiv:2611.25575

Trajectory-Calibrated Merging of Embodied Policies

Wenshuo Chen, Honglei Jin, Le Yang, Jianfei Song, Shaofeng Liang, Lei Wang, Junze Zhu, Kailong Wang 等 14 位作者

原文摘要

Robot foundation models are increasingly adapted into task specialists to expand their embodied capabilities. Yet these capabilities remain fragmented across separate policies, and existing model-merging methods do not reliably preserve them during closed-loop execution: parameter-space alignment provides only an indirect proxy for behavior, while regression may fail to preserve behavior at states not covered by its calibration inputs. We present Trajectory-Calibrated Regression (TCR), an execution-aware, gradient-free framework that consolidates robot specialists into a single policy. Across two complementary timescales, TCR collects calibration inputs from expert rollouts spanning different stages of task execution and from intermediate states within iterative action generation. To account for representation shifts introduced by merging, TCR recomputes each block's input through the already merged blocks. It then uses each expert's computation within the block to form module inputs and fits merged weights to match the experts' linear outputs. A two-pass calibration procedure further refines the merged policy using separate rollout caches while retaining the frozen specialists as teachers. The resulting model preserves the original inference graph and requires neither expert routing nor additional gradient training. On LIBERO, TCR merges four $π_{0.5}$ specialists and achieves 77.83% success across 1,200 evaluation episodes, outperforming FeatCal, RegMean, and direct weight averaging by 9.50, 24.17, and 33.00 percentage points, respectively.

cs.RO · cs.CV

新论文 · v1arXiv:2611.25574

VidToC: Rethinking Video Token Compression for Multimodal Large Language Models

Zihan Xiao, Boshen Xu, Pei Fu, Jianzhong Ju, Zhenbo Luo, Jian Luan, Qin Jin

原文摘要

Extending multimodal large language models (MLLMs) from images to videos faces substantial memory and latency challenges due to the massive number of visual tokens. These challenges have driven the development of numerous visual token compression methods and more efficient LLM architectures. However, existing compression methods are often evaluated with different training setups, tasks, and models, making fair comparison difficult. Meanwhile, the growing use of hybrid Mamba-Transformer architectures raises questions about whether compression strategies transfer across backbones. We present VidToC, a controlled evaluation framework for representative compression algorithms, LLM backbones, and time encoding strategies, covering training-free and fine-tuned settings on three complementary video understanding tasks. Our results reveal task-dependent compression behavior and characterize the effects of compression placement, time encoding, and backbone choice. These findings provide empirical guidance for designing and selecting video token compression strategies across tasks and training settings.

cs.CV

版本更新 · v4arXiv:2502.16759

Can Explanations Improve Recommendations? Evidence from Prediction-Informed Explanations

Yuyan Wang, Pan Li, Minmin Chen

原文摘要

Recommender systems are central to digital platforms, yet they face a fundamental trade-off between accuracy and explainability: Black-box models achieve strong performance but lack interpretability needed for trust and adoption, while existing explainable AI approaches often generate explanations post hoc or sacrifice accuracy. We challenge this view, proposing that explanations, when designed as an integral component of the system and aligned with predictions, can improve \emph{both} interpretability and accuracy. We introduce RecPIE (Recommendation with Prediction-Informed Explanations), a framework that jointly optimizes recommendation predictions and natural-language explanations. RecPIE embeds explanation generation into the learning loop: predictions guide explanation generation (prediction-informed explanations), which are fed back to refine subsequent predictions (explanation-informed predictions) via alternating training. The LLM is fine-tuned using LoRA and reinforcement learning with a usefulness reward derived from recommendation accuracy. We evaluate RecPIE on large-scale point-of-interest recommendation data from Google Maps, where consumer preferences span diverse place categories. RecPIE improves predictive accuracy by 1.7-3.6\% over state-of-the-art baselines and outperforms the strongest full-data baselines using only 25\% of the training data. In human evaluations with 566 participants, RecPIE explanations are preferred 61.5% of the time (versus 16.6% for the best baseline) and rated closer to human-generated explanations. Together, these results suggest that insights about consumer behavior and AI decisions need not be imposed externally or generated post hoc; they can be learned as an integral component \emph{within} the AI systems, allowing systems themselves to discover which insights are useful for improving predictive performance.

cs.IR

新论文 · v1arXiv:2611.25570

Reinforcement-Aligned Probabilistic Weather Forecasting with Discrete Tokens

Jinhan Li, Zhisheng Chen, Yuxuan Li, Yuan Gao, Fan Xu, Hao Wu, Jinlong Du, Kun Wang

原文摘要

Global medium-range weather forecasting must not only generate ensembles efficiently to characterize uncertainty in future states, but also serve downstream applications such as extreme-event warnings and cross-sphere coupled simulations. We introduce WeatherCraft, a framework that unifies probabilistic generation and reinforcement learning alignment through discrete weather tokens, bringing the GPT-style pretraining--post-training paradigm to global medium-range weather forecasting. A multiscale tokenizer represents atmospheric states as a coarse-to-fine token hierarchy, while a Transformer models causal dependencies across time and scale and predicts tokens within each scale in parallel. Direct sampling generates diverse ensembles, and exact token log-probabilities enable Group Relative Policy Optimization (GRPO) to maximize downstream task rewards without differentiating through evaluators or coupled models. On the ERA5 benchmark, WeatherCraft's eight-member ensemble mean reduces day-15 normalized RMSE by 17.6% relative to the strongest evaluated deterministic baseline and, with KV caching, achieves an approximately $18\times$ inference speedup over GenCast. Direct fair-CRPS adaptation improves ensemble probabilistic skill, while reward-based post-training improves forecasts of persistent land temperature extremes and coupled ocean heat extremes. This work establishes discrete autoregressive modeling as a general foundation for scalable, task-alignable Earth system foundation models.

cs.LG

新论文 · v1arXiv:2611.25568

AeroBelief: Budgeted First-Action Verification for Evidence-Guided Aerial Sensing

Tianren Li, Dongbo Li, Hongkai Chen, Jie Liu, Guoliang Xing

原文摘要

Aerial sensing must combine uncertain overhead context with current observations while budgeting prediction, movement and image processing. AeroBelief uses the next executable action as the unit of verification against reference scores. It groups candidate views by first action, scans them with an inexpensive proposer, and selectively queries a structured world model (WM). Calibrated group bounds support strict commitment; a six-query controller connects selective prediction with belief updates, acceptance checks and scheduled sensing. Across four layout families, 40 worlds and 960 closed-loop executions, the AeroBelief controller completes 75.42% of tasks, compared with 73.33% for exhaustive WM scoring and 65.83% for line-of-sight geometry, using 91.81% fewer reference-candidate evaluations than exhaustive scoring. In fixed-reference replay with a six-query cap, proposer-scaled anchored bounds raised first-action certificate availability from 36.25% to 54.69% and reduced mean queries per state by 13.54% relative to unit-scale anchored bounds. A complete ground-hosted pipeline processes 2,685 self-collected 20-megapixel unmanned aerial vehicle (UAV) images from seven DJI Mavic 3M routes. Context-guided admission and feedback retain all eight independently supported target confirmations with 76.05% fewer processed images and 70.02% less accumulated active processing service than uncapped all-frame processing; feedback saves another 5.17% of images relative to static context. The studies connect budgeted action selection with evidence-driven processing of recorded flight imagery.

cs.AI

新论文 · v1arXiv:2611.25567

Online Mistake Detection in Instruction-Conditioned Procedural Video

Amit Feldman, Ilan Shimshoni, Rotem Dror

原文摘要

An assistant monitoring a person performing a procedure is most useful when it can detect mistakes while there is still time to recover. This requires judging actions against task instructions as the video unfolds, identifying the attempted step, locating it in time, and assessing its execution. Existing online detectors do not use the instruction text, while instruction-conditioned methods operate offline. In this paper, we study online, instruction-conditioned mistake detection in egocentric videos. Using only observed video, task instructions, and prior predictions, our model identifies and localizes completed action attempts and classifies them as correct, modified, slip, corrections, or additions; omitted steps are determined after the video ends. We compare a modular pipeline, with separate step identification, temporal localization, and execution verification, to a joint model that predicts all outputs in a single step. To support supervised learning from mistakes, we repartition EgoPER to include erroneous videos in all splits. The approaches achieve similar step-segmentation F1@0.5 (93.6% and 92.1%), while the joint model achieves higher binary mistake-detection F1@0.5 (84.1% vs. 80.7%), with its largest gain on slips (56.4% vs. 37.5%). Component-level analysis suggests that error propagation between stages contributes to the modular approach's lower detection score. On test videos containing mistakes, both designs outperform released AEM and ESTANet checkpoints under our evaluation protocol.

cs.CV · cs.AI

新论文 · v1arXiv:2611.25566

Diagnosis is not a policy: a state-transition audit of caoability-conditioned test-time inference

Yuchong Hu, Zhao Zhang, Kainan Liu, Haoyu Wang, Siyan Zhang, Jing Xiao

原文摘要

Test-time actions can satisfy retrieval, state-realization, program-consistency, or rendering objectives without correcting an answer. Test-time Error Response and Realizability Audit (TERRA) is an evaluation/auditing protocol, not a new inference algorithm. It uses a unified decomposition of intervention outcomes into applicability, target-state realization, endpoint translation, harm, and resource/inference cost. Paired comparisons fix the incoming state and evaluation conditions; cross-model analyses vary the receiving configuration. Executable financial reasoning experiments yield three findings: the same context expansion can harm one configuration and benefit another; sufficient evidence does not ensure correctness because configurations differ in evidence-to-answer translation; and diagnostic headroom---accuracy attainable with idealized diagnostic information over the measured baseline---plus local consistency does not ensure practical gains. Action value is $V(a\mid x,m)$, not diagnosis alone; diagnosis is not a deployable policy $π(a\mid x,m)$.

cs.AI

新论文 · v1arXiv:2611.25562

AZOWL AgentOps: Task-Scoped Governance for Deployed Agent Applications

Taeoh Hong

原文摘要

AZOWL AgentOps provides shared governance for deploying agents and delegating work to them. Human and agent senders can exchange messages or explicitly delegate a business task. Accepted work has a stable identifier that links the sender's authority, the assigned runner, any required approval, and result evidence. The default enrolled customer PC profile is read-only. We compare AZOWL components with a configured LangGraph application in 28 synthetic runs across four paired pilots. Original AZOWL services checked work admission, recipient routing, and workflow prerequisites in isolated tests. The configured LangGraph application supplied business task and dispatch records in application code. Both configurations recorded zero business tasks in dialogue, rejected unauthorized agent handoffs, and issued zero writes when an approved delivery address changed. AZOWL's scope gate, exercised as a component, requested approval again on the same task. A product probe reused one registered provider across two workflows and checked prerequisite enforcement for ticket-task submission. The paired protected writes, including ticket creation, used mock providers. Of two deployed PC inventory reads with verified receipts, we traced the 0.6.57 run from submission through runner pickup and tool execution to the receipt, with matching local and central artifact hashes. A separate deployed GRC audit followed an operator-driven ticket lifecycle on sample records.

cs.CR

新论文 · v1arXiv:2611.25561

Passing NPCR Without Diffusion: Exact Signatures, a Detection Benchmark and an Audit of Published Image Ciphers

Mua'ad Alfarajat

原文摘要

NPCR, the share of ciphertext bytes changed by a one-pixel plaintext change, is reported for nearly every image cipher as evidence of diffusion, with 99.61% the target. Several mechanisms produce values near it, and the mean does not say which. We separate three regimes: no propagation, one round of forward and backward chaining linear over addition or XOR, and independent ciphertexts. For the second an exact count applies. After a least-significant-bit change its mean asymptotically matches the independent-ciphertext reference, while its spread without cancellation is about five to twenty times larger. In simulation, the one-sided NPCR test of Wu, Noonan and Agaian flags one-round additive, mixed and S-box chaining in 2-4% of papers; read in both directions, the same five values flag at least 99.9%. Two-sided testing does the detecting; the spread and the other signatures help diagnose the structure. We then audit twenty-one published schemes. The count matches a reimplementation of one published diffusion layer on every trial and accounts for its reported table, the signatures show two schemes whose printed specifications cannot produce their reported values, and find that three of our own report plain-versus-cipher values and fall to two chosen images; ten more change their encryption material with the plaintext or per encryption, so their NPCR mostly reflects that change. A released checker applies the signatures to published tables, showing what a table measures within stated assumptions; it does not certify security.

cs.CR · cs.MM · eess.IV

新论文 · v1arXiv:2611.25560

Is a Description Worth a Thousand Demonstrations? Specificity and Reliability of Task Descriptions in In-Context Learning

Bruce Nguyen, Bei Jiang, Linglong Kong

原文摘要

A prompt for in-context learning (ICL) can describe a task in addition to, or instead of, demonstrating it through examples. Yet how task descriptions work remains poorly understood theoretically. Through the Bayesian view of ICL, we model descriptions as informative priors characterized by their specificity and reliability. We adapt the effective sample size (ESS) from Bayesian statistics, quantifying a description's predictive value as an equivalent number of demonstrations. For in-context linear regression, we show that a reliable description's ESS grows with its specificity and with the task's dimension. A description that may be irrelevant, however, gains little from its specificity unless demonstrations are given. In practice, do in-context learners use descriptions as a Bayes-optimal predictor would? Our experiments show that meta-trained Transformers come close to the Bayes-optimal ESS when tested on their training distribution. Pretrained language models, however, make too little use of highly specific descriptions when answering without reasoning; most of them closely match the Bayes-optimal predictor only when reasoning is enabled. These results extend the Bayesian analysis of ICL from demonstrations to descriptions.

cs.LG

新论文 · v1arXiv:2611.25557

The Coverage Selection Gap: Ranking, Not Generation, Limits LLM Rubric Construction

Ali Karkehabadi, Jamshid Hassanpour, Rafael E. Banchs

原文摘要

Task-specific rubrics are increasingly used to evaluate open-ended language-model outputs, yet constructing comprehensive evaluation criteria from complex source documents remains challenging. We study automatic rubric construction in professional mergers and acquisitions (M&A) due diligence, where evaluation criteria must be inferred from task instructions and supporting documents without access to an expert-authored reference rubric. We introduce a multi-axis generation framework that explores complementary structural and free-form views of the source evidence to construct a broad pool of candidate evaluation criteria. To distinguish failures of criterion discovery from failures of candidate selection, we introduce the coverage-selection gap, which measures the difference between reference-free ranking and a reference-informed greedy diagnostic under a fixed selection budget. Developed on 16 tasks and evaluated on four held-out tasks, our framework achieves 0.876 macro recall against senior-authored reference rubrics under an LLM-based criterion-equivalence protocol. Ablation studies further show that introducing new structural search axes improves recall by 6-10 percentage points, whereas increasing generation volume along existing search directions yields at most 1.5 points. These findings suggest that, once a high-coverage candidate pool has been constructed, effective reference-free selection becomes a substantial bottleneck in document-grounded rubric construction.

cs.CL · cs.AI

新论文 · v1arXiv:2611.25555

UniReFP: Robust Unified Fingerprinting for Vision Models against Cross-Task Repurposing Attacks

Ziye Geng, Guang Yang, Yihang Chen, Lingfeng Yao, Xing Shi, Miao Pan, Xin Fu, Changqing Luo

原文摘要

We identify, for the first time, a new model modification attack--the cross-task model repurposing attack--that can render fingerprints generated by existing model fingerprinting approaches inapplicable. To address this challenge, we propose UniReFP--a novel unified model fingerprinting framework applicable to diverse vision models across different tasks--that utilizes ownership evidence in a shared class-level semantic space rather than original task-specific output space, enabling effective ownership verification when a protected model is repurposed across tasks with heterogeneous output formats. Specifically, UniReFP constructs a surrogate classifier by repurposing the protected model, enabling fingerprints to be generated independently of the original task-dependent output space of the protected model. Moreover, it optimizes grouped fingerprints to encode ownership evidence: fingerprints within the same group are encouraged to induce consistent semantic responses and intermediate representations on the surrogate classifier, while producing dispersed responses on independently trained reference classifiers. During verification, UniReFP uses a task-aware semantic projection mechanism to map heterogeneous suspect-model outputs into unified semantic-presence vectors, and determines ownership by measuring inner-group consistency. Extensive experimental results validate the effectiveness of UniReFP, consistently showing superior performance under cross-task repurposing attacks, commonly studied model modification attacks, and even their combinations.

cs.CV · cs.CR

版本更新 · v6arXiv:2607.24197

Bregman Power Divergences

Frank Nielsen

原文摘要

Power distances between weighted Euclidean points are fundamental dissimilarity measures in computational geometry, used notably for constructing power diagrams and their dual weighted regular triangulations. Bregman divergences, on the other hand, are fundamental deviation measures in information theory and serve as canonical divergences of dually flat spaces in information geometry. In this work, we introduce Bregman power divergences, an extension of power distances to Bregman divergences, and provide two complementary geometric interpretations of that measure based on (i) the Bregman potential lifting and on (ii) the Bregman Pythagorean theorem. We show that Bregman power divergences between weighted parameters amount to equivalent ordinary power distances expressed using the mixed-parameterization induced by the convex duality of Bregman generators. This equivalence enables Bregman proximity problems to be solved as Euclidean proximity tasks at the expense of using an extra dimension. As an illustration of this equivalence, we reduce the Bregman diameter and closest-pair problems to Euclidean power distance problems. In the second part, we further elucidate the correspondence between Bregman power divergences and power distances through the equivalence of Bregman potential lifting and paraboloid lifting techniques. Last, we consider the space of Bregman and power spheres, and show how to construct a Bregman circumsphere passing through prescribed parameters as a corresponding power circumsphere, called an orthosphere. We further discuss several applications.

cs.IT · cs.CG

版本更新 · v2arXiv:2611.23845

ReflectScene:MLLM-Guided Closed-Loop Diffusion for 3D Indoor Scene Generation

Jiawei Shen, Jie Feng, Junpeng Zhang, Guanbin Li, Wanyi Chen, Weisheng Dong, Guangming Shi

原文摘要

Text-conditioned 3D indoor scene generation requires layouts satisfying geometric, functional, and semantic constraints. However, existing diffusion methods rely primarily on static textual conditions, limiting their ability to assess evolving scenes and translate semantic discrepancies into executable object-level corrections. MLLM-based post-hoc refinement provides semantic feedback but remains decoupled from denoising, potentially disrupting scene coherence. We propose ReflectScene, an MLLM-guided closed-loop diffusion framework. First, we introduce Hierarchical Prior-Grounded Scene Understanding that aligns structured layouts with multi-view visual evidence, to interpret object-specific context and scene global semantics via a frozen multi-modal large language model (MLLM). Second, we propose MLLM-Guided Semantic Reward Evaluation to distill MLLM's diagnoses and assessments on possible candidate corrections into a lightweight Reward Predictor, enabling correction decision making with reduced computation burden. Finally, Decision-Conditioned Trajectory-Consistent Diffusion is proposed to integrate selected corrections through transient conditioning and persistent clean-estimate projection, correcting object layout in the generated scene while preserving non-target arrangements throughout denoising. Experiments on 3D-FRONT demonstrate improved functional consistency and semantic alignment while maintaining competitive generation quality and diversity.

cs.CV · cs.AI

新论文 · v1arXiv:2611.25553

TwoPass: Draft-Guided Few-Step Video Prediction for Robotic Manipulation

Yijun Liu, Shijia Ge, Tongtong Feng, Xin Wang, Yuan Meng, Yuwei Liu, Zhi Wang, Wenwu Zhu

原文摘要

Action-conditioned video prediction lets robots anticipate the visual consequences of their actions for planning and policy evaluation. Existing predictors denoise each future from noise without using their own estimate of it, which limits the quality of the predicted future, in particular of the manipulated objects, and requires many network calls. We present TwoPass, a self-conditioned training-and-inference framework that keeps the standard flow-matching objective. Self-Generated Draft Training supplies a detached clean-video estimate from the shared network as conditioning at the same noisy state. A Refinement-Only Late Adapter applies a dedicated residual transformation when a draft is present. At inference, Draft-Reuse Iterative Refinement obtains both a solver update and the next draft from each velocity prediction, requiring as few as 4 function evaluations (NFE) per video chunk. Training samples discrete solver-query noise levels and updates the pretrained backbone. Extensive experiments on Bridge and Language Table show that TwoPass achieves state-of-the-art results with 4 NFE, far fewer than the 50-59 calls of existing predictors. Code: https://github.com/PlantPotatoOnMoon/TwoPass.

cs.CV

版本更新 · v2arXiv:2609.30798

Evaluating Real-Time Voice Agents: A Survey from Component Quality to Grounded Outcomes

Shivam Negi, Arpit Rawat, Rashi Jain

原文摘要

Real-time voice agents must decide when to speak, recover from disruption, and act only when the conversation requires it. But currently their evaluation is fragmented and is split across speech quality, turn-taking, tool calling, and task completion. We review 136 primary works published up to October 2, 2026, and code each on two axes: the capability an evaluation exercises, namely timing, recovery, grounded outcome, and, where relevant, multiparty conduct; and the strength of its evidence, namely evaluation conditions and grading instrument. The coding locates the gaps. Only 13 of 91 evaluation works check external state or action traces, only 5 of 21 grounded-outcome benchmarks use real human speech, and only 2 of 26 multiparty benchmarks verify state. Three claims follow. Architecture constrains deployment but does not determine results. It has been observed that benchmarks can verify outcomes, but few do. Evaluating an agent solely through one-on-one user interaction leaves critical gaps: it cannot confirm if the agent is speaking to the correct person, verify that person's authority, or identify the proper moments for strategic silence. From the taxonomy we derive TRG, short for Timing-Recovery-Grounded, a minimum reporting standard, and apply it to three published reports with different architectures. In one, tool-selection F1 is 82.5% while exact-call success is 33.0%, with no evidence of external task state. None of the three reports measures task state after an interruption. TRG governs what is reported and assigns no composite score. Per-source codes and tooling accompany the submission.

cs.AI

新论文 · v1arXiv:2611.25551

Anchored Inertial Proposals with Objective Verification for Diffusion-Prior Inverse Problems

Minwoo Kim, Seunghyeok Shin, Dabin Kim, Hongki Lim

原文摘要

Training-free diffusion priors are powerful for inverse problems, but measurement guidance during reverse sampling is local: aggressive updates can improve immediate data fit while disrupting later denoising. We introduce an anchored inertial correction rule that separates proposal generation from acceptance. At each corrected step, the denoiser prediction anchors a frozen clean-space objective combining measurement fit with a noise-level-dependent proximal term. Heavy-ball dynamics with step size and momentum computed from the measured curvature generate a clean-state candidate; a separate verifier evaluates it with the same objective, with the denoiser prediction retained as fallback; both stages are derived from one objective for the whole trajectory, considered one step at a time. The accepted clean state is re-noised to continue sampling. The rule applies to pixel and latent diffusion with differentiable measurement operators, requires no retraining, and adds no denoiser calls inside the correction loop. We analyze how objective verification yields local nonincrease and displacement control. Across natural-image benchmarks, the method achieves competitive reconstruction quality. Code is available at https://github.com/Seunghyeok0715/PIVOT2026.

cs.CV