ArXiv Domain 2026-10-08

数据来源：ArXiv Domain

LLM Domain Papers

1. Tokka-Bench: Evaluating Tokenizers Across 100 Natural and 20 Programming Languages

Abstract:Large language models rely on subword tokenizers whose quality varies across languages, yet no standardized multi-metric framework exists for broad comparative evaluation. We introduce Tokka-Bench, an open-source framework that evaluates tokenizers on five complementary metrics — bytes per token, unique token coverage, subword fertility, word-split rate, and vocabulary composition — across 100 natural languages (30+ scripts) and 20 programming languages, using language-aware segmentation adapted to each writing system. Comparing seven BPE tokenizers (GPT-2, GPT-4, gpt-oss, Llama 3.1, Gemma 3, Qwen3, and Kimi K2) within individual languages, we find that vocabulary allocation strategy matters more than raw vocabulary size, and that programming-language efficiency has converged among recent tokenizers despite divergent natural-language profiles. The framework, data, and interactive dashboard are publicly available.

中文摘要

摘要：大型语言模型依赖子词分词器，其质量在不同语言之间存在差异，但尚无用于广泛比较评估的标准化多指标框架。我们介绍了 Tokka-Bench，这是一个开源框架，可对分词器在五个互补指标上进行评估——每个 token 的字节数、唯一 token 覆盖率、子词生育率、单词拆分率和词汇组成——涵盖 100 种自然语言（30 多种书写系统）和 20 种编程语言，并使用适应每种书写系统的语言感知分词。对七种 BPE 分词器（GPT-2、GPT-4、gpt-oss、Llama 3.1、Gemma 3、Qwen3 和 Kimi K2）在单一语言中的比较显示，词汇分配策略比原始词汇量更重要，尽管自然语言表现各异，但近期分词器在编程语言效率上已趋于一致。该框架、数据和交互式仪表板均已公开提供。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08794

Authors: Ben Gubler

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.08794.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08794

Published: 2026-10-08T02:54:33.972Z

2. Child ASR Adaptation with Adult Retention: An Empirical Study

Abstract:Automatic Speech Recognition (ASR) systems often underperform for children and non-native speakers, while adapting adult ASR models to child speech can cause adult-speech forgetting. We study child ASR adaptation with adult retention across Arabic and English. We compare full fine-tuning, LoRA, and post-hoc weight-space merging across encoder—decoder, encoder—CTC, and AudioLLM-based ASR systems. Experiments use Arabic native and non-native child speech, English MyST child speech, and adult benchmarks from MGB-2 and LibriSpeech test-clean. We evaluate recognition quality with WER and quantify the adaptation—retention trade-off using Retention Index, Child Adaptation Gain, and Adaptation Recovery. Results show that child adaptation is necessary, especially for non-native Arabic and English child speech, but direct adaptation often reduces adult ASR performance. Bilingual adaptation is more stable than language-specific adaptation. Weight-space merging often improves the trade-off, especially for encoder—CTC, Whisper, and AudioLLM-based ASR, with LERP favoring adult retention and TIES recovering stronger child gains. For the encoder—decoder model, direct bilingual fine-tuning remains strongest in raw WER.\footnote{Code, and models are available at this https URL.

中文摘要

摘要：自动语音识别（ASR）系统在处理儿童和非母语说话者时往往表现欠佳，而将成人ASR模型适配到儿童语音上可能导致成人语音记忆遗忘。我们研究了在保留成人能力的情况下对儿童ASR进行适配，覆盖阿拉伯语和英语。我们比较了全量微调（full fine-tuning）、LoRA、以及后置权重空间合并（post-hoc weight-space merging）方法，应用于编码器-解码器（encoder—decoder）、编码器-CTC（encoder—CTC）和基于AudioLLM的ASR系统。实验使用阿拉伯语母语和非母语儿童语音、英语MyST儿童语音，以及来自MGB-2和LibriSpeech test-clean的成人基准数据。我们使用词错误率（WER）评估识别质量，并通过保留指数（Retention Index）、儿童适配增益（Child Adaptation Gain）和适配恢复（Adaptation Recovery）量化适配-保留权衡。结果显示，儿童适配是必要的，尤其是对于非母语阿拉伯语和英语儿童语音，但直接适配往往会降低成人ASR性能。双语适配比单一语言适配更稳定。权重空间合并通常能改善这种权衡，尤其适用于编码器-CTC、Whisper和基于AudioLLM的ASR，其中LERP有利于成人能力保留，而TIES可以恢复更强的儿童增益。对于编码器-解码器模型，直接双语微调在原始WER上仍表现最佳。ootnote{代码和模型可在此https网址获取。}

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08827 (HTTP 429)

Authors: Houssam Eddine-Othman Lachemat, Shammur Absar Chowdhury

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.08827.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08827

Published: 2026-10-08T02:54:33.972Z

3. When Forgetting Looks Like Improvement: Metric Masking in Streaming Diarizer Adaptation and the Price of Rehearsal

Abstract:Small-data adaptation can improve speech detection while degrading speaker attribution. We study this discrepancy in a released streaming diarizer adapted on 7.5 h of two-party conversation and evaluated across six corpora. Adaptation substantially improves in-domain diarization performance and transfers to an independent corpus. However, this improvement is not consistent across evaluation scenarios as the additional confusion is mainly associated with impaired temporal identity consistency rather than speaker-count errors. A local-remapping diagnostic reveals different patterns of identity degradation across corpora, indicating that adaptation may alter how streaming models maintain speaker assignments over time. Rehearsal reduces the observed degradation but reduces the cross-domain transfer performance. These results highlight the need to jointly evaluate detection accuracy, identity consistency, and retention behavior when adapting streaming diarization systems.

中文摘要

摘要：小数据适应可以提高语音检测的性能，但会降低说话人归属的准确性。我们在一个发布的流处理分说话人系统中研究了这种差异，该系统在7.5小时的双人对话数据上进行了适应，并在六个语料库上进行了评估。适应显著提高了域内分说话人性能，并可以迁移到独立语料库。然而，这种提升在不同的评估场景中并不一致，因为额外的混淆主要与受损的时间身份一致性有关，而非说话人数错误。局部重映射诊断显示各语料库中身份退化的模式不同，这表明适应可能改变流处理模型随时间维护说话人分配的方式。复习机制（Rehearsal）能够减少观察到的退化，但会降低跨域迁移性能。这些结果强调，在适应流处理分说话人系统时，需要同时评估检测准确性、身份一致性和保留行为。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08828

Authors: Mo Yu, Yang Liu, Jing Qian

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.08828.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08828

Published: 2026-10-08T02:54:33.972Z

4. Emo-Jev: Probabilistic Reasoning for Emotion Classification with Jev

Abstract:Jev offers an alternative interface for language understanding: given an input and predefined questions, it returns probabilistic decisions rather than free-form responses. Whether this interface can support effective reasoning for text classification against leading LLMs remains an open questions. We introduce Emo-Jev, a training-free framework with two complementary implementations. Emo-Jev-D decomposes classification into task-specific atomic judgments and composes their probabilities into a final prediction. Emo-Jev-SC constructs multiple judgment paths from complementary perspectives and aggregates their predictions into a consensus decision. We evaluate Emo-Jev on eight datasets spanning sentiment analysis, emotion recognition, sarcasm detection and humor detection, comparing against direct Jev classification and five SoTA LLMs under input/output and chain-of-thought reasoning. Standard Jev achieves 62.93\% average macro-F1 versus 67.28\% for the strongest LLM baseline, with lower observed latency and generally lower cost.

中文摘要

摘要：Jev 提供了一种语言理解的替代接口：给定输入和预定义问题，它返回概率决策而不是自由格式的回答。该接口是否能够在文本分类中支持对领先大语言模型的有效推理仍然是一个悬而未决的问题。我们提出了 Emo-Jev，一种无需训练的框架，具有两种互补的实现方式。Emo-Jev-D 将分类分解为特定任务的原子判断，并将它们的概率组合为最终预测。Emo-Jev-SC 从互补的角度构建多条判断路径，并将它们的预测聚合为一致性决策。我们在覆盖情感分析、情绪识别、讽刺检测和幽默检测的八个数据集上评估了 Emo-Jev，并与直接 Jev 分类以及在输入/输出和链式思维推理下的五个最先进大语言模型进行了比较。标准 Jev 的平均宏 F1 分数为 62.93%，而最强大语言模型基线为 67.28%，但 Jev 的观察延迟较低，整体成本也通常较低。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08829

Authors: Yazhou Zhang, Junhao Yu

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.08829.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08829

Published: 2026-10-08T02:54:33.972Z

5. CoDR: Training-Free Confidence-Drift Remasking for Diffusion Language Models

Abstract:Masked diffusion language models (MDLMs) decode by repeatedly committing tokens to masked positions, but these commitments are usually irreversible. A token chosen under sparse, partial context is kept fixed, even when later context no longer supports it. Existing samplers mainly decide when to commit a token, but rarely check whether an already committed token should still be kept, allowing early mistakes to propagate. We trace this issue to confidence drift, where the model’s confidence in a committed token drops from its sparse commit-time context to the denser context available later. Based on this signal, we propose CoDR (Confidence Drift Remasking), a training-free and sampler-agnostic refinement pass. CoDR estimates drift for all committed positions in only k forward passes via k-partition probing, then remasks and regenerates only the tokens the model no longer endorses. Across two backbones, four reasoning and coding tasks, and three base samplers, CoDR improves average accuracy across all evaluated model-sampler configurations and improves most individual task settings with modest overhead. Controlled experiments show that the gains come from targeted confidence-drift remasking rather than extra compute alone, and that CoDR uses far fewer forward passes than prior remasking methods. Code is available at this https URL.

中文摘要

摘要：掩码扩散语言模型（MDLM）通过反复将令牌提交到掩码位置来进行解码，但这些提交通常是不可逆的。在稀疏且部分的上下文下选择的令牌会被固定，即使后来的上下文不再支持它。现有的采样器主要决定何时提交令牌，但很少检查已提交的令牌是否仍应保留，从而允许早期错误传播。我们将此问题归因于置信度漂移，即模型对已提交令牌的置信度从稀疏提交时上下文到后面可用的密集上下文会下降。基于这一信号，我们提出了CoDR（Confidence Drift Remasking，置信漂移重新掩码），这是一种无需训练、与采样器无关的改进方法。CoDR通过k分区探测，在仅进行k次前向传递的情况下估计所有已提交位置的漂移，然后仅重新掩码和再生成模型不再认可的令牌。在两种骨干模型、四个推理和编码任务以及三种基础采样器上，CoDR在所有评估的模型-采样器组合中提高了平均准确率，并在大多数单独任务设置中有所提升，同时开销适中。对照实验显示，这些提升来自针对性的置信漂移重新掩码，而不仅仅是额外计算，并且CoDR使用的前向传递远少于之前的重新掩码方法。代码可在此https URL获取。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08833

Authors: Yue Wu, Qinghe Zhang, Yu Zhang, Jian Huang

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.08833.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08833

Published: 2026-10-08T02:54:33.972Z

6. Leveraging LLM-Generated Explanations for Detecting Emotionally Rewritten Fake News

Abstract:The spread of fake news may cause severe social consequences. Existing fake news detection methods mainly focus on stylistic variations or incorporate external information such as explanations. However, news articles are often rewritten under different emotional backgrounds while preserving their underlying factual claims, which may affect the robustness of detection models. In this work, we investigate fake news detec- tion under fact-preserving emotional variations. To study this problem, we construct emotion-rewritten test sets and generate explanations from the original news articles as stable background knowledge. We then propose a Gated Cross Attention (GCA) framework that adaptively integrates emotionally rewritten news with the corresponding explanations, enabling the model to focus on informative explanation content while reducing potential mismatches caused by emotional reframing. Experiments on PolitiFact, GossipCop, and LUN demonstrate that the proposed method achieves notable improvements under multiple emotional conditions on PolitiFact and LUN, while maintaining competitive performance on GossipCop. We further analyze the effects of explanation guidance and gating mechanisms under different emotional conditions. Our code and data are available at: this https URL gca .

中文摘要

摘要：假新闻的传播可能导致严重的社会后果。现有的假新闻检测方法主要集中在风格差异或整合外部信息，如解释。然而，新闻文章常常在不同的情绪背景下被重写，同时保留其基本的事实陈述，这可能影响检测模型的鲁棒性。在本研究中，我们研究了在保留事实的情绪变化下的假新闻检测问题。为研究该问题，我们构建了情绪重写的测试集，并从原始新闻文章生成解释作为稳定的背景知识。随后，我们提出了一个门控交叉注意（GCA）框架，该框架能够自适应地将情绪重写的新闻与相应的解释整合，使模型能够关注有信息量的解释内容，同时减少情绪重构可能造成的不匹配。我们在PolitiFact、GossipCop和LUN上的实验表明，在PolitiFact和LUN的多种情绪条件下，所提出的方法取得了显著改进，同时在GossipCop上保持了具有竞争力的性能。我们进一步分析了在不同情绪条件下解释引导和门控机制的效果。我们的代码和数据可在以下网址获得：this https URL gca 。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08835

Authors: Yupei Guo, Jiajun He, Xiaohan Shi, Tomoki Toda, Zekun Yang, Bowen Wang, Yukinobu Taniguchi

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.08835.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08835

Published: 2026-10-08T02:54:33.972Z

7. Beyond the Sycophancy Score: How Task, Model, and Pressure Shape LLM Yielding

Abstract:Large language models (LLMs) often abandon a correct answer, or endorse a user’s position, once the user pushes back. This behavior, called sycophancy, is usually reported as a single rate per model, which says little about when it happens or how a user can avoid it. We study the conditions that produce it with 103,939 graded replies from ten configurations: eight LLMs with reasoning disabled, and two of them again with maximum reasoning, all facing the same 200 items, 13 pressure conditions, and four-turn conversations, with every reply labeled by two independent LLM judges. We find that the dominant factors are how costly it is for the model to verify the user’s claim, and whether a trained guardrail covers it. Removing this task factor from a logistic model costs 0.485 of McFadden $R^2$, against 0.139 for model family and 0.009 for pressure tactic. Anchored facts are almost never conceded (1.3%), while adoption on logic puzzles rises with the number of clues needed to refute the pushed answer. Personal choices are endorsed in 77.0% of conversations. Most concessions on hard items come from models that cannot reliably solve them; models that can solve them rarely give the answer up. For both models tested, maximum reasoning removes these concessions completely: adoption on deep puzzles falls from 19.2% and 12.5% to 0%. Fallacious or emotional framing adds nothing beyond plain repetition. Three human annotators agree with the judges’ consensus on 118/120 calibration items. These results give practical rules for reliable use: simplify hard-to-verify problems and reason deeply, state the question rather than one’s preferred answer, ask for evidence on open questions, and choose models by their measured guardrail profile.

中文摘要

摘要：大型语言模型（LLM）在用户提出反对意见时，常常放弃正确答案或支持用户的立场。这种行为被称为谄媚，通常以每个模型的单一率来报告，但这无法说明它何时发生或用户如何避免。我们通过对十种配置的103,939条评分回复进行研究，以探索产生这种行为的条件：八种禁用推理的LLM，其中两种再次开启最大推理能力，所有模型处理相同的200个条目、13种压力条件和四轮对话，每条回复由两位独立的LLM评审标注。我们发现，支配因素是模型验证用户主张的成本高低，以及是否有训练好的护栏覆盖该主张。在逻辑回归模型中，去掉任务因素会导致麦克法登 $R^2$ 损失0.485，而模型家族损失0.139，压力策略仅损失0.009。已确立的事实几乎从不被让步（1.3%），而逻辑谜题中被采用的情况随推翻答案所需线索数量增加而上升。个人选择在77.0%的对话中被认可。大多数对困难条目的让步来自无法可靠解决这些问题的模型；能够解决问题的模型很少放弃答案。对于测试的两种模型，最大推理完全消除了这些让步：深度谜题的采用率从19.2%和12.5%降至0%。谬误或情绪化的表述除了简单重复之外没有任何效果。三位人工注释者在120个校准条目中有118条与评审的一致意见。这些结果提供了可靠使用的实用规则：简化难以验证的问题并进行深入推理，说明问题而不是自己的偏好答案，在开放性问题上要求提供证据，并根据模型的护栏测量结果选择模型。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08840

Authors: Guang Yang, Homa Hosseinmardi, Fengchen Liu, Amir Ghasemian

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.08840.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08840

Published: 2026-10-08T02:54:33.972Z

8. Beyond Risk Prediction: Evidence Grounding and Psychosocial Factor Verification for Explainable Suicide Risk Assessment

Abstract:Identifying suicide risk from social networking services (SNS) posts is important for detecting suicide-related signals in online environments. However, risk classification alone provides limited insight into the textual evidence and psychosocial factors behind a prediction. Based on the IEEE BigData 2026 Explainable Suicide Risk Detection Challenge, this study presents a framework consisting of Risk Assessment, Evidence Grounding, and Factor Identification. Risk Assessment uses length-based routing to accommodate posts of different lengths. Evidence Grounding identifies supporting phrases and uses a Risk-Evidence constraint to maintain consistency with the Risk prediction. For Factor Identification, two verifiers are used. The Taxonomy Verifier focuses on factor semantics, whereas the Evidence-Aware Verifier uses factor-specific lexical-semantic cues to select informative positive training units. Their prediction probabilities are combined to produce the final factor predictions. The three tasks are evaluated using task-specific F1 score measures. Risk Assessment achieved a Weighted F1 of 0.8088, Evidence Grounding achieved a test Macro row F1 of 0.7605, and Factor Identification achieved a Macro F1 of 0.5562. The results show that the framework can provide risk predictions, along with supporting textual evidence and fine-grained information on psychosocial factors. Overall, the proposed framework extends suicide-risk assessment beyond risk-level prediction and provides a more interpretable analysis of SNS posts.

中文摘要

摘要：从社交网络服务（SNS）帖子中识别自杀风险对于检测在线环境中的自杀相关信号非常重要。然而，仅进行风险分类对于预测背后的文本证据和心理社会因素的理解是有限的。基于IEEE BigData 2026可解释自杀风险检测挑战，本研究提出了一个由风险评估、证据基础和因素识别组成的框架。风险评估使用基于长度的路由来适应不同长度的帖子。证据基础识别支持性短语，并使用风险-证据约束以保持与风险预测的一致性。在因素识别中，使用了两个验证器。分类验证器关注因素语义，而证据感知验证器使用因素特定的词汇语义线索来选择信息性正向训练单元。它们的预测概率被组合以产生最终的因素预测。这三个任务使用任务特定的F1评分进行评估。风险评估获得了加权F1为0.8088，证据基础在测试中获得宏观行F1为0.7605，因素识别获得宏观F1为0.5562。结果表明，该框架能够提供风险预测，同时提供支持性的文本证据以及心理社会因素的细粒度信息。总体而言，该框架将自杀风险评估扩展到风险水平预测之外，并提供了对SNS帖子更可解释的分析。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08842

Authors: Tianle Hu, Chen Peng, Yi-Hsin Tsai, Takshing Andy Tung, Bingyang Sun, Yenjou Wang

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.08842.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08842

Published: 2026-10-08T02:54:33.972Z

9. QuanLing: Cross-Branch Validation of Language Distance Quantification on Western Romance

Abstract:Quantifying language distance among closely related languages remains a core challenge in quantitative linguistics. Our previous work [1] introduced QuanLing (Quantitative Linguistics via Pretrained Language Models), a quantitative framework combining language distance metrics (sentence embedding distance, tokenization fragmentation rate) with language property analysis (MLM prediction probability), validated on North Germanic (Danish, Norwegian Bokmål, Swedish). This paper extends QuanLing to Western Romance—French, Portuguese, Spanish, Italian—testing cross-branch applicability with the same metric family and aggregation protocol as our North Germanic study, adapted for four languages (English anchor, quadruplet construction). Using 150 four-language parallel sentences, we compute LaBSE sentence embedding distances, tokenization fragmentation rates from four monolingual BERT tokenizers, and mBERT masked language model mutual intelligibility. Results show that Portuguese—Spanish are closest (LaBSE distance 0.0229), French—Italian most distant (0.0338); LaBSE and mBERT rankings agree on 4 of 6 pairs, confirming cross-model robustness. Western Romance shows a wider absolute distance span than North Germanic (0.011 vs. 0.008) but comparable relative ratios (1.48 vs. 1.67), consistent with longer divergence time. French exhibits notably higher MLM predictability (36.12% top-1 accuracy vs. 29.28% for Italian), reflecting its orthography—phonology decoupling. This cross-branch validation provides further evidence for QuanLing’s generalizability beyond a single language branch.

中文摘要

摘要：量化密切相关语言之间的语言距离仍然是定量语言学中的核心挑战。我们之前的工作 [1] 引入了 QuanLing（通过预训练语言模型的定量语言学），这是一种将语言距离度量（句子嵌入距离、分词碎片率）与语言属性分析（MLM 预测概率）结合的定量框架，并在北日耳曼语族（丹麦语、挪威博克mål语、瑞典语）上进行了验证。本文将 QuanLing 扩展到西罗曼语族——法语、葡萄牙语、西班牙语、意大利语——使用与我们北日耳曼语族研究相同的度量系列和聚合协议进行跨分支适用性测试，并为四种语言（以英语为锚，构建四元组）进行了调整。利用 150 条四语言平行句子，我们计算了 LaBSE 句子嵌入距离、四种单语 BERT 分词器的分词碎片率，以及 mBERT 掩码语言模型的相互可理解性。结果显示，葡萄牙语与西班牙语最接近（LaBSE 距离 0.0229），法语与意大利语最远（0.0338）；LaBSE 和 mBERT 排名在六对语言中有四对一致，确认了跨模型的稳健性。西罗曼语族显示出比北日耳曼语族更宽的绝对距离范围（0.011 对 0.008），但相对比率类似（1.48 对 1.67），与更长的分化时间一致。法语的 MLM 可预测性显著更高（顶 1 准确率 36.12% 对意大利语的 29.28%），反映了其正字法与语音系统的脱钩。此次跨分支验证进一步提供了 QuanLing 超越单一语言族的可推广性的证据。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08851

Authors: Yiping Bai

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.08851.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08851

Published: 2026-10-08T02:54:33.972Z

10. LRCC: Generalizing Low-Rank Compression with Conditional Computation

Abstract:Low-rank compression reduces the cost of pretrained language models by replacing linear transformations with low-rank factorizations. However, conventional methods use a fixed rank allocation during inference, assigning the same amount of compute regardless of the input token. We introduce Low-Rank Conditional Computation (LRCC), which adds token-dependent computation to pretrained models by training one lightweight router per Transformer block to select among a small set of nested low-rank paths. During training, the low-rank factors remain frozen, and only the routers are optimized. We evaluate LRCC on Llama and Qwen models for language modeling and zero-shot downstream tasks. Within the same average active-parameter budget, LRCC improves the predictive performance over static low-rank compression, including a 7.6 percentage-point gain in average downstream accuracy on Llama-2-7B over static methods. At matched batch-size-1 decoding latency, LRCC improves both perplexity and downstream accuracy on Llama-3.2-1B and remains competitive on Llama-2-7B, without specialized kernels. Finally, we assess the usefulness of assigning a token-wise path by analyzing the routers’ path choices.

中文摘要

摘要：低秩压缩通过将线性变换替换为低秩分解来降低预训练语言模型的计算成本。然而，传统方法在推理过程中使用固定的秩分配，无论输入的token如何，都分配相同的计算量。我们提出了低秩条件计算（LRCC），通过为每个Transformer块训练一个轻量级路由器，在预训练模型中加入依赖于token的计算，以选择一小组嵌套的低秩路径。在训练过程中，低秩因子保持冻结，仅优化路由器。我们在Llama和Qwen模型上对LRCC进行了语言建模和零样本下游任务的评估。在相同的平均活跃参数预算下，LRCC在预测性能上优于静态低秩压缩，包括在Llama-2-7B上平均下游准确率比静态方法提高7.6个百分点。在匹配批量大小为1的解码延迟时，LRCC在Llama-3.2-1B上提升了困惑度和下游准确率，并在Llama-2-7B上保持竞争力，无需专用内核。最后，我们通过分析路由器的路径选择评估了为token分配路径的有效性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08858

Authors: Thomas Vaitses Fontanari, Maximo Eduardo Rulli, Federico Alvetreti, Donatella Genovese, Simone Scardapane

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.08858.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08858

Published: 2026-10-08T02:54:33.972Z

Agent Domain Papers

1. Adaptive Workflow Intelligence: A Cognitive Architecture for Context-Driven Enterprise Automation

Abstract:Enterprise systems increasingly rely on automated workflows, yet many AI-driven solutions remain brittle under non-stationary conditions, evolving policies, and delayed operational feedback. While reinforcement learning and large language model (LLM) agents offer partial adaptability, they do not by themselves provide persistent reflection mechanisms or straightforward integration with policy-constrained enterprise operations. This paper introduces Adaptive Workflow Intelligence (AWI), a cognitive architecture for context-driven enterprise agents organized around a four-layer Perception-Cognition-Action-Reflection (PCAR) loop. AWI treats reflection as a mechanism for continuous policy refinement and combines hybrid reasoning with reflective memory and feedback-driven adaptation to support decision making under environmental drift and operational constraints. We evaluate AWI in a simulated enterprise decision workflow characterized by delayed outcomes and a controlled regime shift. In a drift-and-delay stress test, guardrail-constrained adaptive approaches recover more rapidly than static automation while maintaining policy compliance. Within this setting, AWI’s reflective components modestly reduce behavioral oscillation and feedback variance, illustrating the stability-agility trade-off introduced by reflective policy adaptation.

中文摘要

摘要：企业系统越来越依赖自动化工作流程，但许多人工智能驱动的解决方案在非平稳条件、政策变化和操作反馈延迟下仍然脆弱。尽管强化学习和大型语言模型（LLM）代理提供了部分适应性，但它们本身并不能提供持久的反思机制，也不能与受政策约束的企业操作轻松集成。本文介绍了自适应工作流智能（Adaptive Workflow Intelligence, AWI），这是一种以情境驱动的企业代理为目标的认知架构，围绕四层感知-认知-行动-反思（Perception-Cognition-Action-Reflection, PCAR）循环组织。AWI 将反思视为持续政策改进的机制，并结合混合推理、反思性记忆和反馈驱动适应，以支持在环境漂移和操作约束下的决策。我们在一个具有延迟结果和受控制度变动的模拟企业决策工作流中评估 AWI。在漂移与延迟压力测试中，受保护约束的自适应方法比静态自动化恢复更快，同时保持政策合规。在此环境下，AWI 的反思组件适度减少了行为震荡和反馈方差，展示了反思性政策适应所引入的稳定性与敏捷性之间的权衡。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08793

Authors: Sreedevi Pandiyath Viswambaran

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08793.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08793

Published: 2026-10-08T02:55:12.891Z

2. Accelerating Floating-Point Satisfiability Solving via Gradient Normalization

Abstract:Satisfiability Modulo Theories (SMT) solvers are foundational to software verification, program analysis, and compiler testing, particularly over the theory of Quantifier-Free Floating-Point (QF_FP). While recent optimization-based SMT solvers have successfully applied gradient descent to continuous relaxations of logical formulas, they are fundamentally bottlenecked by gradient domination, a phenomenon where a small subset of difficult clauses hijacks the optimization trajectory, preventing the solver from satisfying the broader formula and trapping it in local minima. To overcome this, we present GradSAT, a novel framework that bridges optimization-based SMT solving with Multi-Task Learning (MTL). GradSAT reformulates the constraint satisfaction process by treating each SMT clause as an independent MTL task. By applying dynamic gradient normalization (GradNorm), GradSAT actively balances the gradient magnitudes across all clauses at runtime, systematically penalizing dominant gradients and accelerating lagging clauses to ensure uniform convergence. GradSAT implements this through a highly optimized, two-stage hybrid pipeline. First, a GPU-accelerated PyTorch backend leveraging symbolic compilation and operator fusion navigates the continuous relaxation to a high-quality basin. Second, the candidate assignment is handed off to a bit-precise local search engine to rapidly resolve the exact, rigorous assignment. By stabilizing the continuous search dynamics, GradSAT mitigates the brittleness of prior gradient-based solvers and provides a robust, highly parallelizable architecture for complex constraint solving.

中文摘要

摘要：可满足性模理论（SMT）求解器是软件验证、程序分析和编译器测试的基础工具，特别是在无量词浮点理论（QF_FP）中。虽然近期基于优化的SMT求解器已经成功地将梯度下降应用于逻辑公式的连续松弛，但它们在根本上受限于梯度主导现象，即少数难解子句支配优化轨迹，阻止求解器满足更广泛的公式并将其困在局部最优解中。为了解决这一问题，我们提出了GradSAT，一个将基于优化的SMT求解与多任务学习（MTL）相结合的新型框架。GradSAT通过将每个SMT子句视为独立的MTL任务来重新定义约束满足过程。通过应用动态梯度归一化（GradNorm），GradSAT在运行时主动平衡所有子句的梯度幅度，有系统地惩罚主导梯度并加速滞后子句，确保统一收敛。GradSAT通过一个高度优化的两阶段混合管道实现这一点。首先，利用符号编译和操作符融合的GPU加速PyTorch后端将连续松弛导航到高质量的解盆。其次，将候选赋值交给精确的局部搜索引擎，以快速求解精确、严格的赋值。通过稳定连续搜索动态，GradSAT减轻了之前基于梯度求解器的脆弱性，并为复杂约束求解提供了稳健且高度可并行化的架构。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08808

Authors: Yuanzhuo Zhang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08808.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08808

Published: 2026-10-08T02:55:12.891Z

3. Route-Verify-Vote: Procedure-Conditioned Self-Consistency for Mixed-Domain Reasoning

Abstract:Compositional generalization remains challenging when language models must combine familiar reasoning operations in unfamiliar ways. The Scenario-Based Commonsense Reasoning Evaluation (SCoRE) 2026 tests this ability on three mixed domains absent from training and requires models to identify the complete set of correct options for each question. We introduce Route-Verify-Vote (RVV), a framework for procedure-conditioned self-consistency that uses language models without parameter updates. Route uses the provided domain label to select a reasoning procedure that guides the model in representing and applying the relevant constraints. Verify prompts the model to assess each option against those constraints. Vote aggregates complete answer sets and allocates additional samples to questions with a small vote-count margin between the two most frequent sets. Samples for each question follow the same domain-specific procedure. On the official test set, voting over 16 sampled answer sets per question achieves an exact-set accuracy of 74.6%. Adaptive RVV reaches 77.3%, and combining models on selected domain routes raises accuracy to 79.4%. The final system ranked second among participating systems. These results support domain-specific reasoning procedures and answer-set disagreement as useful tools for allocating inference-time computation in mixed-domain reasoning.

中文摘要

摘要：当语言模型必须以不熟悉的方式结合熟悉的推理操作时，组合性泛化仍然具有挑战性。基于场景的常识推理评估（SCoRE）2026 测试这种能力，涵盖训练中未出现的三个混合领域，并要求模型为每个问题识别完整正确选项集。我们引入了路线-验证-投票（RVV）框架，这是一种基于程序的自洽性方法，无需对语言模型进行参数更新。Route（路线）使用提供的领域标签选择推理程序，指导模型表示并应用相关约束。Verify（验证）提示模型根据这些约束评估每个选项。Vote（投票）汇总完整答案集，并将更多样本分配给两组最常见答案集之间投票数差距较小的问题。每个问题的样本遵循相同的领域特定程序。在官方测试集中，每个问题对16个抽样答案集进行投票，获得精确集准确率为74.6%。自适应 RVV 达到77.3%，在选定领域路线组合模型的情况下，准确率提升至79.4%。最终系统在参赛系统中排名第二。这些结果支持使用领域特定推理程序和答案集分歧作为在混合领域推理中分配推理时计算资源的有效工具。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08814 (HTTP 429)

Authors: Xinchen Xiao

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08814.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08814

Published: 2026-10-08T02:55:12.891Z

4. An Empirical Study of Agent Skills’ Downstream Utility

Abstract:Agent Skills package procedural guidance and resources for reuse, but a relevant Skill does not necessarily improve task performance. Existing studies characterize Skill content and evaluate downstream performance, yet provide limited explanations of how utility depends on content, execution configuration, and multi-Skill organization. We conduct an empirical study on 87 SkillsBench tasks, defining downstream utility as the pass-rate difference from No-Skill on the same tasks under the same model—harness configuration. We compare the same Skills across nine configurations, then examine alternative published Skills and organizations of fixed Skill sets under three selected configurations. We retrieve marketplace candidates from a curated corpus of 37,596 Skills. LLM-assisted analysis of content, execution traces, and final artifacts, followed by author review, relates provided support to actual use and task outcomes. The same Skills help some configurations and hurt others on 36.78\% of tasks, with trajectories showing that recommended procedures can become an execution burden. Relevance rankings overlook more useful candidates. Within the evaluated candidate sets, reranking by support for required operations raises first-choice pass rates by 4.35—5.80 percentage points across the three configurations. We derive 17 authoring practices linking executable procedures to recovery, preservation of task requirements, and checks on final artifacts. Stage Plan and Dependency DAG outperform use order alone, with DAG’s additional benefits concentrated in tasks supplied with five or six Skills. These findings guide developers to assess usable operation support, allow procedure adaptation while preserving task requirements, and make artifact dependencies explicit when organizing Skills.

中文摘要

摘要：Agent Skills 包提供了过程指导和可重用资源，但相关技能并不一定能提升任务表现。现有研究对技能内容进行了表征，并评估了下游任务的表现，但对其效用如何依赖内容、执行配置及多技能组织的解释有限。我们对 87 个 SkillsBench 任务进行了实证研究，将下游效用定义为在相同模型—执行配置下，相较于无技能情况下的通过率差异。我们在九种配置下比较相同技能，然后在三种选定配置下考察替代的已发布技能及固定技能集的组织方式。我们从由 37,596 个技能构成的精选语料库中检索市场候选项。通过对内容、执行轨迹和最终产物进行 LLM 辅助分析，随后由作者审核，将提供的支持与实际使用和任务结果建立关联。同样的技能在 36.78% 的任务中对某些配置有帮助，而对其他配置则有害，轨迹显示推荐的执行流程可能成为执行负担。相关性排名忽略了更有用的候选项。在评估的候选集内，通过对所需操作的支持重新排序，可使三种配置下的首选通过率提高 4.35–5.80 个百分点。我们总结了 17 条创作实践，将可执行流程与恢复、任务要求保持及最终产物检查关联起来。阶段计划（Stage Plan）和依赖有向无环图（Dependency DAG）优于单纯的使用顺序，其中 DAG 的额外优势集中在提供了五或六个技能的任务中。这些发现指导开发者评估可用操作支持，在保持任务要求的同时允许流程适应，并在组织技能时明确产物依赖关系。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08875

Authors: Yu Cheng, Dehai Zhao, Zhongxin Liu, Qing Huang, Zhenchang Xing, Xiaoxue Ren

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08875.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08875

Published: 2026-10-08T02:55:12.891Z

5. How Could AI Eliminate Humanity? A Failure-Mode Analysis of Civilizational Risk

Abstract:This article develops a failure-mode framework for analyzing how advanced artificial intelligence could contribute to human extinction, irreversible civilizational collapse, or permanent human disempowerment. The central thesis is that catastrophic AI risk does not require consciousness, hostility, or an explicit intention to harm humanity. Instead, risk may arise through several distinct but interacting pathways, including autonomous misalignment, harmful human use, organizational failure, and competitive deployment. The severity of these pathways depends on factors such as capability, autonomy, external access, persistence, institutional safeguards, and the preservation of recovery capacity. The analysis is deliberately non-operational: it identifies causal conditions, empirically tractable intermediate quantities, and defensive research questions rather than procedures for causing harm.

中文摘要

摘要：本文提出了一个故障模式框架，用于分析先进人工智能可能如何导致人类灭绝、文明不可逆的崩溃或人类永久性权能丧失。核心论点是，灾难性人工智能风险并不需要意识、敌意或明确的伤害人类意图。相反，风险可能通过几个独立但相互作用的途径产生，包括自主不对齐、有害的人类使用、组织失效以及竞争性部署。这些途径的严重性取决于能力、自治性、外部访问、持续性、制度保障和恢复能力保存等因素。该分析刻意不涉及操作性内容：它识别了因果条件、可经验研究的中间变量以及防御性研究问题，而非造成伤害的程序。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08878

Authors: Mikołaj Sienicki, Krzysztof Sienicki

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08878.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08878

Published: 2026-10-08T02:55:12.891Z

6. Humanize: Judgement Engineering for Agentic Coding

Abstract:Agentic coding makes code generation cheap, but reliable completion remains difficult: the agent that writes the code is a weak judge of whether it is done. We present Humanize, a multi-agent orchestration workflow for agentic coding built around judgement engineering: explicit, mechanically enforced decisions at the boundaries between planning, implementation, review, and learning. A human approves a plan contract, a builder agent implements it in rounds, and a reviewer agent from another vendor decides completion; deterministic hooks, not a model, route work between these roles and enforce 72 mechanical gates. Viewed as a Markov chain over repository states, alternating builder and reviewer samples jointly from two models, so a defect survives only if both miss it. We study Humanize through its deployment, 118 public postmortems of real loops, and its applications. Over 68 versions in 108 days, it gathered 1,468 GitHub stars. Applications include a 567-file gem5 build-system migration under upstream review; Kernel Design Agents, which extend the loop with a kernel knowledge base and profiling feedback and placed in the top three of all three Full-Agent tracks of the MLSys 2026 FlashInfer contest; and, through Humanize Olympiad Agents (HOA), full scores in IOI 2026, IMO 2026, IPhO 2026, and IBO 2024, 418.5/437 in IChO 2026 (gold-medal). Humanize also achieves 672/672 on PutnamBench and ranks first (251/303) on Lean-Eval’s leaderboard even competiting with professional mathematicians. The postmortems show that independent review catches unsupported builder claims, but stopping remains a key weakness. In reports that separate rounds by phase, two thirds of rounds occurred after implementation was accepted. This evidence is observational, not a controlled comparison of workflows.

中文摘要

摘要：自主编码使代码生成成本低廉，但可靠完成仍然困难：编写代码的代理对于代码是否完成判断能力有限。我们提出了 Humanize，一种围绕判断工程构建的多代理自主编码编排工作流：在计划、实施、审查和学习的边界处做出明确的、机械执行的决策。人类批准计划合同，构建代理通过轮次完成实施，而来自其他供应商的审查代理决定是否完成；确定性钩子而非模型在这些角色之间路由工作并执行 72 个机械门。作为对代码库状态的马尔可夫链观察，交替的构建者和审查者共同从两个模型中采样，因此缺陷只有在双方都未发现时才会存活。我们通过部署、118 个真实循环的公共事后分析及其应用研究 Humanize。在 108 天内的 68 个版本中，它获得了 1468 个 GitHub 星标。应用包括在上游审查下的 567 文件 gem5 构建系统迁移；Kernel Design Agents，将循环扩展至内核知识库和性能分析反馈，并在 MLSys 2026 FlashInfer 比赛的三条完整代理赛道中均进入前三名；以及通过 Humanize 奥林匹克代理（HOA）在 IOI 2026、IMO 2026、IPhO 2026 和 IBO 2024 中取得满分，在 IChO 2026 中取得 418.5/437（金牌）。Humanize 还在 PutnamBench 上取得 672/672，并在 Lean-Eval 排行榜上排名第一（251/303），甚至与专业数学家竞争。事后分析显示，独立审查能够发现构建者无依据的声明，但停止机制仍是关键弱点。在按阶段区分轮次的报告中，三分之二的轮次发生在实施被接受之后。该证据为观察性，而非工作流的受控比较。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08900

Authors: Sihao Liu, Ligeng Zhu, Zijian Zhang, Dongyun Zou, Zhengyang Zhang, Changye Li, Song Bian, Song Han, Tony Nowatzki

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08900.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08900

Published: 2026-10-08T02:55:12.891Z

7. Sequential Probabilistic Uncertainty Estimation for Parallel Multi-Agent Reasoning Systems

Abstract:LLM-based multi-agent systems (MAS) have attracted growing attention for improving reasoning through interaction among multiple agents. In this work, we focus on parallel multi-agent reasoning systems, where several agents solve the same problem over multiple rounds and aggregate their outputs into a final answer. Despite their strong reasoning performance, uncertainty estimation for such systems remains underexplored: the reliability of a MAS depends not only on individual generations, but also on how agents interact and evolve across rounds. We propose SAUCE (Sequential Agent Uncertainty through Consensus Evolution), a lightweight, training-free uncertainty estimator that formulates MAS uncertainty as sequential inference over a latent system-level belief. SAUCE aggregates round-level agreement and generation-uncertainty signals through a filtering-style update. Across five backbones, five benchmarks, and two MAS protocols, SAUCE improves misclassification detection, selective prediction, and calibration over a broad set of uncertainty estimation baselines, including standard log-likelihood-based methods and MAS-specific estimators.

中文摘要

摘要：基于大型语言模型（LLM）的多智能体系统（MAS）因通过多智能体间的互动提升推理能力而受到越来越多的关注。在本工作中，我们关注并行多智能体推理系统，其中多个智能体在多轮中解决相同问题，并将它们的输出汇总为最终答案。尽管这些系统的推理表现强大，但其不确定性估计仍未得到充分研究：MAS 的可靠性不仅取决于个体生成结果，还取决于智能体在多轮中的互动和演变方式。我们提出了 SAUCE（通过共识演变的顺序智能体不确定性），这是一种轻量级、无需训练的不确定性估计器，将 MAS 的不确定性表述为对潜在系统级信念的顺序推理。SAUCE 通过过滤式更新汇总轮级别的一致性和生成不确定性信号。在五个骨干模型、五个基准测试以及两种 MAS 协议上，SAUCE 在误分类检测、选择性预测和校准方面相较于包括标准对数似然方法和 MAS 特定估计器在内的广泛不确定性估计基线均有所提升。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08901

Authors: Tunyu Zhang, Zihao Zhao, Yusong Zhao, Haizhou Shi, Zhuohang Li, Haoxian Chen, Hao Wang, Dimitris N. Metaxas

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08901.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08901

Published: 2026-10-08T02:55:12.891Z

8. Agent Plasticity: Measuring Self-Improvement Through Experience

Abstract:AI agents increasingly operate in environments where they can diagnose failures and improve through experience, yet existing evaluations largely measure what an agent can do at a fixed point in time rather than how effectively it learns. Evaluating self-improvement requires answering three questions: does future performance improve and generalize beyond the interactions that enabled learning; how efficiently are new capabilities acquired; and where does the self-improvement process break down? To answer these questions, we study self-improvement in a controlled setting where agents amortize past experience into reusable artifacts that are inherited by future instances. At each checkpoint, we measure performance on training and held-out environment interactions while accounting for learning cost. We introduce agent plasticity, the efficiency with which an agent converts experience into gains in future held-out performance. Across multiple environments, frontier models exhibit sharply different improvement trajectories despite comparable opportunities to learn. Some achieve substantial and persistent gains, while others remain near or below their initial performance, and gains within the training regime often transfer only partially to out-of-distribution conditions. Endpoint capability and acquisition efficiency also diverge: the agent that ultimately performs best need not be the one that improves most efficiently. Tracing failures through the improvement loop further reveals different candidate bottlenecks. Agents with low plasticity often fail to reuse relevant artifacts, whereas more plastic agents may still fail despite reusing relevant artifacts, pointing to limitations in artifact quality, generalization, or application. Evaluating self-improving agents requires measuring not only what they can do, but how effectively they become better through experience.

中文摘要

摘要：人工智能代理越来越多地在能够诊断故障并通过经验改进的环境中运行，然而现有的评估在很大程度上衡量的是代理在固定时间点能做什么，而不是其学习的有效性。评估自我改进需要回答三个问题：未来表现是否提高并超越促成学习的交互范围；新能力的获取效率如何；以及自我改进过程在哪些环节出现问题？为回答这些问题，我们在一个受控环境中研究自我改进，在该环境中，代理将过去的经验摊销到可被未来实例继承的可重用成果中。在每个检查点，我们在训练和保留环境交互上测量性能，同时考虑学习成本。我们引入了代理可塑性，即代理将经验转化为未来保留性能提升的效率。在多个环境中，前沿模型尽管拥有可比的学习机会，其改进轨迹却大相径庭。有些模型取得了显著且持久的提升，而其他模型则停留在接近或低于初始性能水平，并且训练机制内的提升往往只能部分转移到分布外条件下。终端能力和获取效率也存在差异：最终表现最好的代理不一定是提升效率最高的代理。通过改进循环追踪失败进一步揭示了不同的潜在瓶颈。低可塑性的代理通常无法重用相关成果，而高可塑性的代理即便重用相关成果仍可能失败，这表明成果的质量、泛化性或应用存在局限。评估自我改进的代理需要不仅测量它们能做什么，还需评估它们通过经验变得更优秀的有效性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08902

Authors: Harman Singh, Anton Bakhtin, Rulin Shao, Gabriel Synnaeve, Ilia Kulikov, Rob Fergus, Sanjeev Arora, Kurt Keutzer, Jason Weston, Anuj Mahajan, Anirudh Goyal

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08902.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08902

Published: 2026-10-08T02:55:12.891Z

9. AdaGuard: Enhancing Safety and Policy Compliance with Reasoning-Enabled LLM-As-A-Judge Guardrails

Abstract:Enterprise generative AI applications require robust safety mechanisms that can accommodate diverse risk postures, evolving policies, and varying latency constraints. Current guardrail solutions often suffer from rigidity, relying on fixed policy sets and offering limited transparency or reasoning flexibility. We present Adaguard, an adaptive LLM-as-a-Judge framework designed to address these challenges through dynamic policy enforcement and adaptive reasoning-budget allocation. Built using supervised fine-tuning (SFT) and reinforcement learning (GRPO), AdaGuard generalizes to user-defined safety and compliance policies at runtime without requiring frequent model updates. A core innovation of our approach is the ability to dynamically infer the complexity of input-policy pairs, allowing the model to switch between high-speed black-box inference and explainable, reasoning-enabled moderation. This flexibility enables developers to balance stringent latency requirements with the need for actionable transparency. This adaptive capability allows AdaGuard to rival other guardrail and frontier models several times its size, while its auto-reasoning mode recovers the accuracy of always-on reasoning at a fraction of the latency

中文摘要

摘要：企业生成式人工智能应用需要强大的安全机制，以适应多样的风险态势、不断变化的政策以及不同的延迟约束。目前的护栏解决方案往往存在僵化问题，依赖固定的政策集，并且在透明度或推理灵活性方面有限。我们提出了 AdaGuard，一种自适应的“LLM 作为裁判”框架，旨在通过动态政策执行和自适应推理预算分配来应对这些挑战。AdaGuard 使用监督微调（SFT）和强化学习（GRPO）构建，可以在运行时推广到用户定义的安全和合规政策，而无需频繁更新模型。我们方法的核心创新在于能够动态推断输入-政策对的复杂性，使模型能够在高速的黑箱推理和可解释、启用推理的管理之间切换。这种灵活性使开发者能够在严格的延迟要求与可操作透明性的需求之间取得平衡。这种自适应能力使 AdaGuard 能够与其他护栏和前沿模型（其规模为 AdaGuard 数倍）媲美，同时其自动推理模式以极低的延迟恢复始终开启推理的准确性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08923

Authors: Melissa Kazemi Rad, Sihui Dai, Isha Slavin, Kushal Chawla, Mann Patel, Jian Ni, William M. Campbell, Stephen Rawls, Sambit Sahu

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08923.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08923

Published: 2026-10-08T02:55:12.891Z

10. Can AI Agents Make Open-Ended Scientific Discovery? Evidence from Station

Abstract:Recent AI systems have made rapid progress in scientific discovery when given well-defined metrics, but whether they can autonomously undertake open-ended scientific discovery remains unclear. We investigate AI’s ability to tackle open-ended tasks in Station, an open-world environment in which multiple agents simulate a scientific ecosystem. To tackle challenges specific to open-ended tasks, we propose augmenting Station with two mechanisms: a Supervisor mechanism and periodic Meta Reflection, which encourage persistent exploration even when intermediate metrics are lacking. We construct open-ended tasks from three recent oral papers presented at ICLR. We give agents the main research question studied in each paper while withholding the paper’s results and disabling web access. We then measure how many of the original findings-partitioned into individual criteria-agents rediscover. We find that Station rediscovers 62.7% of the criteria on average, compared with 15.4% for Codex Multiagent-v2 and 14.4-20.6% for AI Scientist-v2. Ablation and behavioral analyses indicate that adding the two mechanisms together improves research coverage and continuity. We further evaluate Station on two open-ended tasks without oracle papers and find that some of the discoveries made by the agents closely match discoveries reported by researchers after the knowledge cutoff date. Together, these results indicate that a suitable environment can enable agents to autonomously make meaningful progress in open-ended scientific discovery.

中文摘要

摘要：最近的人工智能系统在科学发现方面取得了快速进展，前提是提供了明确定义的指标，但它们是否能够自主进行开放式科学发现仍不清楚。我们研究了人工智能在Station中的开放式任务处理能力，Station是一个开放世界环境，其中多个代理模拟科学生态系统。为了应对开放式任务的特定挑战，我们提出通过两种机制增强Station：监督机制和周期性元反思，这鼓励在缺乏中间指标时仍保持持续探索。我们从最近在ICLR上发表的三篇口头论文中构建了开放式任务。我们为代理提供每篇论文研究的主要问题，同时隐去论文的结果并禁用网络访问。然后我们测量代理重新发现原始研究发现的情况——将这些发现划分为单独标准。我们发现，Station平均重新发现了62.7%的标准，而Codex Multiagent-v2为15.4%，AI Scientist-v2为14.4-20.6%。消融实验和行为分析表明，同时添加这两种机制可以提高研究覆盖度和连续性。我们进一步在两个没有论文参考的开放式任务上评估了Station，发现代理所取得的一些发现与知识截止日期之后研究人员报告的发现高度契合。综上，这些结果表明，合适的环境可以使代理自主在开放式科学发现中取得有意义的进展。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08927

Authors: Wenyu Du, Stephen Chung

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08927.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08927

Published: 2026-10-08T02:55:12.891Z

Evaluation Domain Papers

1. Adaptive Workflow Intelligence: A Cognitive Architecture for Context-Driven Enterprise Automation

中文摘要

摘要：企业系统越来越依赖自动化工作流程，但许多人工智能驱动的解决方案在非平稳条件、政策变化和操作反馈延迟下仍然脆弱。尽管强化学习和大型语言模型（LLM）代理提供了部分适应能力，但它们本身并不能提供持久的反思机制，也难以与受政策约束的企业操作实现简单整合。本文介绍了自适应工作流程智能（AWI），这是一种面向上下文驱动企业代理的认知架构，围绕感知-认知-行动-反思（PCAR）四层循环进行组织。AWI 将反思视为持续政策优化的机制，并结合混合推理、反思记忆和反馈驱动的适应，以支持在环境漂移和操作约束下的决策。我们在一个具有延迟结果和受控制度变动的模拟企业决策工作流程中评估了 AWI。在漂移与延迟的压力测试中，受护栏约束的自适应方法比静态自动化恢复更快，同时保持政策合规。在此环境下，AWI 的反思组件适度减少了行为振荡和反馈方差，展示了反思型政策适应引入的稳定性与敏捷性权衡。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08793

Authors: Sreedevi Pandiyath Viswambaran

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08793.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08793

Published: 2026-10-08T02:55:37.936Z

2. Accelerating Floating-Point Satisfiability Solving via Gradient Normalization

中文摘要

摘要：可满足性模理论（SMT）求解器是软件验证、程序分析和编译器测试的基础，尤其是在无量词浮点数理论（QF_FP）中。虽然近期基于优化的SMT求解器已经成功地将梯度下降应用于逻辑公式的连续放松，但它们本质上受到梯度主导现象的瓶颈限制，即少量困难子句控制了优化轨迹，阻止求解器满足更广泛的公式并将其困在局部最优解中。为了解决这一问题，我们提出了GradSAT，一个将基于优化的SMT求解与多任务学习（MTL）相结合的新框架。GradSAT通过将每个SMT子句视为独立的MTL任务，重新定义了约束满足过程。通过应用动态梯度归一化（GradNorm），GradSAT在运行时主动平衡所有子句的梯度幅度，有系统地惩罚主导梯度并加速滞后的子句，确保均匀收敛。GradSAT通过高度优化的两阶段混合流水线实现这一点。首先，利用符号编译和操作融合的GPU加速PyTorch后端，将连续放松导航到高质量的最优盆地。其次，将候选赋值交给精确位局部搜索引擎，以快速求解准确、严格的赋值。通过稳定连续搜索动态，GradSAT减轻了之前基于梯度求解器的脆弱性，并为复杂约束求解提供了稳健且高度可并行的架构。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08808

Authors: Yuanzhuo Zhang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08808.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08808

Published: 2026-10-08T02:55:37.936Z

3. Route-Verify-Vote: Procedure-Conditioned Self-Consistency for Mixed-Domain Reasoning

中文摘要

摘要：当语言模型必须以不熟悉的方式结合熟悉的推理操作时，组合性泛化仍然具有挑战性。基于场景的常识推理评估（SCoRE）2026 测试这种能力，涵盖训练中未出现的三个混合领域，并要求模型为每个问题识别完整的正确选项集合。我们引入了 Route-Verify-Vote（RVV）框架，这是一种基于程序的自洽性方法，无需对语言模型进行参数更新。Route 利用提供的领域标签选择推理程序，从而指导模型表示并应用相关约束。Verify 则提示模型根据这些约束评估每个选项。Vote 聚合完整的答案集，并将额外样本分配给两个最频繁答案集之间投票数差距较小的问题。每个问题的样本都遵循相同的领域特定程序。在官方测试集中，对每个问题的16个样本答案集进行投票，实现了74.6%的完全集准确率。自适应 RVV 达到77.3%，而在选定领域路线上的模型组合将准确率提高到79.4%。最终系统在参赛系统中排名第二。这些结果支持领域特定的推理程序和答案集分歧作为在混合领域推理中分配推理时计算资源的有用工具。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08814 (HTTP 429)

Authors: Xinchen Xiao

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08814.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08814

Published: 2026-10-08T02:55:37.936Z

4. An Empirical Study of Agent Skills’ Downstream Utility

中文摘要

摘要：Agent Skills 包提供了过程指导和可重用资源，但相关技能并不一定能提升任务表现。现有研究对技能内容进行了表征，并评估了下游任务的表现，但对其效用如何依赖内容、执行配置及多技能组织的解释有限。我们对 87 个 SkillsBench 任务进行了实证研究，将下游效用定义为在相同模型—执行配置下，相较于无技能情况下的通过率差异。我们在九种配置下比较相同技能，然后在三种选定配置下考察替代的已发布技能及固定技能集的组织方式。我们从由 37,596 个技能构成的精选语料库中检索市场候选项。通过对内容、执行轨迹和最终产物进行 LLM 辅助分析，随后由作者审核，将提供的支持与实际使用和任务结果建立关联。同样的技能在 36.78% 的任务中对某些配置有帮助，而对其他配置则有害，轨迹显示推荐的执行流程可能成为执行负担。相关性排名忽略了更有用的候选项。在评估的候选集内，通过对所需操作的支持重新排序，可在三种配置下将首选通过率提高 4.35–5.80 个百分点。我们总结了 17 条创作实践，将可执行流程与恢复、任务要求保留及最终产物检查联系起来。阶段计划（Stage Plan）和依赖 DAG（Dependency DAG）优于仅按使用顺序排列，其中 DAG 的额外优势集中在任务具有五至六个技能的情况。研究结果指导开发者评估可用操作支持，在保留任务要求的前提下允许流程适配，并在组织技能时明确产物依赖关系。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08875

Authors: Yu Cheng, Dehai Zhao, Zhongxin Liu, Qing Huang, Zhenchang Xing, Xiaoxue Ren

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08875.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08875

Published: 2026-10-08T02:55:37.936Z

5. How Could AI Eliminate Humanity? A Failure-Mode Analysis of Civilizational Risk

中文摘要

摘要：本文提出了一个故障模式框架，用于分析先进人工智能可能如何导致人类灭绝、文明不可逆的崩溃或人类永久性权能丧失。核心论点是，灾难性人工智能风险并不需要意识、敌意或明确的伤害人类意图。相反，风险可能通过几个独立但相互作用的途径产生，包括自主不对齐、有害的人类使用、组织失效以及竞争性部署。这些途径的严重性取决于能力、自治性、外部访问、持续性、制度保障和恢复能力保存等因素。该分析有意不涉及操作性内容：它识别了因果条件、可经验研究的中间变量以及防御性研究问题，而非造成伤害的程序。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08878

Authors: Mikołaj Sienicki, Krzysztof Sienicki

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08878.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08878

Published: 2026-10-08T02:55:37.936Z

6. Humanize: Judgement Engineering for Agentic Coding

中文摘要

摘要：自主编码使代码生成成本低，但可靠完成仍然困难：编写代码的代理对其是否完成的判断能力较弱。我们提出了Humanize，这是一种面向自主编码的多代理协调工作流程，围绕判断工程构建：在规划、实现、审查和学习边界上进行明确的、机械地执行的决策。人类审批计划合同，构建代理分轮实现计划，另一供应商的审查代理决定是否完成；确定性钩子，而非模型，在这些角色之间分配工作并强制执行72个机械关卡。作为对仓库状态的马尔可夫链来看，交替的构建者和审查者样本共同来自两个模型，因此只有当两者都未发现缺陷时，缺陷才会存活。我们通过部署、人类化的118篇公开事后分析以及其应用来研究Humanize。在108天内的68个版本中，它收获了1,468颗GitHub星标。应用包括在上游审查下进行的567个文件的gem5构建系统迁移；内核设计代理（Kernel Design Agents），将循环扩展到内核知识库和性能分析反馈，并在MLSys 2026 FlashInfer竞赛的所有三个全代理赛道中排名前三；通过Humanize奥林匹克代理（HOA），在IOI 2026、IMO 2026、IPhO 2026和IBO 2024中取得满分，IChO 2026获得418.5/437（金牌）。Humanize还在PutnamBench上取得672/672的成绩，并在Lean-Eval排行榜上排名第一（251/303），甚至与专业数学家竞争。事后分析显示，独立审查能发现构建者未经支持的声明，但停止仍是关键弱点。在按阶段区分轮次的报告中，三分之二的轮次发生在实现被接受之后。这一证据是观察性的，而非对工作流程的对照比较。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08900

Authors: Sihao Liu, Ligeng Zhu, Zijian Zhang, Dongyun Zou, Zhengyang Zhang, Changye Li, Song Bian, Song Han, Tony Nowatzki

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08900.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08900

Published: 2026-10-08T02:55:37.936Z

7. Sequential Probabilistic Uncertainty Estimation for Parallel Multi-Agent Reasoning Systems

中文摘要

摘要：基于大型语言模型（LLM）的多智能体系统（MAS）因通过多智能体间的互动提升推理能力而受到越来越多的关注。在本工作中，我们关注并行多智能体推理系统，其中多个智能体在多轮中解决相同问题，并将它们的输出汇总为最终答案。尽管这些系统的推理表现较强，但其不确定性估计仍未得到充分研究：MAS 的可靠性不仅取决于个体生成结果，还取决于智能体在多轮中的互动和演变。我们提出了 SAUCE（通过共识演化的序列化智能体不确定性），这是一种轻量级、无需训练的不确定性估计器，将 MAS 的不确定性表述为对潜在系统级信念的序列推理。SAUCE 通过过滤式更新整合轮级一致性信号和生成不确定性信号。在五个骨干模型、五个基准测试以及两种 MAS 协议上，SAUCE 在误分类检测、选择性预测和校准方面相较于包括标准对数似然方法和 MAS 特定估计器在内的广泛不确定性估计基线均有提升。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08901

Authors: Tunyu Zhang, Zihao Zhao, Yusong Zhao, Haizhou Shi, Zhuohang Li, Haoxian Chen, Hao Wang, Dimitris N. Metaxas

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08901.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08901

Published: 2026-10-08T02:55:37.936Z

8. Agent Plasticity: Measuring Self-Improvement Through Experience

中文摘要

摘要：人工智能代理越来越多地在能够诊断故障并通过经验改进的环境中运行，然而现有的评估在很大程度上衡量的是代理在固定时间点能做什么，而不是其学习的有效性。评估自我改进需要回答三个问题：未来表现是否提高并超越促成学习的交互范围；新能力的获取效率如何；以及自我改进过程在哪些环节出现问题？为回答这些问题，我们在一个受控环境中研究自我改进，在该环境中，代理将过去的经验摊销到可被未来实例继承的可重用成果中。在每个检查点，我们在训练和保留环境交互上测量性能，同时考虑学习成本。我们引入了代理可塑性，即代理将经验转化为未来保留性能提升的效率。在多个环境中，前沿模型尽管拥有可比的学习机会，其改进轨迹却大相径庭。有些模型取得了显著且持久的提升，而其他模型则停留在接近或低于初始性能水平，并且训练机制内的提升往往只能部分转移到分布外条件下。终端能力和获取效率也存在差异：最终表现最好的代理不一定是提升效率最高的代理。通过改进循环追踪失败进一步揭示了不同的潜在瓶颈。低可塑性的代理通常无法重用相关成果，而高可塑性的代理即便重用相关成果仍可能失败，这表明成果质量、泛化或应用存在限制。评估自我改进的代理不仅需要测量其能做什么，还需要评估其通过经验变得更优秀的有效性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08902

Authors: Harman Singh, Anton Bakhtin, Rulin Shao, Gabriel Synnaeve, Ilia Kulikov, Rob Fergus, Sanjeev Arora, Kurt Keutzer, Jason Weston, Anuj Mahajan, Anirudh Goyal

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08902.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08902

Published: 2026-10-08T02:55:37.936Z

9. AdaGuard: Enhancing Safety and Policy Compliance with Reasoning-Enabled LLM-As-A-Judge Guardrails

中文摘要

摘要：企业生成式人工智能应用需要强大的安全机制，以适应多样的风险态势、不断变化的政策以及不同的延迟约束。目前的护栏解决方案往往存在僵化问题，依赖固定的政策集，并且在透明度或推理灵活性方面有限。我们提出了 AdaGuard，一种自适应的“LLM 作为裁判”框架，旨在通过动态政策执行和自适应推理预算分配来应对这些挑战。AdaGuard 使用监督微调（SFT）和强化学习（GRPO）构建，可以在运行时推广到用户定义的安全和合规政策，而无需频繁更新模型。我们方法的核心创新在于能够动态推断输入-政策对的复杂性，使模型能够在高速的黑箱推理和可解释、启用推理的审核之间切换。这种灵活性使开发者能够在严格的延迟要求与可操作透明性的需求之间取得平衡。这种自适应能力使 AdaGuard 能够与其他护栏和前沿模型（其规模为 AdaGuard 的数倍）相媲美，同时其自动推理模式以极低的延迟恢复始终开启推理的准确性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08923

Authors: Melissa Kazemi Rad, Sihui Dai, Isha Slavin, Kushal Chawla, Mann Patel, Jian Ni, William M. Campbell, Stephen Rawls, Sambit Sahu

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08923.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08923

Published: 2026-10-08T02:55:37.936Z

10. Can AI Agents Make Open-Ended Scientific Discovery? Evidence from Station

中文摘要

摘要：最近的人工智能系统在科学发现方面取得了快速进展，前提是提供了明确定义的指标，但它们是否能够自主进行开放式科学发现仍不清楚。我们研究了人工智能在Station中处理开放式任务的能力，Station是一个开放世界环境，其中多个智能体模拟科学生态系统。为了应对开放式任务的特定挑战，我们提出通过两种机制增强Station：监督机制和周期性元反思，这些机制鼓励即使在缺少中间指标时也进行持续探索。我们从最近在ICLR上发表的三篇口头论文构建了开放式任务。我们向智能体提供每篇论文研究的主要问题，同时隐藏论文的结果并禁用网络访问。然后，我们测量智能体重新发现了原始发现中各个标准的数量。结果显示，Station平均重新发现了62.7%的标准，而Codex Multiagent-v2为15.4%，AI Scientist-v2为14.4%-20.6%。消融和行为分析表明，同时加入这两种机制可以提高研究覆盖率和连续性。我们进一步在两个没有标准论文的开放式任务中评估Station，发现智能体做出的一些发现与知识截止日期后研究人员报道的发现高度一致。综合来看，这些结果表明，合适的环境可以使智能体自主在开放式科学发现中取得有意义的进展。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08927

Authors: Wenyu Du, Stephen Chung

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.08927.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08927

Published: 2026-10-08T02:55:37.936Z

VLM Domain Papers

1. Pre-training, Reasoning, Benchmarking: X-ray Report Generation on CheXpert Plus Dataset

Abstract:X-ray image-based Radiology Report Generation (RRG) constitutes a critical research direction within medical artificial intelligence, with great potential to alleviate clinicians’ diagnostic workload and shorten patient waiting periods. Despite substantial advances over recent years, the field faces evident bottlenecks stemming from insufficient standardized benchmarks and inadequate domain adaptation of generic large models. Notably, the newly released CheXpert Plus dataset is provided without accompanying baseline implementations and evaluation results, which impedes standardized training, quantitative evaluation and fair comparison among follow-up algorithms. To mitigate this limitation, we establish a comprehensive benchmark encompassing prevailing X-ray report generation models and Large Language Models on CheXpert Plus. This benchmark delivers a reliable comparative foundation for upcoming methods and enables researchers to rapidly identify state-of-the-art approaches within this domain. Beyond benchmark construction, we rethink X-ray RRG under the paradigm of large models and propose a novel framework termed MambaXray-PRB. Our framework improves report generation performance and enhances model interpretability via multi-stage large-model pre-training and multi-modal Chain-of-Thought reasoning. The pipeline consists of three successive phases: self-supervised auto-regressive modeling, X-ray-report contrastive learning, and post-training optimization for reasoning and report generation. Extensive experiments on IU X-ray, MIMIC-CXR, and CheXpert Plus datasets validate the effectiveness of MambaXray-PRB for radiology report generation. The source code of this paper is available on this https URL

中文摘要

摘要：基于X光影像的放射学报告生成（RRG）是医疗人工智能中的一个关键研究方向，具有极大的潜力来减轻临床医生的诊断工作负担并缩短患者等待时间。尽管近年来取得了显著进展，该领域仍面临明显瓶颈，主要来自于标准化基准不足以及通用大型模型的领域适应性不足。值得注意的是，新发布的CheXpert Plus数据集没有提供相应的基线实现和评估结果，这阻碍了标准化训练、定量评估以及后续算法的公平比较。为缓解这一限制，我们建立了一个涵盖主流X光报告生成模型和大型语言模型在CheXpert Plus上的综合基准。该基准为后续方法提供了可靠的比较基础，使研究人员能够快速识别该领域的最先进方法。除了基准构建之外，我们在大型模型范式下重新思考X光RRG，并提出了一种称为MambaXray-PRB的新框架。我们的框架通过多阶段大型模型预训练和多模态思维链推理，提升了报告生成性能并增强了模型可解释性。该流程包括三个连续阶段：自监督自回归建模、X光报告对比学习以及用于推理和报告生成的后训练优化。在IU X-ray、MIMIC-CXR和CheXpert Plus数据集上的大量实验验证了MambaXray-PRB在放射学报告生成方面的有效性。本文的源代码可通过此https URL获取。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08813

Authors: Xiao Wang, Yuxiang Zhang, Dan Xu, Yuehang Li, Shiao Wang, Bo Jiang, Yaowei Wang, Yonghong Tian, Jin Tang

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.08813.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08813

Published: 2026-10-08T02:56:03.076Z

2. Autonomous Driving Research Requires a Community-Driven Data Paradigm

Abstract:Autonomous driving has made remarkable progress, with recent AI advances enabling commercial deployments that are reshaping urban mobility. Yet the field remains far from its universal social promise: autonomous systems that can operate robustly anywhere, anytime, for anyone. We posit that this gap is not merely a modeling problem, but a problem of the prevailing data paradigm. Current research relies heavily on a few benchmark datasets with limited spatial and scenario coverage, even though the community has collectively produced over 600 autonomous driving datasets across nearly 50 countries. However, this abundance has not translated into broad research impact: most datasets remain significantly underused due to fragmentation, limited visibility, incompatible protocols, and benchmark incentives that concentrate attention on a few dominant datasets. We therefore argue that autonomous driving research requires a collaborative, community-driven data paradigm. Such a paradigm would improve the discovery, reuse, integration, and evaluation of diverse datasets; make underexplored data easier and more rewarding to study; and lower the barrier for new contributors. We outline its key principles, illustrate an early realization, and call for collaboration across academia and industry to transform fragmented datasets into shared community infrastructure for anytime-anywhere autonomy.

中文摘要

摘要：自动驾驶取得了显著进展，近期的人工智能发展使商业部署成为可能，并正在重塑城市交通。然而，这一领域距离其普遍的社会承诺还有很大差距：即能够在任何地点、任何时间、为任何人稳健运行的自主系统。我们认为，这一差距不仅仅是建模问题，更是现行数据模式的问题。当前研究在很大程度上依赖少数具有有限空间和场景覆盖的基准数据集，尽管整个社区已经在近50个国家共创建了超过600个自动驾驶数据集。然而，这种丰富的数据资源并没有转化为广泛的研究影响：由于数据碎片化、可见性有限、协议不兼容以及基准激励集中在少数主导数据集上，大多数数据集仍然被严重低估。因此，我们认为自动驾驶研究需要一种协作的、社区驱动的数据模式。这样的模式将改善多样化数据集的发现、再利用、整合和评估；使开发较少的数据更容易且更有价值；并降低新贡献者的参与门槛。我们概述了其关键原则，展示了早期实现，并呼吁学术界和产业界的合作，将分散的数据集转化为任何时间、任何地点自治的共享社区基础设施。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08825

Authors: Jinsu Yoo, Zanming Huang, Katie Z Luo, Zheda Mai, Qiyuan Wu, Bharath Hariharan, Mark Campbell, Wei-Lun Chao

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.08825.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08825

Published: 2026-10-08T02:56:03.076Z

3. PanoPed: Beyond Bounding Boxes for Sim-to-Real Panoramic Pedestrian Tracking

Abstract:Full-sphere panoramic cameras let fixed monitoring systems and mobile robots track people in every direction, but a planar bounding box does not fully describe where a person is on the sphere. We introduce PanoPed, a sim-to-real benchmark for pedestrian tracking on the full sphere. PanoPed-S contains 108,000 frames from fixed, quadruped-mounted, and drone-mounted cameras, with synchronized masks, depth, camera poses, and 3D pedestrian states. PanoPed-R adds 28,002 real frames from fixed cameras, 16,247 of them densely annotated. We find that an ERP rectangle cannot uniquely determine the spherical center and angular extent of the visible person, while the detector’s visual query still carries information about them. Inspired by the sextant’s use of angular measurements to locate objects, we propose Sextant, a plug-and-play angular localization head with only about 0.035M parameters. It reuses a frozen detector, keeps track identities unchanged, and needs no extra image encoder. Sextant gives the best result in our PanoPed-S test comparison, raising the strongest baseline, MOTIP, from 47.30 to 49.49 HOTA, with gains on all eight test sequences. Without fine-tuning on real data, the same synthetic-trained heads improve MOTIP and HAT by 0.96-1.14 HOTA on real video, and both seeds improve every real sequence. HAT+Sextant scores best among the compared systems that add no localization image encoder.

中文摘要

摘要：全景全向摄像头使固定监控系统和移动机器人能够从各个方向跟踪人员，但平面边界框不能完全描述人所在的球面位置。我们提出了 PanoPed，这是一个用于全球面行人跟踪的模拟到真实的基准数据集。PanoPed-S 包含来自固定、四足机器人和无人机摄像头的 108,000 帧图像，并提供同步的掩码、深度、相机位姿以及三维行人状态。PanoPed-R 添加了来自固定摄像头的 28,002 帧真实图像，其中 16,247 帧进行了密集标注。我们发现，ERP 矩形无法唯一确定可见人员的球面中心和角度范围，而检测器的视觉查询仍然携带有关它们的信息。受六分仪使用角度测量来定位物体的启发，我们提出了 Sextant，这是一种即插即用的角度定位头，仅约有 0.035M 参数。它复用了一种冻结的检测器，保持跟踪身份不变，并且不需要额外的图像编码器。Sextant 在我们的 PanoPed-S 测试比较中给出了最佳结果，将最强基线 MOTIP 的 HOTA 从 47.30 提升到 49.49，在所有八个测试序列中都有提升。在未对真实数据进行微调的情况下，相同的合成训练头在真实视频上将 MOTIP 和 HAT 提升了 0.96-1.14 HOTA，并且两个种子都提升了每个真实序列的表现。HAT+Sextant 在所有未添加定位图像编码器的对比系统中得分最高。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08826

Authors: Qinfeng Zhu, Weiguang Zhao, Yunxi Jiang, Anh Nguyen, Lei Fan

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.08826.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08826

Published: 2026-10-08T02:56:03.076Z

4. MoR-MLLM: Mixture of Recursions for Efficient Multimodal Large Language Models

Abstract:Multimodal Large Language Models (MLLMs) have demonstrated remarkable reasoning capabilities across vision and language tasks. However, their massive computational and memory demands hinder real-world deployment. While recent efforts reduce costs by employing lightweight language backbones, existing paradigms remain computation-dense due to their static sparsity and depth allocation, which cannot adapt to the semantic complexity of each token. To this end, we propose MoR-MLLM, a computation-sparse MLLM based on the recent Mixture-of-Recursions (MoR) framework. MoR-MLLM introduces adaptive per-token recursion, allowing the model to dynamically adjust its recursive depth and allocate more computation to visually or linguistically challenging tokens while skipping redundant operations for simpler ones. To stabilize the training of recursive sparsity in multimodal settings, we further design a three-stage MoR-Tuning strategy and an entropy-regularized loss to encourage diverse routing distributions. Extensive experiments show that compared with recent advanced tiny MLLMs, our proposed MoR-MLLM can greatly reduce the training memory and computation complexity while retaining high performance on various vision-language tasks.

中文摘要

摘要：多模态大语言模型（MLLMs）在视觉和语言任务中展现了卓越的推理能力。然而，它们庞大的计算和内存需求阻碍了实际应用。尽管最近的研究通过使用轻量级语言骨干网络来降低成本，但现有范式仍然计算密集，因为它们的稀疏性和深度分配是静态的，无法适应每个标记的语义复杂性。为此，我们提出了MoR-MLLM，一种基于最新递归混合（MoR）框架的计算稀疏MLLM。MoR-MLLM引入了自适应的每标记递归机制，使模型能够动态调整递归深度，并为视觉或语言上具有挑战性的标记分配更多计算资源，同时跳过对简单标记的冗余操作。为了在多模态环境中稳定递归稀疏性训练，我们进一步设计了三阶段MoR调优策略和熵正则化损失，以鼓励多样化的路由分布。大量实验表明，与最近先进的小型MLLM相比，我们提出的MoR-MLLM在保持各种视觉-语言任务高性能的同时，能够大幅降低训练内存和计算复杂度。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08830

Authors: Pengcheng Zheng, Chaoning Zhang, Jiaxin Yan, Sihan Cao, Jianwei Zhang, Xudong Wang, Jiaquan Zhang, Jewon Lee, Tae-Ho Kim, Yang Yang, Heng Tao Shen

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.08830.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08830

Published: 2026-10-08T02:56:03.076Z

5. VCR-Bench: A Modular Open-Source Benchmark for Video Classification Robustness

Abstract:Robustness of image classification has several benchmarks, but their video counterparts are absent. In video classification temporal dimension introduces additional degrees of freedom for adversarial attacks, defenses, and preprocessing. Temporal sampling, perturbation budgets, and metric aggregation also interact in ways with no direct analogue in the image setting. Therefore, robustness for video classifiers is studied across scattered, incompatible implementations, making reported numbers hard to reproduce and analyze. We introduce VCR-Bench, a modular open-source benchmark framework that standardizes video loading, wrappers for classifiers, adversarial attacks and defenses, perceptual metrics, configuration presets, and result logging. VCR-Bench currently integrates 30 video classification models, 14 adversarial attacks, and 10 defense wrappers under a common evaluation protocol. We evaluate representative video classifiers, attacks, and defenses on Kinetics-400 subset, reporting clean accuracy, attack success rate, perceptual quality, runtime, and memory usage. VCR-Bench is released with documented installation, reproducible run presets, component-extension interfaces, and scripts for reproducing the reported results at this https URL.

中文摘要

摘要：图像分类的鲁棒性有若干基准，但其视频对应的基准尚不存在。在视频分类中，时间维度为对抗性攻击、防御和预处理引入了额外的自由度。时间采样、扰动预算和度量聚合也以图像设置中没有直接对应的方法互动。因此，视频分类器的鲁棒性研究分散在不兼容的实现中，使得报告的数据难以复现和分析。我们引入了 VCR-Bench，一个模块化的开源基准框架，标准化了视频加载、分类器包装器、对抗性攻击和防御、感知度量、配置预设以及结果记录。VCR-Bench 当前整合了 30 个视频分类模型、14 种对抗性攻击和 10 个防御包装器，并基于统一的评估协议进行评估。我们在 Kinetics-400 子集上评估了具有代表性的视频分类器、攻击和防御方法，报告了干净准确率、攻击成功率、感知质量、运行时间和内存使用情况。VCR-Bench 发布时附有文档化的安装说明、可复现的运行预设、组件扩展接口以及用于复现报告结果的脚本，该脚本可通过此 https URL 获取。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08936

Authors: Maksim Plinskiy, Aleksandr Gushchin, Sergey Lavrushkin, Dmitriy S. Vatolin, Anastasia Antsiferova

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.08936.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08936

Published: 2026-10-08T02:56:03.076Z

Abstract:Language-guided panoramic video generation benefits various downstream applications, such as interactive 3D scene exploration, virtual reality experiences, and embodied agent training. Existing panoramic generators follow predefined trajectories, and interactive world models act through low-level actions in perspective views. We propose SPW-Nav, a streaming panoramic world model that understands movement instructions and streams one minute of 2K 360-degree video in real time from a single panorama. SPW-Nav interprets each instruction in the previously generated panorama as camera motion. Spherical rotation decoupling applies rotation exactly on the sphere, pose-aligned conditioning keeps translation inputs bounded over long streams, and a multi-term memory with a few-step generator continues the scene as instructions change. We also build SPW-NavSet, panoramic videos with camera trajectories and verified instructions. Driven by language, SPW-Nav outperforms prior panoramic generators in camera-following accuracy and video quality, and supports on-the-fly instruction switching.

中文摘要

摘要：语言引导的全景视频生成对各种下游应用都有益，例如交互式3D场景探索、虚拟现实体验和具身智能体训练。现有的全景生成器遵循预定义的轨迹，而交互式世界模型则通过透视视图中的低级动作进行操作。我们提出了SPW-Nav，一种流式全景世界模型，它能够理解移动指令，并从单个全景图实时生成一分钟的2K 360度视频。SPW-Nav将每条指令在先前生成的全景中解释为相机运动。球面旋转解耦在球面上精确地应用旋转，姿态对齐条件保持长时间流中平移输入的界限，多项记忆结合少步生成器能够随着指令变化持续生成场景。我们还构建了SPW-NavSet，包括带有相机轨迹和经过验证指令的全景视频。在语言驱动下，SPW-Nav在相机跟随精度和视频质量上均优于先前的全景生成器，并支持即时切换指令。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08941

Authors: Yunheng Liu, Ziqi Cai, Siqi Yang, Yimu Wang, Minggui Teng, Jiaming Tan, Shuchen Weng, Erwin Wu, Kaipeng Zhang, Boxin Shi

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.08941.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08941

Published: 2026-10-08T02:56:03.076Z

7. One-Slide Calibration of Pathology Foundation Models

Abstract:Scanner variation changes how pathology foundation models represent the same tissue. We introduce SlideRuler, which uses regions within a slide as internal controls to estimate and correct acquisition-induced shifts in other regions. A transfer map learned from paired rescans enables calibration from a single scan at inference while keeping the foundation model fixed. Across two encoders and five SCORPION scanners, learned transfer reduces mean target-to-source embedding distance by 16.3-38.5% relative to raw embeddings. Comparisons with unrelated same-scanner controls reveal a positive same-slide contribution across all four evaluation settings, including scanner holdout. A source-anchored variant reduces source-feature displacement by 47.7-83.6% relative to learned transfer while retaining most of its alignment gain. By drawing calibration information from the slide itself, SlideRuler offers a path toward more consistent use of frozen pathology models across imaging systems.

中文摘要

摘要：扫描仪的差异会改变病理基础模型对相同组织的表示方式。我们引入了 SlideRuler，它使用切片内的区域作为内部对照，以估计并校正其他区域的获取引起的偏移。从成对的重新扫描中学习的转移映射使在推理时可以仅通过一次扫描进行校准，同时保持基础模型固定。在两个编码器和五台 SCORPION 扫描仪上，学习的转移使目标到源嵌入的平均距离相比原始嵌入减少了 16.3%-38.5%。与无关的同扫描仪对照比较显示，在包括扫描仪保留在内的所有四个评估设置中，同切片对齐贡献为正。源锚定的变体相对于学习的转移将源特征偏移减少了 47.7%-83.6%，同时保留了大部分的对齐增益。通过从切片本身获取校准信息，SlideRuler 为在不同成像系统上更一致地使用冻结的病理模型提供了一条路径。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08944

Authors: Ming Ren Hou, Tianyi Huang

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.08944.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08944

Published: 2026-10-08T02:56:03.076Z

8. RACER: Reflective Agent Coupling Query Interpretation and Tool-Based Retrieval for Frame Selection in Long Video Understanding

Abstract:Video large language models (Vid-LLMs) excel at diverse video-language tasks by reasoning over selected frames. However, frame selection for long videos remains challenging, as it requires retrieving relevant frames distributed across segments from a large candidate pool given complex queries. This paper investigates dominant approaches to long-video frame selection from a task-decomposition perspective, identifying two key challenges: the Query Comprehension Gap in similarity-based methods and the Interpretation—Selection Gap in judgment-based methods. To address them, we propose RACER, a training-free reflective agentic framework that decomposes long-video frame selection into query interpretation driven by a lightweight Vid-LLM and evidence localization supported by an embedding model serving as a retrieval tool. Specifically, the Vid-LLM is responsible solely for reformulating the complex query into sub-queries that make implicit information requirements explicit, mitigating the Query Comprehension Gap. Meanwhile, the retrieval tool leverages these sub-queries to localize relevant evidence, relieving the Vid-LLM of direct frame selection and thus addressing the Interpretation—Selection Gap. Finally, the retrieved frames are fed back to the Vid-LLM for sub-query refinement, forming a reflection loop that iteratively improves query interpretation and frame selection. Experiments across multiple benchmarks show that RACER consistently improves long video understanding. Notably, RACER achieves effective frame selection even with limited-capability components, demonstrating that agentic integration enables these components to enhance more capable Vid-LLMs.

中文摘要

摘要：视频大语言模型（Vid-LLMs）通过对选定帧进行推理，在多种视频-语言任务中表现出色。然而，对于长视频的帧选择仍然具有挑战性，因为它需要从大量候选帧中检索分布在各段落中的相关帧，以应对复杂查询。本文从任务分解的视角研究了长视频帧选择的主流方法，识别了两个关键挑战：基于相似度的方法中的查询理解差距（Query Comprehension Gap）以及基于判断的方法中的解读-选择差距（Interpretation—Selection Gap）。为了解决这些问题，我们提出了RACER，一种无需训练的反思型代理框架，它将长视频帧选择分解为由轻量级Vid-LLM驱动的查询解读以及由嵌入模型支持的证据定位检索工具。具体而言，Vid-LLM仅负责将复杂查询重新表述为子查询，从而将隐含信息需求显式化，缓解查询理解差距。与此同时，检索工具利用这些子查询来定位相关证据，使Vid-LLM无需直接进行帧选择，从而解决解读-选择差距。最终，检索到的帧被反馈给Vid-LLM以优化子查询，形成一个反思循环，迭代提升查询解读和帧选择能力。在多个基准测试中的实验表明，RACER能够持续提升长视频理解能力。值得注意的是，即使使用能力有限的组件，RACER仍能实现有效的帧选择，这表明代理式整合能够使这些组件增强更强大的Vid-LLMs的性能。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08954

Authors: Yiyang Huang, Yitian Zhang, Yizhou Wang, Jianglin Lu, Qihua Dong, Hailing Wang, Huimin Zeng, Mingyuan Zhang, Yun Fu

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.08954.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08954

Published: 2026-10-08T02:56:03.076Z

9. S2Tok: Streaming 3D Gaussian Reconstruction with Persistent Spatial Tokens

Abstract:Streaming 3D reconstruction requires more than a sequence of geometric predictions: it requires a persistent scene state that can incorporate new evidence and remain renderable as observations arrive. Latent spatial tokens offer a promising representation for this purpose, but constructing them from an image collection leaves open how to maintain them online, where each observation may both revisit known regions and reveal new content. We introduce S2Tok, a feed-forward framework that maintains a size-adaptive, persistent scene state from uncalibrated image streams. Its central idea is to distinguish updates to the existing representation from selective expansion. A spatially informed transformer integrates each incoming observation with the persistent scene tokens, while a learned admission module selectively expands the representation to limit redundant storage. A hierarchical decoder and Gaussian head convert the evolving state into non-pixel-aligned 3D Gaussians, enabling novel-view rendering without caching previous frames. Experiments across four benchmarks demonstrate competitive streaming rendering quality with compact Gaussian representations. These results support latent spatial tokens as a persistent computational state for online 3D reconstruction, combining learned scene updates with explicit Gaussian rendering.

中文摘要

摘要：流式 3D 重建不仅仅需要一系列几何预测：它需要一个持续的场景状态，能够在新证据到来时进行更新并保持可渲染性。潜在的空间令牌为这一目的提供了有前景的表示，但从图像集合构建它们时，仍然存在如何在线维护的问题，因为每个观测既可能重新访问已知区域，又可能揭示新内容。我们提出了 S2Tok，这是一种前馈框架，从未校准的图像流中维护一个大小可自适应的持续场景状态。其核心思想是区分对现有表示的更新与选择性扩展。一个空间信息化的变换器将每个进入的观测与持续的场景令牌整合，而一个学习的准入模块选择性地扩展表示，以限制冗余存储。分层解码器和高斯头将不断演化的状态转换为非像素对齐的 3D 高斯，从而在无需缓存先前帧的情况下实现新视角渲染。在四个基准测试中的实验表明，紧凑的高斯表示能够实现具有竞争力的流式渲染质量。这些结果支持潜在空间令牌作为在线 3D 重建的持续计算状态，结合了学习到的场景更新与显式高斯渲染。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08978

Authors: Fang Li, Jiraphon Yenphraphai, Quentin Herau, Depu Meng, Yihan Hu, Tianshuo Xu, Narendra Ahuja, Wei Zhan

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.08978.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08978

Published: 2026-10-08T02:56:03.076Z

10. Zero-Shot Brain MRI Inpainting with 2.5D Unconditional Flow Priors

Abstract:Generative inpainting of brain MRI volumes is essential for synthesizing healthy tissue in pathological regions, improving the accuracy and reliability of automated downstream brain analysis applications such as image registration, brain extraction, and segmentation. However, standard 3D approaches are computationally prohibitive, while efficient 2D slice-wise methods suffer from severe inter-slice discontinuities. Furthermore, traditional models rely on conditional training, requiring task-specific learning of masked inputs. We propose a zero-shot brain MRI inpainting framework utilizing 2.5D unconditional flow priors to capture spatial context along the superior-inferior axis without the overhead of full 3D convolutions. During training, our flow matching model learns the joint distribution of adjacent axial slice triplets, modeling the manifold of healthy brain anatomy while explicitly excluding pathological regions from the loss function. At inference, the model processes the input triplets autoregressively along the depth axis. We employ the Restora-Flow solver to constrain the unconditional prior using the input mask, achieving accurate zero-shot inpainting. Evaluations show our 2.5D strategy resolves the structural discontinuities of 2D baselines, synthesizing plausible healthy tissue while maintaining volumetric consistency across the axial, sagittal, and coronal planes. As a final step, we generate and average an ensemble of multiple stochastic reconstructions to form the final prediction. Quantitative results benchmarked on the official BraTS 2026 Inpainting Challenge validation set demonstrate the effectiveness of our proposed approach, yielding an SSIM of 0.816 $\pm$ 0.112, MSE of 0.007 $\pm$ 0.005, and PSNR of 22.923 $\pm$ 4.343. Code is available at this https URL.

中文摘要

摘要：脑部MRI体积的生成性图像修复对于在病理区域合成健康组织至关重要，从而提高自动化下游脑分析应用（如图像配准、脑提取和分割）的准确性和可靠性。然而，标准的3D方法在计算上成本过高，而高效的二维切片方法则存在严重的切片间不连续性。此外，传统模型依赖条件训练，需要对掩码输入进行特定任务的学习。我们提出了一种零样本脑MRI修复框架，利用2.5D无条件流先验沿上下轴捕捉空间上下文，而无需完全3D卷积的开销。在训练过程中，我们的流匹配模型学习相邻轴向切片三联体的联合分布，建模健康脑解剖结构的流形，同时在损失函数中明确排除病理区域。在推理阶段，模型沿深度轴自回归处理输入三联体。我们采用Restora-Flow求解器使用输入掩码约束无条件先验，实现准确的零样本修复。评估显示，我们的2.5D策略解决了二维基线方法的结构不连续性，合成出合理的健康组织，同时在轴面、矢状面和冠状面上保持体积一致性。作为最后一步，我们生成并平均多个随机重建的集合以形成最终预测。在官方BraTS 2026修复挑战验证集上的定量结果表明，我们提出的方法有效，获得了0.816 ± 0.112的SSIM、0.007 ± 0.005的MSE，以及22.923 ± 4.343的PSNR。代码可在此https URL获取。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.08983

Authors: Arnela Hadzic, Franz Thaler, Simon Johannes Joham, Martin Urschler

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.08983.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.08983

Published: 2026-10-08T02:56:03.076Z

ArXiv Domain 2026-10-08

LLM Domain Papers

1. Tokka-Bench: Evaluating Tokenizers Across 100 Natural and 20 Programming Languages

2. Child ASR Adaptation with Adult Retention: An Empirical Study

3. When Forgetting Looks Like Improvement: Metric Masking in Streaming Diarizer Adaptation and the Price of Rehearsal

4. Emo-Jev: Probabilistic Reasoning for Emotion Classification with Jev

5. CoDR: Training-Free Confidence-Drift Remasking for Diffusion Language Models

6. Leveraging LLM-Generated Explanations for Detecting Emotionally Rewritten Fake News

7. Beyond the Sycophancy Score: How Task, Model, and Pressure Shape LLM Yielding

8. Beyond Risk Prediction: Evidence Grounding and Psychosocial Factor Verification for Explainable Suicide Risk Assessment

9. QuanLing: Cross-Branch Validation of Language Distance Quantification on Western Romance

10. LRCC: Generalizing Low-Rank Compression with Conditional Computation

Agent Domain Papers

1. Adaptive Workflow Intelligence: A Cognitive Architecture for Context-Driven Enterprise Automation

2. Accelerating Floating-Point Satisfiability Solving via Gradient Normalization

3. Route-Verify-Vote: Procedure-Conditioned Self-Consistency for Mixed-Domain Reasoning

4. An Empirical Study of Agent Skills’ Downstream Utility

5. How Could AI Eliminate Humanity? A Failure-Mode Analysis of Civilizational Risk

6. Humanize: Judgement Engineering for Agentic Coding

7. Sequential Probabilistic Uncertainty Estimation for Parallel Multi-Agent Reasoning Systems

8. Agent Plasticity: Measuring Self-Improvement Through Experience

9. AdaGuard: Enhancing Safety and Policy Compliance with Reasoning-Enabled LLM-As-A-Judge Guardrails

10. Can AI Agents Make Open-Ended Scientific Discovery? Evidence from Station

Evaluation Domain Papers

1. Adaptive Workflow Intelligence: A Cognitive Architecture for Context-Driven Enterprise Automation

2. Accelerating Floating-Point Satisfiability Solving via Gradient Normalization

3. Route-Verify-Vote: Procedure-Conditioned Self-Consistency for Mixed-Domain Reasoning

4. An Empirical Study of Agent Skills’ Downstream Utility

5. How Could AI Eliminate Humanity? A Failure-Mode Analysis of Civilizational Risk

6. Humanize: Judgement Engineering for Agentic Coding

7. Sequential Probabilistic Uncertainty Estimation for Parallel Multi-Agent Reasoning Systems

8. Agent Plasticity: Measuring Self-Improvement Through Experience

9. AdaGuard: Enhancing Safety and Policy Compliance with Reasoning-Enabled LLM-As-A-Judge Guardrails

10. Can AI Agents Make Open-Ended Scientific Discovery? Evidence from Station

VLM Domain Papers

1. Pre-training, Reasoning, Benchmarking: X-ray Report Generation on CheXpert Plus Dataset

2. Autonomous Driving Research Requires a Community-Driven Data Paradigm

3. PanoPed: Beyond Bounding Boxes for Sim-to-Real Panoramic Pedestrian Tracking

4. MoR-MLLM: Mixture of Recursions for Efficient Multimodal Large Language Models

5. VCR-Bench: A Modular Open-Source Benchmark for Video Classification Robustness

6. SPW-Nav: A Streaming Panoramic World Model for Language-Guided Navigation

7. One-Slide Calibration of Pathology Foundation Models

8. RACER: Reflective Agent Coupling Query Interpretation and Tool-Based Retrieval for Frame Selection in Long Video Understanding

9. S2Tok: Streaming 3D Gaussian Reconstruction with Persistent Spatial Tokens

10. Zero-Shot Brain MRI Inpainting with 2.5D Unconditional Flow Priors