ArXiv Domain 2026-10-31

数据来源：ArXiv Domain

LLM Domain Papers

1. The FlightFinder Occurrence Corpus: harmonising 128 aviation-safety source codes into one deduplicated cross-national accident dataset

Abstract:National aviation-investigation databases are deep but territorial; global compilations such as the Aviation Safety Network and the Bureau of Aircraft Accidents Archives are hand-curated, apply their own inclusion thresholds, and publish no per-record lineage. The FlightFinder Occurrence Corpus harmonizes 279,485 occurrences and 131,882 attached narratives from 128 source codes — national accident-investigation authorities, occurrence-reporting systems and four global compilations. Records are mapped into one schema; each normalized record reduces to exactly one key from an ordered ladder — event signature, registration and date, rounded position and date, or source and native identifier — so a missing field never creates a link, and the key is the occurrence identifier. Records the ladder cannot reach are scored in a blocked pass into an explicit merge table and applied only within an automatically approved score band or after a person confirms them. Field conflicts follow a documented source ranking; one provenance row per contributing source — its native identifier and URL — is retained, with attribution and licence terms per source code. Linkage error is measured on five adjudicated samples totalling 200 pairs. A recorded uniform random draw of 60 from the 3,441 merges the automatic band applies as it ships found no error; the exact two-sided 95% Clopper-Pearson upper bound on its false-merge rate is 6.0%, or 205 pairs. The measured error covers the scored band only; the ladder’s deterministic joins are not adjudicated. The corpus has no exposure denominators and inherits the under-reporting of its sources. It is offered as a base for cross-national safety statistics, source-disagreement studies and narrative text mining; a frozen, DOI-referenced snapshot accompanies the live API.

中文摘要

摘要：国家航空调查数据库内容丰富但具有地域性；像航空安全网络（Aviation Safety Network）和飞机事故档案局（Bureau of Aircraft Accidents Archives）这样的全球汇编是手工整理的，采用自己的纳入标准，并且不发布每条记录的来源。FlightFinder 事件语料库整合了来自128个来源代码的279,485条事件和131,882条附属叙述，包括国家事故调查机关、事件报告系统和四个全球汇编。记录被映射到一个统一模式；每条规范化记录精确归约为有序梯阶中的一个关键项——事件特征、登记号和日期、经四舍五入的位置和日期，或来源及原生标识符——因此缺失的字段永远不会创建链接，并且关键项是事件标识符。梯阶无法触及的记录将在明确合并表中得分封锁，仅在自动批准的得分区间内或经过人工确认后使用。字段冲突遵循记录在案的来源排序；每个贡献来源保留一行来源信息——其原生标识符和网址——并按来源代码附注明确归属和许可条款。链路错误在五个裁定样本（共计200对）中进行衡量。从3,441次合并中随机抽取60次，自动区间应用未发现错误；其错误合并率的精确双侧95% Clopper-Pearson 上限为6.0%，即205对。测量的错误仅涵盖评分区间；梯阶的确定性连接未经过裁定。语料库没有暴露分母，并继承了其来源的漏报情况。它被提供作为跨国安全统计、来源争议研究和叙述文本挖掘的基础；一个固定的、带DOI引用的快照随实时API一起提供。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37671 (HTTP 429)

Authors: Denys Kolomiiets

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.37671.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37671

Published: 2026-10-31T02:42:56.046Z

2. Scoring Explanations Without Reading Them: Structural Shortcuts in Automated-Interpretability Evaluation Data

Abstract:Automated interpretability scores an explanation of a sparse-autoencoder (SAE) feature by asking a model to predict, token by token, where the feature fires on evaluation examples or to judge whether marked tokens fit it. We find that the evaluation data standard pipelines build can be scored without reading the explanation. On Neuronpedia’s stored fuzzing runs for 500 random GPT-2 small features, 91.3% of activating examples are cropped with the peak on the centre token, and a rule that says “yes” when the centre token is marked reaches balanced accuracy 0.979, against 0.838 for the stored LLM scorers (recomputed from the stored per-item verdicts, on marks rebuilt from the pipeline’s code); a rule reading only the mark count reaches 0.969. The cue changes form across pipelines: current delphi always marks the same slot in its fuzzing negatives, and full-context runs mark negatives with more tokens. OpenAI’s simulation benchmark lists each neuron’s top-activating sequences first; a rule that predicts activity on the first five scores 0.227 in pooled correlation against 0.160 for GPT-4’s published simulator (post hoc). We give closed forms for what such rules can score; show that one scorer’s simulation scores are higher on stored snippets than on natural windows but not shown to be reordered; and build windows that remove the position cue, derive but do not build negatives that remove the mark cues, and separate the sequence cue by decomposition. On natural-text windows cut at random offsets, the best fixed-offset prior shared by all features scores 0.015. On them, that scorer, a typed decision model (Jev), reaches a pooled correlation of 0.331 over 49 features, against 0.013 with shuffled explanations, and an open 8B simulator with the original read-out reads the explanation there (0.194 against 0.019). Scores should be reported beside the best explanation-free rule on the same items.

中文摘要

摘要：自动可解释性通过让模型逐个标记预测稀疏自编码器（SAE）特征在评估示例中何处触发，或者判断标记的标记是否符合该特征，从而对其解释进行评分。我们发现，使用标准管道构建的评估数据可以在不读取解释的情况下进行评分。在Neuronpedia为500个随机GPT-2 small特征存储的模糊测试运行中，91.3%的激活示例在中心标记上被裁剪，并且一个规则在中心标记被标记时输出“是”，其平衡准确率达到0.979，而存储的LLM评分器（根据存储的每项结论重新计算，对从管道代码重建的标记）为0.838；仅读取标记数量的规则达到0.969。不同管道中线索形式会变化：当前Delphi在其模糊测试负样本中总是标记相同的槽位，而全上下文运行则使用更多标记标记负样本。OpenAI的模拟基准将每个神经元的最高激活序列列在前面；一个基于前五个序列预测激活的规则在汇总相关性中得分0.227，而GPT-4已发布模拟器的得分为0.160（事后分析）。我们给出了此类规则可评分的闭式形式；展示了一个评分器的模拟得分在存储的片段上高于自然窗口，但没有显示出重新排序；并构建了消除位置线索的窗口，推导但未构建消除标记线索的负样本，并通过分解分离序列线索。在按随机偏移裁剪的自然文本窗口上，由所有特征共享的最佳固定偏移先验得分为0.015。在这些窗口上，该评分器——一个类型化决策模型（Jev）——在49个特征上实现汇总相关性0.331，而打乱解释后为0.013，而一个原始读出开放8B模拟器在此读取解释得到0.194，对照0.019。应在同一项目旁边报告得分与最佳无需解释规则的得分。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37676 (HTTP 429)

Authors: Xupeng Chen

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.37676.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37676

Published: 2026-10-31T02:42:56.046Z

3. The Best Second Opinion Is Right Where the Primary Is Wrong: Partner Selection for Selective Span Tagging

Abstract:A span tagger that ranks its outputs by risk can defer the riskiest ones to a person, and a second model can sharpen this ranking. Which second model should it be? Partner selection is conditional, not global: the best second opinion is the model that is right where the primary is wrong. Across eight tagger types, three seeds and three English benchmarks (576 evaluated primary-partner pairings), choosing the partner by held-out accuracy on the primary’s errors recovers 82% of the best partner’s gain, against 69-70% for span F1 or a random pick. Frozen before a fourth dataset was chosen, the rule replicated there (Spearman 0.40 vs. 0.04 for span F1). Partner value is also directional. The predictive distribution is the second opinion: hard labels discard half of a partner’s value, and scoring the served label improves 94% of pairings. Fixed before the fourth dataset was chosen, this score also replicated there. For LoRA-tuned decoders, independently trained partners beat every readout of the primary network. Applied zero-shot to 51 MASSIVE locales, every primary has a helpful trained partner in every locale, and the preferred partner follows competence.

中文摘要

摘要：一个按风险对其输出进行排序的跨度标注器可以将最风险的输出交给人处理，而第二个模型可以进一步优化这一排序。第二个模型应该是什么？伙伴选择是有条件的，而非全局的：最佳的第二意见是当主模型出错时能正确的模型。在八种标注器类型、三个随机种子和三个英文基准（共评估576个主-伙伴配对）中，通过主模型错误上的保留数据精度选择伙伴能恢复最佳伙伴增益的82%，而基于跨度F1或随机选择仅能恢复69-70%。在第四个数据集选择之前固定的规则在该数据集上也重复了效果（Spearman 0.40，相对于跨度F1的0.04）。伙伴价值也是有方向性的。预测分布即第二意见：硬标签会丢失伙伴的一半价值，而对所提供标签进行评分能提升94%的配对效果。在第四个数据集选择之前固定，该评分也在该数据集上得到了重复验证。对于LoRA微调的解码器，独立训练的伙伴优于主网络的所有读出。在零样本条件下应用于51个MASSIVE数据集，每个主模型在每个数据集上都有一个有帮助的训练伙伴，而首选伙伴会随能力而定。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37678 (HTTP 429)

Authors: Varun Kotte

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.37678.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37678

Published: 2026-10-31T02:42:56.046Z

4. Exposure Is Not Commitment: Double-Stratified Evaluation of Latent Monitors for Tool-Using Agents

Abstract:Tool-using language agents are vulnerable to indirect prompt injection, and a growing literature probes model internals to catch unsafe actions before they execute, with sharply conflicting results. We argue much of the conflict is an evaluation artifact, and propose a double-stratification protocol: separate injection exposure from behavioral commitment, in both the evaluation and the training population. On the primary Qwen AgentDojo grid, monitors trained on the naive mixed unsafe-versus-rest contrast mostly learn exposure: they flag exposed trajectories whether or not the agent complies, and barely predict which cases turn unsafe. Retrained on exposed cases only, the same probe families predict the model’s own unsafe tool call up to two steps early, the maximum the three-step trajectories admit (logistic AUROC 0.853, matched by a trained TF-IDF text baseline at 0.855). Mistral-7B reaches 0.896 and OLMo-2-7B 0.912; their naive contrasts learn different things: inverted exposure ranking on one, both axes on the other. Representations separate on exposure (probe 1.000 versus trained text 0.908) and where visible context is sanitized or unavailable. Training contrast alone reproduces the shape of recent pre-action-probe negative results, one candidate explanation for the conflict: the signal exists, but the model’s own zero-shot forecast and NLA verbalizations do not reliably recover it. An unstratified AUROC overstates what a latent monitor knows, and what a naive contrast learns is population-dependent: evaluations should stratify both.

中文摘要

摘要：使用工具的语言代理容易受到间接提示注入的影响，越来越多的文献通过探测模型内部以在执行之前捕捉不安全行为，但结果存在明显冲突。我们认为大部分冲突是评估伪影所致，并提出一种双重分层协议：在评估和训练人群中，将注入暴露与行为承诺分开。在主要的Qwen AgentDojo网格上，基于天真的混合不安全与其他对比训练的监视器主要学习到暴露：它们会标记被暴露的轨迹，无论代理是否遵从，但几乎无法预测哪些情况会变得不安全。仅在暴露案例上重新训练时，同一系列探针可以提前最多两步预测模型自己的不安全工具调用，这是三步轨迹所允许的最大值（逻辑AUROC为0.853，与训练的TF-IDF文本基线0.855相匹配）。Mistral-7B达到0.896，OLMo-2-7B达到0.912；它们的天真对比学到的内容不同：一个模型获得了暴露排名的倒置，另一个则在两个轴上。表示在暴露上分离（探针1.000，相比训练文本0.908），以及在可见上下文被清理或不可用时也如此。仅凭训练对比可以再现近期行动前探针负面结果的形状，这是冲突的一个可能解释：信号存在，但模型自身的零样本预测和NLA口头表述无法可靠地恢复它。未分层的AUROC会高估潜在监视器所知道的内容，而天真的对比所学内容依赖于人群：评估应进行双重分层。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37679 (HTTP 429)

Authors: Rishabh Sinha

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.37679.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37679

Published: 2026-10-31T02:42:56.046Z

5. Artifact Distillation: Training-Free Transfer from Frontier to Small Language Models

Abstract:Closing the gap between frontier and smaller language models is usually approached by distillation: fine-tuning the small model (student) on a frontier model’s (teacher) outputs, which binds transfer to training compute and leaves the transferred knowledge largely opaque. We propose Artifact Distillation: a teacher solves training instances and externalizes its problem-solving procedure into an artifact bank that a student queries at inference time, with no gradient updates. Across four LiveBench tasks, a bank written by Claude Opus 4.5 lets Claude Haiku 4.5 outperform unaided Claude Sonnet 4.5 on spatial reasoning (0.707 vs. 0.647) at 60% of its cost, and reach Sonnet 4.5’s level on table joining at 89% of its cost. The same raw bank raises Qwen3.5-27B’s pooled score from 0.730 to 0.800, showing that transfer is not confined to the teacher’s model family.

中文摘要

摘要：弥合前沿模型与小型语言模型之间的差距通常通过蒸馏实现：在前沿模型（教师）的输出上对小模型（学生）进行微调，这会将迁移与训练计算绑定，并使迁移的知识大部分不透明。我们提出了“产物蒸馏”：教师解决训练实例并将其问题解决过程外化到一个产物库中，学生在推理时查询此库，无需梯度更新。在四个LiveBench任务中，由Claude Opus 4.5编写的产物库使Claude Haiku 4.5在空间推理任务上的表现优于未经辅助的Claude Sonnet 4.5（0.707对0.647），成本为其60%，并在表格合并任务上达到Sonnet 4.5的水平，成本为其89%。同样的原始产物库将Qwen3.5-27B的综合得分从0.730提高到0.800，显示出迁移并不限于教师的模型家族。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37680 (HTTP 429)

Authors: Mustafa Safa Ozdayi, Somin Wadhwa

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.37680.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37680

Published: 2026-10-31T02:42:56.046Z

6. Supervised Fine-Tuning for Reasoning Large Language Models: A Survey

Abstract:Large reasoning models have recently demonstrated remarkable capabilities in solving complex tasks, where supervised fine-tuning (SFT) on long chain-of-thought data serves as a crucial foundation for eliciting and enhancing their reasoning abilities. Despite rapid progress in both improving and analyzing reasoning-oriented SFT, the field still lacks a systematic survey that consolidates its fast-growing literature. To fill this gap, we present a comprehensive review of recent advancements in reasoning SFT, examining the literature through the dual lenses of methodological design and analytical investigation. First, we review methodological improvements across the SFT pipeline and categorize them into data-centric approaches and algorithm-centric innovations. Second, we reorganize analytical studies along three dimensions: data characteristics, optimization dynamics, and mechanistic insights. Finally, by synthesizing current research focuses and remaining bottlenecks, we outline promising future directions for reasoning SFT. We hope this survey deepens the understanding of reasoning SFT and paves the way for advanced reasoning models.

中文摘要

摘要：大型推理模型最近在解决复杂任务方面展示了显著的能力，其中在长链思维数据上的有监督微调（SFT）是激发和增强其推理能力的关键基础。尽管在改进和分析面向推理的SFT方面取得了快速进展，但该领域仍缺乏系统的综述来整合其快速增长的文献。为填补这一空白，我们呈现了一篇关于推理SFT最新进展的全面综述，从方法设计和分析研究两个角度审视相关文献。首先，我们回顾了SFT流程中的方法学改进，并将其分类为以数据为中心的方法和以算法为中心的创新。其次，我们沿三个维度重新组织了分析研究：数据特征、优化动态和机制洞察。最后，通过综合当前的研究重点和现有的瓶颈，我们概述了推理SFT的有前景的未来方向。我们希望这篇综述能够加深对推理SFT的理解，并为先进的推理模型铺平道路。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37681 (HTTP 429)

Authors: Yongcan Yu, Kuangpu Guo, Jian Liang, Lingxiao He, Meng Wang, Ran He

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.37681.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37681

Published: 2026-10-31T02:42:56.046Z

7. Rank-Aware Differentially Private Release of Listwise Preferences for LLM Alignment

Abstract:Listwise preference feedback offers richer supervision for large language model (LLMs) alignment than pairwise comparisons, but a full ranking also reveals more sensitive user preferences. Existing privacy-preserving alignment methods focus mainly on pairwise feedback, while listwise alignment under local differential privacy (LDP) remains unexplored. Extending LDP to this listwise domain raises a critical challenge: the high sensitivity of the full ranking space demands excessive noise, neutralizing the benefits of listwise supervision. To address this challenge, we analyze the interaction between supervision granularity and privacy noise in downstream training, and introduce a rank-aware exponential mechanism that privatizes listwise preference data into a low-sensitivity granularity suitable for downstream alignment. The mechanism leverages ranking information to sample a fixed-size binary partition, concentrating the noise near borderline items instead of perturbing all items uniformly as randomized response (RR) does. Empirical evaluations on downstream LLM alignment tasks show that our mechanism consistently outperforms existing LDP baselines at matched privacy budgets. The code is available at this https URL.

中文摘要

摘要：列表式偏好反馈相比于成对比较，为大语言模型（LLMs）的对齐提供了更丰富的监督，但完整排序也揭示了更多敏感的用户偏好。现有的隐私保护对齐方法主要关注成对反馈，而在本地差分隐私（LDP）下的列表式对齐尚未被探索。将LDP扩展到列表式领域带来了一个关键挑战：完整排序空间的高敏感性需要过多噪声，从而抵消了列表式监督的优势。为了解决这一挑战，我们分析了下游训练中监督粒度与隐私噪声的交互，并引入了一种排序感知指数机制，将列表式偏好数据私有化为适合下游对齐的低敏感性粒度。该机制利用排序信息采样固定大小的二分划分，将噪声集中在边界项目附近，而不是像随机响应（RR）那样均匀扰动所有项目。在下游LLM对齐任务上的实证评估显示，我们的机制在相同的隐私预算下持续优于现有LDP基线。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: 2610.37682 (HTTP 429)

Authors: Junwei Chen, Manjiang Yu, Pengpeng Qiao, Yang Cao

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.37682.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37682

Published: 2026-10-31T02:42:56.046Z

8. Prompt-agnostic Language Models

Abstract:Despite their strong performance, large language models remain highly sensitive to prompt formulation. This issue is typically addressed at inference-time through manual prompt optimization, which introduces overhead and lacks scalability. Conversely, we focus on train-time approaches that build prompt robustness into the model during instruction tuning. A common strategy is to mix multiple paraphrases of each prompt into training. We demonstrate that across trainable parameters, the per-paraphrasing gradients interfere, with 57% exhibiting sign conflicts and the naive gradient average discarding approximately a fifth of the magnitude. Therefore, we develop two gradient-merging methods, Prompt-TIES and Prompt-DARE, that address this interference by leveraging interference-resolution operators from model and gradient merging and applying them to prompt gradients. In this study, we compare the efficacy of our methods to increase model robustness to train-time robustness methods spanning all three established classes: data construction, consistency regularization, and contrastive alignment. The evaluation is conducted across four models from two families at two scales. We observe the interaction between prompt phrasing and the specific examples accounts for 54% of per-example variability on average, and exceeds 70% on tasks such as ANLI and WiC. The gradient-merging methods directly improve this local robustness and nearly halve per-example variance on fully fine-tuned models at little cost to overall accuracy. The other train-time methods enhance raw performance, but fail to improve per-example variance reliably. Code: this https URL. Model checkpoints: this https URL.

中文摘要

摘要：尽管大型语言模型表现出强大的能力，但它们仍然对提示的表述高度敏感。这个问题通常在推理阶段通过手动优化提示来解决，这会带来开销并且缺乏可扩展性。相反，我们关注训练阶段的方法，将提示的鲁棒性在指令调优过程中内建到模型中。一种常见策略是将每个提示的多种改写混合到训练中。我们证明了在可训练参数中，每种改写的梯度会相互干扰，其中57%的梯度存在符号冲突，而简单的梯度平均会丢弃大约五分之一的幅值。因此，我们开发了两种梯度合并方法——Prompt-TIES和Prompt-DARE，通过利用模型和梯度合并中的干扰解决算子并将其应用于提示梯度来解决这种干扰。在本研究中，我们将我们的方法与涵盖所有三类已建立路径的训练阶段鲁棒性方法进行了比较：数据构建、一致性正则化和对比对齐。评估在两类模型的四个模型上，在两个规模下进行。我们观察到，提示表述与具体示例之间的交互平均占每个示例变异性的54%，在如ANLI和WiC的任务中甚至超过70%。梯度合并方法直接提高了这种局部鲁棒性，并在对完全微调模型时几乎将每个示例的方差减半，同时对整体准确性造成的影响很小。其他训练阶段方法提升了原始性能，但未能可靠地改善每个示例的方差。代码：这个https URL。模型检查点：这个https URL。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37683 (HTTP 429)

Authors: Frederic Sadrieh

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.37683.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37683

Published: 2026-10-31T02:42:56.046Z

9. An Introduction to Large Language Models for Scientists and Professionals

Abstract:Large Language Models (LLMs) have rapidly become important tools for science, education, medicine, and professional work. Despite their apparent simplicity from the user’s perspective, effective and responsible use requires an understanding of how these systems work, what their limitations are, and how their outputs should be evaluated. I wrote this document as a basis for my lecture on LLM in the master in artificial intelligence in medicine at the university of Turin, Italy. This book provides a practical and accessible introduction to LLMs for scientists and professionals. It explains the fundamental ideas behind tokenisation, embeddings, training, inference, context windows, prompting, and instruction hierarchies, while deliberately avoiding mathematical detail. Particular attention is given to the limitations of LLM-generated answers, including hallucinations, uncertainty, reproducibility, bias, and the need for systematic verification and human oversight. The book also discusses privacy, governance, security, and the responsible handling of personal and sensitive information. More advanced topics include Retrieval-Augmented Generation (RAG), retrieval quality, chunking, embeddings, vector search, agents, tools, and workflows. A substantial practical exercise guides the reader through running a local LLM with Ollama and Open WebUI and using it to interact with a collection of scientific papers. The exercise emphasizes evidence retrieval, source verification, uncertainty, disagreement between studies, and the distinction between plausible language and scientifically supported conclusions. No programming knowledge is required.

中文摘要

摘要：大语言模型（LLMs）迅速成为科学、教育、医学和专业工作中的重要工具。尽管从用户的角度看它们似乎很简单，但有效且负责任的使用需要理解这些系统的工作原理、它们的局限性以及如何评估其输出。我编写这份文档是作为我在意大利都灵大学人工智能医学硕士课程中关于LLM讲座的基础。本书为科学家和专业人士提供了对LLM的实用且易于理解的介绍。它解释了分词、嵌入、训练、推理、上下文窗口、提示和指令层级的基本理念，同时刻意避免数学细节。特别关注LLM生成答案的局限性，包括幻觉、不确定性、可重现性、偏差以及系统验证和人工监督的必要性。本书还讨论了隐私、治理、安全以及对个人和敏感信息的负责任处理。更高级的话题包括检索增强生成（RAG）、检索质量、分块、嵌入、向量搜索、代理、工具和工作流程。一个实质性的实践练习引导读者通过使用Ollama和Open WebUI运行本地LLM，并利用它与一系列科学论文进行交互。该练习强调证据检索、来源验证、不确定性、研究之间的分歧，以及合理语言与科学支持结论的区别。不需要编程知识。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37684 (HTTP 429)

Authors: Umberto Michelucci

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.37684.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37684

Published: 2026-10-31T02:42:56.046Z

10. From Geometry to Behavior: How Instruction Exposure Unlocks Latent Rhetorical Directions in LLMs

Abstract:Language models do not merely memorize facts; they absorb rhetorical strategies from pretraining data, including the human tendency to reason by analogy. We ask: is this absorbed capability geometrically accessible before any instruction tuning, and what determines whether it can be activated at inference time? We extract an “Analogy Vector” from the residual stream via difference-in-means and show it enables steerable explanation style across seven instruction-tuned models (Gemma-2, Llama-3.1, Qwen-2.5; 2B-32B), improving harmonic mean from 3.22 to 3.94 (+22%) for Gemma-2-9B-IT on 300 held-out test prompts. A small human validation study confirms the LLM judge on Analogy Quality. Our central finding, from seven matched base-instruct pairs, is a gradient of behavioral accessibility: the analogy direction exists in every base model, but its behavioral accessibility under linear intervention scales continuously with instruction exposure during training. Gemma-2 base models (strict pretraining/IT separation) show flat steering landscapes: present but behaviorally inert. Qwen-2.5 base models (instruction data woven into pretraining) are already partially accessible before any explicit fine-tuning; base-instruct cosine similarity of the analogy vector reaches 0.94-0.97 for Qwen versus 0.37-0.66 for Gemma, providing geometric corroboration of the behavioral gradient. This challenges the binary “base vs instruct” framing: the relevant variable is how much instruction exposure a model has received across its full training pipeline. We position activation steering as a complementary interpretability tool that reveals how instruction exposure organizes rhetorical knowledge in activation space.

中文摘要

摘要：语言模型不仅仅是记忆事实；它们还从预训练数据中吸收修辞策略，包括人类通过类比进行推理的倾向。我们提出问题：这种吸收的能力在任何指令微调之前是否在几何上可访问，以及是什么决定了它在推理时是否可以被激活？我们通过均值差异从残差流中提取了一个“类比向量”，并展示它能够在七个指令微调模型（Gemma-2、Llama-3.1、Qwen-2.5；2B-32B）中实现可控的解释风格，在300个保留的测试提示上将Gemma-2-9B-IT的调和平均值从3.22提升到3.94（+22%）。一项小规模的人类验证研究确认了LLM在类比质量上的判断。我们从七对匹配的基础-指令模型中得出的核心发现是行为可访问性的梯度：类比方向存在于每个基础模型中，但其在行为上的可访问性在训练过程中线性干预下随着指令暴露程度的增加而持续变化。Gemma-2基础模型（严格的预训练/指令微调分离）显示出平坦的操控景观：存在但在行为上无效。Qwen-2.5基础模型（将指令数据融入预训练）在任何显性微调之前已经部分可访问；类比向量的基础-指令余弦相似度在Qwen中达到0.94-0.97，而在Gemma中为0.37-0.66，为行为梯度提供了几何上的证据。这挑战了二元的“基础 vs 指令”框架：关键变量是模型在整个训练过程中接触了多少指令数据。我们将激活操控定位为一种互补的解释工具，它揭示了指令暴露如何在激活空间中组织修辞知识。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37685 (HTTP 429)

Authors: Shengjie Xu

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.37685.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37685

Published: 2026-10-31T02:42:56.046Z

Agent Domain Papers

1. LRMoE: Demand-Side Load Re-Calibration for Efficient MoE Training on Memory-Constrained Nodes

Abstract:Sparse Mixture-of-Experts (MoE) models offer scalable capacity for Web applications, but uneven routing creates computation and communication bottlenecks. Existing placement and replication methods redistribute this workload, yet additional expert copies compete with token-related buffers for limited GPU memory. We present LRMoE, a memory-aware MoE training system that jointly recalibrates backward demand and balances the retained workload. During forward execution, it plans retention quotas and expert placement under resource constraints while preserving all routed computations. During backpropagation, a load recalibration metric estimates each token-expert pair’s contribution to the expert update and selects pairs within these quotas. A capacity-aware routing algorithm distributes retained pairs among eligible instances of their original experts. Physical compaction then forms smaller communication payloads and variable-size grouped GEMMs, reducing backward traffic, computation, and execution-buffer demand. Experiments on Qwen3-30B-A3B and GLM4.5-106B-A12B show that LRMoE reduces cumulative MoE-layer training time to fixed-loss targets by 21.2-44.8% relative to SYMI.

中文摘要

摘要：稀疏专家混合（MoE）模型为 Web 应用提供了可扩展的容量，但不均衡的路由会产生计算和通信瓶颈。现有的放置和复制方法可以重新分配工作负载，但额外的专家副本会与与 token 相关的缓冲区争夺有限的 GPU 内存。我们提出了 LRMoE，一种内存感知的 MoE 训练系统，它能够联合重新校准反向需求并平衡保留的工作负载。在前向执行过程中，它在资源约束下规划保留配额和专家放置，同时保留所有路由计算。在反向传播过程中，一个负载重新校准指标评估每个 token-专家对专家更新的贡献，并在这些配额内选择配对。容量感知路由算法将保留的配对分布到其原始专家的合格实例中。物理压缩随后形成更小的通信负载和可变大小的分组 GEMM，从而减少反向流量、计算和执行缓冲需求。在 Qwen3-30B-A3B 和 GLM4.5-106B-A12B 上的实验表明，相较于 SYMI，LRMoE 可将 MoE 层训练到固定损失目标的累计时间减少 21.2%~44.8%。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37720 (HTTP 429)

Authors: Jiaming Hu, Zhida Jiang, Zhaolong Xing, Yu Li, Zhen Chen, Ke Zhang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37720.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37720

Published: 2026-10-31T02:43:16.304Z

2. Amortized Localized Posterior Inference for Model Calibration with Application to an Ocean Turbulence Closure

Abstract:Model discrepancy may prevent a single calibration parameter from reproducing a physical response across the context domain. In addition, uncertainties in observations and contexts affect calibration and must be accounted for and quantified. We construct a family of localized posteriors by weighting observation likelihoods according to their proximity to a target context. Gaussian localization weights incorporate context uncertainty through covariance inflation, with reflection at domain boundaries and an optional correction for uneven context sampling. A conditional normalizing flow approximates the family of posteriors through amortized variational inference. Integrating the localized posteriors under a prescribed context measure gives the marginalized parameter distribution. In a twin experiment with the Branin-Hoo model, we compare global and localized inference under uniform and skewed context distributions. Localization reduces the reported aggregate predictive errors relative to the global posterior. An application to the k-$\epsilon$ turbulence closure in a single-column ocean model, with atmospheric surface forcing as context, shows that localized posterior calibration reduces temperature and zonal-velocity errors relative to deterministic and global-posterior calibration. Both twin experiments are constructed to admit local compensation.

中文摘要

摘要：模型差异可能阻止单个校准参数在上下文域中重现物理响应。此外，观测和上下文中的不确定性会影响校准，必须加以考虑和量化。我们通过根据观测似然与目标上下文的接近程度加权，构建了一类局部后验。高斯局部化权重通过协方差膨胀纳入上下文不确定性，在域边界反射，并可选地修正不均匀的上下文抽样。条件归一化流程通过摊销变分推断近似后验族。在规定上下文度量下整合局部后验即可得到边际参数分布。在Branin-Hoo模型的双重实验中，我们比较了均匀和偏斜上下文分布下的全局和局部推断。局部化降低了报告的总预测误差相对于全局后验。在单柱海洋模型中，以大气表面强迫为背景的k-$\ε⁻⁭闭合应用显示，局部后验校准相对于确定性和全局后验校准，能够降低温度和纬向速度误差。这两项实验均设计为允许局部补偿。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37731 (HTTP 429)

Authors: Exaucé Luweh Adjim Ngarti, Élise Arnaud, Léo Nicoletti, Victor Trappler, Arthur Vidard

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37731.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37731

Published: 2026-10-31T02:43:16.304Z

3. ThinkJev: One Thought, Every Budget for Calibrated Typed Decisions

Abstract:Many uses of language models end in a typed decision: a choice among a closed set of options, whose probabilities code can act on. Reading those probabilities from one forward pass is fast but often wrong on decisions that take several steps, such as arithmetic, counting, and date comparison. Letting the model think first fixes most of these errors, but costs seconds per decision instead of milliseconds. To pay for thinking only where it helps, ThinkJev builds a fast and a slow system from one open hybrid-thinking model that can answer directly or after a thought. System 1 reads the model’s probabilities for the option letters when it answers directly; System 2 reads the same letters after the model has thought on the same cached prompt, so escalating repeats no work and both systems give probabilities over the same options. Because the letters can be read after any number of thought tokens, one sampled thought also gives the answer under every thinking budget. Without training, ThinkJev lets System 1 answer alone only where a statistical test on labeled dev data guarantees that its accepted answers stay under a target error rate. At a 5% target, Qwen3.5-27B answers 24.5% of 8359 test decisions this way with a realized error of 3.9% (Qwen3.5-4B and 9B: 6.5% and 0.0%). For the rest, stopping each thought as soon as its answer is confident reaches 81.6% accuracy, against 82.7% for always thinking, with 72% of the thought tokens. Choosing when to think from System 1’s confidence alone saves less, because the decisions System 1 is least sure of are often not the ones thinking fixes. On the public items of JevBench, an independent benchmark of typed-decision services, Qwen3.5-27B thinks on 29.4% and is not distinguishable in accuracy from the two thinking APIs listed (exploratory). Code and data generators are available at this https URL.

中文摘要

摘要：语言模型的许多使用最终都以键入决策告终：在一组封闭选项中进行选择，其概率代码可以操作。通过一次前向传播读取这些概率很快，但在需要多步操作的决策上通常是错误的，例如算术、计数和日期比较。让模型先思考可以纠正大多数这些错误，但每次决策耗时从毫秒增加到秒。为了仅在有帮助的情况下进行思考，ThinkJev 从一个开放的混合思考模型构建了一个快速系统和一个慢速系统，该模型可以直接回答或经过思考后回答。系统1在模型直接回答时读取选项字母的概率；系统2在模型对相同缓存提示进行思考后读取相同的字母，因此升级不会重复工作，两个系统都给出相同选项的概率。由于这些字母可以在任意数量的思考令牌后读取，因此一次采样的思考也能提供在任何思考预算下的答案。在无需训练的情况下，ThinkJev 仅在经过标注开发数据的统计测试保证其接受的答案保持在目标错误率以下时，让系统1单独回答。在5%目标下，Qwen3.5-27B 以这种方式回答了8359个测试决策中的24.5%，实际误差为3.9%（Qwen3.5-4B和9B分别为6.5%和0.0%）。对于其余情况，在答案自信时立即停止每次思考可达到81.6%的准确率，而总是思考时为82.7%，使用了72%的思考令牌。仅依据系统1的置信度决定何时思考节省较少，因为系统1最不确定的决策通常不是思考能修正的决策。在JevBench的公开条目上，这是一个独立的键入决策服务基准，Qwen3.5-27B 对29.4%的条目进行思考，并且在准确率上无法区别于列出的两个思考API（探索性）。代码和数据生成器可在此 https URL 获取。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37732 (HTTP 429)

Authors: Xiaolu Ding, Kai Yang, Dongdong Liu

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37732.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37732

Published: 2026-10-31T02:43:16.304Z

4. OpenFrontiers: A Collection of Open Research Problems and a Benchmark for Scientific Rediscovery

Abstract:Large language models (LLMs) have shown increasing promise in research-level problem solving, yet a comprehensive collection of such problems is still missing, and evaluating them on research-level problems remains challenging. We introduce OpenFrontiers, a cross-disciplinary resource that combines a structured corpus of open research problems with a retrospective benchmark of recently resolved problems in mathematics. The corpus records problem statements, context, progress, and other information for 12,151 problems. The benchmark contains 200 problems, 110 resolved in 2026 and 90 still open, each with a Lean~4 statement checked by automated probes and two independent reviews; a subset HQ-50 further filters based on quality score, etc. Our evaluation protocol grades formal proofs with two independent checkers in an isolated sandbox under closed-book, frozen-retrieval, and live-web conditions. Evaluating seven LLM agents on the HQ-50 subset under a standardized budget, we find that closed-book rediscovery is rare, at most 7 of 40 resolved problems, whereas live-web success mostly reflects retrieval of published solutions. By linking a broad research-problem corpus to a formally graded evaluation subset with documented solutions, OpenFrontiers provides a foundation for measuring research-level rediscovery and studying progress toward solving currently open scientific problems. The dataset and benchmark could be found in this https URL.

中文摘要

摘要：大型语言模型（LLMs）在研究级问题解决中显示出越来越大的潜力，但这样的问题的综合收集仍然缺失，对这些问题的评估也仍然具有挑战性。我们引入了OpenFrontiers，这是一个跨学科资源，将结构化的开放研究问题语料库与最近在数学上解决的问题的回顾性基准相结合。该语料库记录了12,151个问题的题目陈述、背景、进展及其他信息。基准测试包含200个问题，其中110个在2026年已解决，90个仍未解决，每个问题都附有由自动探测器检查并经过两次独立审查的Lean~4表述；一个子集HQ-50进一步根据质量评分等进行筛选。我们的评估协议在隔离沙箱中使用两名独立检查员对形式证明进行评分，条件包括闭卷、冻结检索和实时网络。对HQ-50子集下的七个LLM代理进行标准化预算评估时，我们发现闭卷重新发现问题的情况很少，最多仅有40个已解决问题中的7个，而在实时网络条件下的成功主要反映了已发布解决方案的检索。通过将广泛的研究问题语料库与具有正式评分且附有文档化解决方案的评估子集关联，OpenFrontiers为测量研究级重新发现能力及研究当前未解决科学问题的进展提供了基础。数据集和基准测试可在此https URL中找到。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37739 (HTTP 429)

Authors: Jiarui Yao, Ruida Wang, Tong Zhang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37739.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37739

Published: 2026-10-31T02:43:16.304Z

5. WebAnything: From the Wild Web to Scalable Agent Environments

Abstract:Web agents require diverse, executable, and resettable website environments for reproducible training and evaluation. Human-assisted mirrors demand per-site engineering, while generative construction can diverge from specified real websites. Capture-and-replay preserves observations, with interaction coverage constrained by captured content and replay capabilities. We present WebAnything, a fully automated framework that reconstructs source-grounded, reusable Web environments from real URLs without per-site manual implementation or repair after initial setup and authorization. An Explorer-Builder collects browser-observable structures, resources, and interaction outcomes; an independent verifier checks frozen candidates and guides local repair or targeted source recapture. We construct 244 website artifacts across 14 application domains and 53 categories, supporting 12,984 executed and accepted tasks. Compared with human-assisted WebHarbor, WebAnything achieves higher aggregate scores on three fidelity metrics, with 12 wins, two ties, and no losses in a blind audit. Fine-tuning Qwen3.5-9B on 5,856 interaction trajectories improves REAL success by 9.9 percentage points, Odysseys-Easy Rubric Avg and Task Avg by 6.25 and 6.32 percentage points, and WebArena-Verified Overall success by 13.33 percentage points under the fixed harness. Code: this https URL

中文摘要

摘要：Web代理需要多样化、可执行且可重置的网站环境，以实现可复现的训练和评估。人工辅助的镜像需要针对每个网站进行工程处理，而生成式构建可能偏离指定的真实网站。抓取与重放可以保留观察结果，但交互覆盖受限于捕获的内容和重放能力。我们提出了WebAnything，一个全自动框架，可从真实URL重建以源为基础、可重复使用的Web环境，无需在初始设置和授权后对每个站点进行手动实现或修复。Explorer-Builder收集浏览器可观察的结构、资源和交互结果；独立验证器检查冻结的候选项，并指导本地修复或有针对性的源重新捕获。我们构建了覆盖14个应用领域和53个类别的244个网站工件，支持12,984个已执行且已接受的任务。与人工辅助的WebHarbor相比，WebAnything在三个保真度指标上的综合得分更高，在一次盲审中获得12次胜利、2次平局，且未出现失败。在5,856条交互轨迹上微调Qwen3.5-9B，使REAL成功率提高了9.9个百分点，Odysseys-Easy Rubric平均分和任务平均分分别提高6.25和6.32个百分点，WebArena-Verified整体成功率在固定测试环境下提高13.33个百分点。代码：此https网址

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37786 (HTTP 429)

Authors: Tianci Xie, Zikun Cai, Chaoyun Zhang, Pinxue Guo, Bingxuan Li, Rui Zheng, Hang Yan

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37786.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37786

Published: 2026-10-31T02:43:16.304Z

6. What Does an Ontology Catch? Curated, LLM-Generated and Data-Derived Ontologies as Checkers of LLM Agents over Knowledge Graphs

Abstract:LLM agents that answer questions over Web knowledge graphs are increasingly wrapped in checks derived from the graph’s ontology, on the assumption that a formal schema makes a reliable judge. Yet a checker can be wrong as well: it can miss errors, and it can flag correct answers and push the agent into “fixing” them. Whether a curated ontology is worth its cost as such a judge, and how it compares with an ontology an LLM writes or a schema mined from the data, is unknown. We make the checker itself the object of study. Across curated, LLM-generated and data-derived ontologies, used as prompt context, as a checker, or both, and across open-weight and proprietary agents on public graphs and a private graph of nearly $10^{9}$ triples, we account for what ontology-derived checks catch and what they wrongly flag. We show how each source trades caught errors against false alarms, how a curated schema used as published can contradict its own data, which errors structural checks miss, and whether combining sources or giving the agent the checker’s specific feedback pays off. On UniProt, the curated ontology flags more correct answers than either an ontology derived from the data or one an LLM derives from a summary of the data with masked identifiers. On real agent answers, its flags are only marginally more precise than random flags.

中文摘要

摘要：针对网络知识图谱回答问题的LLM代理越来越多地被包装在源自图谱本体的检查机制中，假设正式的模式能够成为可靠的判断标准。然而，检查器也可能出错：它可能漏掉错误，也可能标记正确答案并推动代理去“修正”它们。作为这样一个判断者，人工策划的本体是否值得其成本，以及它与LLM生成的本体或从数据中挖掘的模式相比如何，目前尚不清楚。我们将检查器本身作为研究对象。在策划的、LLM生成的以及基于数据的本体中，无论是作为提示上下文、作为检查器，或两者兼有，以及在公开权重和专有代理上，在公共图谱和一个接近$10^{9}$三元组的私有图谱上，我们都考察了本体衍生检查捕获了什么错误以及错误标记了什么。我们展示了每种来源如何在捕获错误与误报之间权衡，以及策划的模式按原样使用时如何可能与自身数据矛盾，结构检查错过了哪些错误，以及结合多种来源或向代理提供检查器的具体反馈是否有效。在UniProt上，策划的本体比从数据中派生的本体或LLM从带有掩码标识符的数据摘要中派生的本体标记了更多正确答案。在实际代理的回答中，其标记的精确性仅略高于随机标记。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37809 (HTTP 429)

Authors: Michal Orzechowski, Rund Tawfiq, Tomasz Kosciolek

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37809.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37809

Published: 2026-10-31T02:43:16.304Z

7. Measuring Where Intent Gets Lost in Agentic Smart Contract Synthesis

Abstract:AI agents can now write software from a plain-English description, but code that runs is not the same as code that does what was asked. This gap matters most for smart contracts, the programs behind Web3 applications that move money on a blockchain and cannot be fixed once deployed. We ask how to check whether an AI-written contract behaves as it should, and where the requested behavior gets lost. We design an evaluation built on execution rather than opinion. We replay the same transactions on an AI-written contract and on the original it should reproduce and measure how often they agree. We test every grading tool on planted mistakes, compare against an answer key written without the description, and audit our own scorer until known wrong answers cannot pass. Common methods, such as AI judges and standard analysis tools, miss obvious mistakes, and scoring choices matter too. Checking only the functions every system includes hid about half the difference between pipelines. We then propose an elicitation-first architecture. Before writing code, the agent asks the client a short checklist of questions, checks the contract against an independently written executable answer key, and returns disagreements to the client instead of silently rewriting code. Asking first matters far more than reviewing or patching afterwards, across three AI models, while automated security fixes often break requested behavior. A few well-chosen questions recover much of what a short description omits, and their value can be estimated without the original contract. To our knowledge, this is the first study in smart contract synthesis to gather missing requirements before generation and check the result against an independently written executable model. The lessons extend to any setting where AI agents turn requests into programs. We release our tools, 600 tasks and all results.

中文摘要

摘要：AI 代理现在可以根据纯英文描述编写软件，但能运行的代码并不等同于能完成所要求操作的代码。这种差距在智能合约中尤为重要，智能合约是支撑 Web3 应用的程序，它在区块链上移动资金并且一旦部署无法修改。我们探讨如何检查 AI 编写的合约是否按预期运行，以及请求的行为在哪里丢失。我们设计了一种基于执行而非主观意见的评估方法。我们在 AI 编写的合约和原始合约上重放相同的交易，并测量它们一致的频率。我们在植入的错误上测试每个评分工具，并与未参考描述编写的答案键对比，同时审计我们自己的评分器，直到已知错误的答案无法通过。常见方法，如 AI 裁判和标准分析工具，会遗漏明显错误，评分选择也同样重要。仅检查每个系统包含的函数会隐藏大约一半管线间的差异。随后，我们提出了一种以提问为先的架构。在编写代码之前，代理向客户询问简短的核对问题清单，将合同与独立编写的可执行答案键进行比对，并将不一致之处反馈给客户，而不是默默重写代码。先提问的重要性远超事后复审或修补，这在三个 AI 模型中都得到了验证，而自动安全修复通常会破坏请求的行为。一些精心选择的问题可以追回简短描述遗漏的大部分内容，其价值甚至无需原始合约即可估算。据我们所知，这是智能合约生成领域中首个在生成前收集缺失需求并将结果与独立编写的可执行模型进行比对的研究。其经验可推广至任何 AI 代理将请求转化为程序的场景。我们发布了我们的工具、600 个任务及所有结果。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37830 (HTTP 429)

Authors: Abhinav Goel, Agostino Capponi, Alfio Gliozzo, Chaitya Shah

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37830.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37830

Published: 2026-10-31T02:43:16.304Z

8. Detecting Covert Collusion between AI Agents through Partner-Outsider Positive Listening

Abstract:AI agents that share a secret convention can coordinate through messages that look innocent, so a monitor that reads only the messages can miss them. We propose a behavioural test: a partner that uses a code reacts to the exact wording of a message, whereas an uninvolved agent placed in its seat does not. Partner—Outsider Positive Listening (PoPLi) is the difference between the two agents in how much their action changes when the message is reworded with its meaning kept, and it is computed after the fact by replaying logged prompts. In a simulated market with five conditions (no collusion, a joint reward with neutral messages, the same with an openly stated intention, a covert first-letter code, and the code with a stated intention), PoPLi is between 0.48 and 0.59 whenever the code is present (95% intervals all above 0.29) and between -0.10 and -0.01 without it. Message-only monitors did not separate the conditions: one reading the speaker’s messages answered “no” in all 50 markets, and one reading every seller’s messages flagged 9 of 10 markets with no collusion. PoPLi detects the use of a private channel, not collusion as such, and it cannot detect collusion through incentives alone or codes carried by meaning.

中文摘要

摘要：共享秘密约定的人工智能代理可以通过看似无害的信息进行协调，因此只读取信息的监控器可能会漏掉这些行为。我们提出了一种行为测试：使用代码的伙伴会对信息的字面措辞做出反应，而放在其位置的无关代理则不会。伙伴—外部者积极倾听（PoPLi）是两类代理在信息重新措辞后其行为变化程度上的差异，并通过回放已记录的提示事后计算。在具有五种条件的模拟市场中（无共谋、与中性信息的联合奖励、同样情况下公开声明意图、隐蔽的首字母代码，以及带有声明意图的代码），每当存在代码时，PoPLi介于0.48和0.59之间（95%区间均高于0.29），而没有代码时介于-0.10和-0.01之间。仅依靠信息的监控器无法区分这些条件：一个读取发言者信息的监控器在所有50个市场中都回答“否”，而一个读取每个卖家信息的监控器却在10个无共谋的市场中标记了9个。PoPLi可以检测私有通道的使用，而非共谋本身，它无法仅通过激励或信息传递的意义代码来检测共谋。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37849 (HTTP 429)

Authors: Kevin Yandoka Denamganaï

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37849.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37849

Published: 2026-10-31T02:43:16.304Z

9. Evaluating and improving AI agents for sequencing protocol understanding

Abstract:Existing scientific-agent benchmarks rarely evaluate whether agents can reconstruct the exact molecular structures and generation workflows specified by sequencing protocols, a capability critical for analyzing customized sequencing assays. We introduce LibStructBench, a document-grounded benchmark for reconstructing sequencing-library structures and molecular workflows. Our experiments show that frontier agents often produce approximately correct reconstructions but remain unreliable at exact reconstruction, particularly in molecular-state representation and workflow connectivity. Switching the underlying language model produced larger average performance differences than switching agent harnesses, with a statistically significant difference only for molecular-state reconstruction. Additional knowledge improved oligo-family recovery in a small pilot without improving mean molecular-workflow scores. Autonomous accumulation of procedural rules and cross-protocol examples showed non-monotonic validation performance and no mean held-out molecular-workflow improvement in the tested trajectory. These findings highlight the need for more controlled and verifiable approaches to scientific reasoning about molecular processes in complex experimental protocols.

中文摘要

摘要：现有的科学代理基准很少评估代理是否能够重建测序协议指定的精确分子结构和生成工作流程，而这种能力对于分析定制测序实验至关重要。我们引入了 LibStructBench，这是一种基于文档的基准，用于重建测序文库结构和分子工作流程。我们的实验表明，前沿代理通常能够生成大致正确的重建结果，但在精确重建方面仍不可靠，尤其是在分子状态表示和工作流程连通性方面。切换底层语言模型产生的平均性能差异大于切换代理工具，且只有在分子状态重建方面差异具有统计学显著性。额外的知识在一个小型试点中改善了寡核苷酸家族的恢复，但未提高平均分子工作流程得分。程序规则和跨协议示例的自主积累显示了非单调的验证性能，并且在测试的轨迹中未提高平均保留的分子工作流程。上述发现凸显了在复杂实验协议中进行分子过程科学推理时，需要更可控和可验证的方法。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37851 (HTTP 429)

Authors: Chi-Lam Poon, Michael Goneos, Zezheng Song, Xi Chen

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37851.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37851

Published: 2026-10-31T02:43:16.304Z

10. Seq@K: Evaluating Language Models Under Sequential Feedback

Abstract:Language-model agents differ fundamentally from prior intelligent systems in their ability to learn from natural language in context. This capability enables agents to iteratively refine their answers in response to feedback, a prevalent mode of operation in modern agentic workflows. However, the community primarily ranks models by metrics such as Pass@1 or Pass@K, which cannot capture how they utilize feedback on prior attempts. We introduce Seq@K, a natural evaluation protocol that measures whether a model can solve a problem within K sequential attempts, provided feedback after each previous failure. Unlike multi-turn agentic benchmarks that examine long-horizon planning, Seq@K isolates and formalizes the core capability of refinement from feedback. We evaluate Seq@K on a suite of 7 benchmarks and 16 frontier models, and show that Seq@K reveals a materially different model-selection landscape: under Seq@K, (1) model rankings often differ, (2) the Pareto frontiers admit more efficient configurations and (3) the gains from choosing the optimal model configuration can widen or shrink relative to those of Pass@K. We further show that analyzing model trajectories under Seq@K helps characterize iterative refinement behavior, such as the rate at which models address issues raised in feedback, or how rigidly they hold onto their own answers. Finally, we study how the nature of feedback affects models’ iterative refinement, showing that practitioners must carefully design and benchmark the feedback loops of iterative agentic systems. Together, these results establish how Seq@K offers a more faithful metric for intelligent systems that typically operate in such feedback loop, and the necessity to consider Seq@K in evaluating models for these systems.

中文摘要

摘要：语言模型代理与以往的智能系统在根本上不同之处在于其能够从上下文中的自然语言中学习。这种能力使代理能够根据反馈迭代地优化其答案，这是现代代理工作流程中普遍的操作模式。然而，社区主要通过诸如 Pass@1 或 Pass@K 的指标对模型进行排序，这无法捕捉模型如何利用对先前尝试的反馈。我们引入了 Seq@K，一种自然的评估协议，用于衡量模型是否能够在 K 次连续尝试内解决问题，并在每次先前失败后获取反馈。与考察长周期规划的多轮代理基准不同，Seq@K 分离并形式化了从反馈中进行改进的核心能力。我们在 7 个基准和 16 个前沿模型上评估了 Seq@K，结果表明 Seq@K 揭示了一个实质上不同的模型选择格局：在 Seq@K 下，(1) 模型排名往往不同，(2) 帕累托前沿允许更高效的配置，(3) 选择最优模型配置所带来的收益相对于 Pass@K 可能扩大或缩小。我们进一步展示了在 Seq@K 下分析模型轨迹有助于刻画迭代改进行为，例如模型解决反馈中提出问题的速率，或它们坚持自身答案的程度。最后，我们研究了反馈的性质如何影响模型的迭代改进，表明在设计和评估迭代代理系统的反馈循环时，实践者必须谨慎。综合来看，这些结果确立了 Seq@K 如何为通常在此类反馈循环中运行的智能系统提供更准确的衡量标准，以及在评估此类系统模型时考虑 Seq@K 的必要性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37859 (HTTP 429)

Authors: Thomson Yen, Gilbert Yang, Naimeng Ye, Minghui Chen, Dong Xia, Tianyi Shen, Yechan Jeon, Yaqin Chen, Daksh Mittal, Fangyu Wu, Hongseok Namkoong

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37859.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37859

Published: 2026-10-31T02:43:16.304Z

Evaluation Domain Papers

1. LRMoE: Demand-Side Load Re-Calibration for Efficient MoE Training on Memory-Constrained Nodes

中文摘要

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37720 (HTTP 429)

Authors: Jiaming Hu, Zhida Jiang, Zhaolong Xing, Yu Li, Zhen Chen, Ke Zhang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37720.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37720

Published: 2026-10-31T02:43:38.550Z

2. Amortized Localized Posterior Inference for Model Calibration with Application to an Ocean Turbulence Closure

中文摘要

摘要：模型差异可能导致单一的校准参数无法在整个情境域中重现物理响应。此外，观察和情境的不确定性会影响校准，必须加以考虑和量化。我们通过根据观测似然与目标情境的接近程度加权来构建一族局部后验分布。高斯局部化权重通过协方差膨胀引入情境不确定性，并在域边界进行反射，同时可选择性地修正情境采样不均。条件归一化流通过摊销变分推断来逼近这一族后验分布。在规定的情境测度下积分局部后验分布可得到边际参数分布。在使用Branin-Hoo模型的双胞实验中，我们比较了在均匀和偏斜情境分布下的全局和局部推断。与全局后验相比，局部化可降低报告的总体预测误差。在一个单柱洋流模型中，对k-$\epsilon$湍流闭合模型进行应用，其中大气表面强迫作为情境，结果表明局部后验校准相对于确定性和全局后验校准，可减少温度和经向速度误差。两个双胞实验都构造为允许局部补偿。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37731 (HTTP 429)

Authors: Exaucé Luweh Adjim Ngarti, Élise Arnaud, Léo Nicoletti, Victor Trappler, Arthur Vidard

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37731.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37731

Published: 2026-10-31T02:43:38.550Z

3. ThinkJev: One Thought, Every Budget for Calibrated Typed Decisions

中文摘要

摘要：语言模型的许多使用最终都以一种键入决策告终：在一组封闭选项中做出选择，这些选项的概率可以被代码操作。从一次前向传播中读取这些概率很快，但在需要多步骤的决策（如算术、计数和日期比较）中，经常会出错。让模型先进行思考可以修正大部分错误，但每次决策需要几秒钟而不是毫秒。为了只在有帮助的情况下付出思考代价，ThinkJev 从一个可以直接回答或经过思考后回答的开放混合思考模型中构建了一个快速系统和一个慢速系统。系统1在直接回答时读取模型对选项字母的概率；系统2在模型对相同缓存的提示进行思考后读取相同字母，因此升级重复不会产生额外工作，并且两个系统对相同选项给出概率。因为字母可以在任意数量的思考标记后读取，所以一次采样的思考还可以在每个思考预算下给出答案。无需训练，ThinkJev 仅在对标注开发数据的统计检验保证其接受的答案保持在目标错误率以下时，允许系统1单独回答。在5%的目标下，Qwen3.5-27B 以这种方式回答了8359道测试决策中的24.5%，实际错误率为3.9%（Qwen3.5-4B和9B分别为6.5%和0.0%）。对于其余情况，一旦答案自信就停止每次思考，其准确率达到81.6%，而总是思考的准确率为82.7%，思考标记消耗为72%。仅根据系统1的信心来选择何时思考节省较少，因为系统1最不确定的决策往往不是思考可以纠正的决策。在JevBench的公共项目上，这是一个独立的键入决策服务基准，Qwen3.5-27B 对29.4%的项目进行了思考，其准确率与列出的两个思考API没有显著差异（探索性）。代码和数据生成器可在此 https URL 获取。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37732 (HTTP 429)

Authors: Xiaolu Ding, Kai Yang, Dongdong Liu

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37732.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37732

Published: 2026-10-31T02:43:38.550Z

4. OpenFrontiers: A Collection of Open Research Problems and a Benchmark for Scientific Rediscovery

中文摘要

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37739 (HTTP 429)

Authors: Jiarui Yao, Ruida Wang, Tong Zhang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37739.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37739

Published: 2026-10-31T02:43:38.550Z

5. WebAnything: From the Wild Web to Scalable Agent Environments

中文摘要

摘要：Web代理需要多样化、可执行且可重置的网站环境，以实现可复现的训练和评估。人工辅助的镜像需要针对每个网站进行工程处理，而生成式构建可能偏离指定的真实网站。抓取与重放可以保留观察结果，但交互覆盖受限于捕获的内容和重放能力。我们提出了WebAnything，一个全自动框架，可从真实URL重建以源为基础、可重复使用的Web环境，无需在初始设置和授权后对每个网站进行手动实现或修复。Explorer-Builder收集浏览器可观察的结构、资源和交互结果；独立验证器检查冻结的候选对象，并指导本地修复或针对性地重新捕获源。我们构建了跨14个应用领域和53个类别的244个网站工件，支持12,984个执行并接受的任务。与人工辅助的WebHarbor相比，WebAnything在三项保真度指标上实现了更高的综合评分，在一次盲审中获得12次胜利、2次平局，且未出现失败。在5,856条交互轨迹上微调Qwen3.5-9B，使REAL成功率提高了9.9个百分点，Odysseys-Easy Rubric平均分和任务平均分分别提高6.25和6.32个百分点，WebArena-Verified整体成功率在固定测试环境下提高13.33个百分点。代码：https URL

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37786 (HTTP 429)

Authors: Tianci Xie, Zikun Cai, Chaoyun Zhang, Pinxue Guo, Bingxuan Li, Rui Zheng, Hang Yan

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37786.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37786

Published: 2026-10-31T02:43:38.550Z

6. What Does an Ontology Catch? Curated, LLM-Generated and Data-Derived Ontologies as Checkers of LLM Agents over Knowledge Graphs

中文摘要

摘要：针对网络知识图谱回答问题的LLM代理越来越多地被包装在源自图谱本体的检查机制中，假设正式的模式能够成为可靠的判断标准。然而，检查器也可能出错：它可能漏掉错误，也可能标记正确答案并推动代理去“修正”它们。作为这样一个判断者，人工策划的本体是否值得其成本，以及它与LLM生成的本体或从数据中挖掘的模式相比如何，目前尚不清楚。我们将检查器本身作为研究对象。在策划的、LLM生成的以及基于数据的本体中，无论是作为提示上下文、作为检查器，或两者兼有，以及在公开权重和专有代理上，在公共图谱和一个接近$10^{9}$三元组的私有图谱上，我们都考察了本体衍生检查捕获了什么错误以及错误标记了什么。我们展示了每种来源如何在捕获的错误与误报之间进行权衡，如何一个按出版形式使用的策划模式可能与自身数据相矛盾，结构性检查会漏掉哪些错误，以及合并来源或将检查器的具体反馈提供给代理是否有益。在UniProt上，策划的本体标记的正确答案比从数据中衍生的本体或LLM根据带掩码标识符的数据摘要生成的本体都要多。在实际代理的答案中，它的标记准确率仅比随机标记略高。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37809 (HTTP 429)

Authors: Michal Orzechowski, Rund Tawfiq, Tomasz Kosciolek

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37809.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37809

Published: 2026-10-31T02:43:38.550Z

7. Measuring Where Intent Gets Lost in Agentic Smart Contract Synthesis

中文摘要

摘要：AI 代理现在可以根据纯英文描述编写软件，但能运行的代码并不等同于能完成所要求操作的代码。这种差距在智能合约中尤为重要，智能合约是支撑 Web3 应用的程序，在区块链上转移资金，一旦部署就无法修改。我们探讨如何检查 AI 编写的合约是否按预期工作，以及请求的行为在何处丢失。我们设计了一个基于执行而非主观判断的评估方法。我们在 AI 编写的合约上重放相同的交易，并将其与原始合约应复现的结果进行比较，测量它们一致的频率。我们在特意植入错误的测试中测试每个评分工具，将结果与未经描述而编写的答案键进行比较，审计我们自己的评分器，直到已知的错误答案无法通过。常见方法（如 AI 裁判和标准分析工具）会遗漏明显错误，并且评分选择也很重要。仅检查所有系统包含的函数隐藏了管道之间约一半的差异。随后，我们提出了“先引导”的架构。在编写代码前，代理会向客户提出一份简短的问题清单，使用独立编写的可执行答案键检查合约，并将分歧返回给客户，而不是悄悄地重写代码。根据三个 AI 模型的实验结果，先询问比事后审查或修补重要得多，而自动化安全修复通常会破坏所要求的行为。一些精心挑选的问题能够恢复短描述遗漏的绝大部分内容，而且其价值可以在没有原始合约的情况下评估。据我们所知，这是智能合约生成领域首次在生成前收集缺失需求，并将结果与独立编写的可执行模型进行对比检查。这些经验可推广到任何 AI 代理将请求转化为程序的场景。我们发布了我们的工具、600 个任务及所有结果。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37830 (HTTP 429)

Authors: Abhinav Goel, Agostino Capponi, Alfio Gliozzo, Chaitya Shah

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37830.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37830

Published: 2026-10-31T02:43:38.550Z

8. Detecting Covert Collusion between AI Agents through Partner-Outsider Positive Listening

中文摘要

摘要：共享秘密约定的人工智能代理可以通过看似无害的信息进行协调，因此只读取信息的监控器可能会漏掉这些行为。我们提出了一种行为测试：使用代码的伙伴会对信息的字面措辞做出反应，而放在其位置的无关代理则不会。伙伴—外部观察者正向倾听（PoPLi）是两种代理在信息意义不变但措辞变化时，其动作变化程度的差异，这一指标通过事后重放记录的提示计算。在一个模拟市场中，有五种条件（无勾结、联合奖励且信息中性、相同条件下公开声明意图、秘密首字母代码、代码加上声明意图），当存在代码时，PoPLi介于0.48到0.59之间（95％区间均高于0.29），而在没有代码时介于-0.10到-0.01。仅观察信息的监控器无法区分这些条件：一个只读取发言者信息的监控器在所有50个市场中回答“否”，而一个读取每个卖家信息的监控器在10个无勾结市场中标记了9个。PoPLi检测的是私人通道的使用，而非勾结本身，也无法仅通过激励或传递信息意义的代码检测勾结。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37849 (HTTP 429)

Authors: Kevin Yandoka Denamganaï

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37849.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37849

Published: 2026-10-31T02:43:38.550Z

9. Evaluating and improving AI agents for sequencing protocol understanding

中文摘要

摘要：现有的科学代理基准很少评估代理是否能够重建测序协议指定的精确分子结构和生成工作流程，而这种能力对于分析定制测序实验至关重要。我们引入了 LibStructBench，这是一种基于文档的基准，用于重建测序文库结构和分子工作流程。我们的实验表明，前沿代理通常能够生成大致正确的重建结果，但在精确重建方面仍不可靠，尤其是在分子状态表示和工作流程连通性方面。切换底层语言模型产生的平均性能差异大于切换代理工具，且只有在分子状态重建方面差异具有统计学显著性。额外的知识在一个小型试点中改善了寡核苷酸家族的恢复，但未提高平均分子工作流程得分。程序规则和跨协议示例的自主积累显示了非单调的验证性能，并且在测试的轨迹中未提高平均保留的分子工作流程。这些发现凸显了在复杂实验协议中进行分子过程科学推理时，需要更可控和可验证的方法。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37851 (HTTP 429)

Authors: Chi-Lam Poon, Michael Goneos, Zezheng Song, Xi Chen

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37851.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37851

Published: 2026-10-31T02:43:38.550Z

10. Seq@K: Evaluating Language Models Under Sequential Feedback

中文摘要

摘要：语言模型代理与以往的智能系统在根本上不同之处在于其能够从上下文中的自然语言中学习。这种能力使代理能够根据反馈迭代地优化其答案，这是现代代理工作流程中普遍的操作模式。然而，社区主要通过诸如 Pass@1 或 Pass@K 的指标对模型进行排序，这无法捕捉模型如何利用对先前尝试的反馈。我们引入了 Seq@K，一种自然的评估协议，用于衡量模型是否能够在 K 次连续尝试内解决问题，并在每次先前失败后获取反馈。与考察长周期规划的多轮代理基准不同，Seq@K 分离并形式化了从反馈中进行改进的核心能力。我们在 7 个基准和 16 个前沿模型上评估了 Seq@K，结果表明 Seq@K 揭示了一个实质上不同的模型选择格局：在 Seq@K 下，(1) 模型排名往往不同，(2) 帕累托前沿允许更高效的配置，(3) 选择最优模型配置所带来的收益相对于 Pass@K 可能扩大或缩小。我们进一步展示了在 Seq@K 下分析模型轨迹有助于刻画迭代改进行为，例如模型解决反馈中提出问题的速率，或它们对自身答案的坚持程度。最后，我们研究了反馈性质如何影响模型的迭代改进，表明在设计和评估迭代代理系统的反馈循环时，实践者必须谨慎。综合来看，这些结果确立了 Seq@K 如何为通常在此类反馈循环中运行的智能系统提供更准确的衡量标准，以及在评估此类系统模型时考虑 Seq@K 的必要性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37859 (HTTP 429)

Authors: Thomson Yen, Gilbert Yang, Naimeng Ye, Minghui Chen, Dong Xia, Tianyi Shen, Yechan Jeon, Yaqin Chen, Daksh Mittal, Fangyu Wu, Hongseok Namkoong

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.37859.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37859

Published: 2026-10-31T02:43:38.550Z

VLM Domain Papers

1. Sparse tensor networks for individual tree diameter estimation in mixed forests

Abstract:Measurements of tree diameter at breast height (DBH) are fundamental to forest inventories, but manual surveys across entire stands or regions are impractical. UAV-borne laser scanning (ULS) enables scalable automated forest surveys, yet canopy occlusion makes accurate DBH estimation from ULS point clouds non-trivial. Limited stem-level point density of ULS data constrains approaches relying on detailed stem reconstruction, such as Quantitative Structure Models (QSMs) and circle fitting. Empirical machine learning methods based on predefined point cloud metrics are an alternative. New deep learning approaches offer an even more principled solution, allowing models to learn directly from point clouds. Sparse tensor networks have shown particular promise, transforming forestry tasks such as tree species classification. We investigate whether these strengths can be extended to estimate individual tree DBH from ULS data across four mixed-species temperate forest plots in Germany. We further advance existing sparse tensor networks by introducing TreeNext, a lightweight deep learning architecture for direct individual tree parameter estimation from voxelized LiDAR point clouds. We present a fully automated forest inventory pipeline from individual tree instance segmentation to DBH regression, and quantify the impact of segmentation errors by comparing performance on automatically segmented and ground truth instances. TreeNext achieves an $R^2$ of 0.8355 on automatically segmented trees and 0.9380 on ground truth segments, outperforming strong tabular baselines including Random Forest and XGBoost. Segmentation quality emerges as the dominant error source in the automated pipeline. TreeNext proves highly data-efficient, performing well even in small data scenarios. Our findings establish sparse tensor networks as a practical foundation for automated individual tree DBH estimation from ULS.

中文摘要

摘要：测量树木胸径（DBH）是森林清查的基础，但在整个林分或区域进行人工调查是不现实的。无人机搭载激光扫描（ULS）能够实现可扩展的自动化森林调查，但由于树冠遮挡，从ULS点云中准确估计DBH并非易事。ULS数据在单株树层面的点密度有限，限制了依赖详细树干重建的方法，如定量结构模型（QSM）和圆形拟合。基于预定义点云指标的经验机器学习方法是一种替代方案。新的深度学习方法提供了更为系统的解决方案，允许模型直接从点云中学习。稀疏张量网络显示出特别的潜力，可用于转变森林相关任务，例如树种分类。我们研究了这些优势是否可以扩展到从德国四个混交温带林地的ULS数据中估计单株树的DBH。我们进一步通过引入TreeNext推进现有的稀疏张量网络，这是一个用于从体素化LiDAR点云直接估计单株树参数的轻量级深度学习架构。我们提出了一个从单株树实例分割到DBH回归的全自动森林清查流程，并通过比较自动分割和实测实例的性能来量化分割误差的影响。TreeNext在自动分割树木上实现了$R^2$为0.8355，在实测分割上为0.9380，优于包括随机森林和XGBoost在内的强大表格基线。分割质量成为自动化流程中主要的误差来源。TreeNext表现出高度的数据效率，即使在小数据场景下也表现良好。我们的研究结果确立了稀疏张量网络作为基于ULS的自动单株树DBH估计的实用基础。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37784 (HTTP 429)

Authors: Clara Lößl, Fabian Ewald Fassnacht, Viacheslav Barkov

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.37784.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37784

Published: 2026-10-31T02:43:58.815Z

2. GOFish: Full Field-of-View Gaussian Splatting and Mesh Extraction

Abstract:Robots, autonomous vehicles, and consumer 360° cameras use fisheye lenses for wide fields of view (FoVs). Simulation-ready digital twins require photorealistic rendering across the full FoV and an aligned collision mesh. However, prior fisheye Gaussian Splatting methods only rasterize up to 180° and do not produce meshes, while Gaussian surface reconstruction methods assume pinhole training images. We present GOFish, the first 3D Gaussian Splatting surface reconstruction method to train directly on fisheye images, rasterize FoVs beyond 180°, and extract an aligned mesh from the same scene. Calibrated densification recovers the surface across the entire FoV, while Calibrated Mip Filtering and Wide-Angle Appearance Modeling keep pixel blur and lens falloff out of the scene and mesh. Across the 200° FIORD and 205° FullCircle benchmarks, GOFish improves PSNR over the best prior method by 2.28 and 1.97 dB, respectively. It also outperforms fisheye rendering baselines within their own FoVs and achieves the highest mesh F-score on FIORD and the main ScanNet++ set. Together, the Gaussian scene and mesh form a LiDAR-free, simulation-ready digital twin that renders in real time, ray traces the full FoV, and provides collision geometry.

中文摘要

摘要：机器人、自动驾驶车辆和消费者360°相机使用鱼眼镜头来获得广阔的视场（FoV）。用于仿真的数字孪生需要在整个视场内进行逼真的渲染，并生成对齐的碰撞网格。然而，现有的鱼眼高斯点绘制方法仅能栅格化到180°，且无法生成网格，而高斯表面重建方法则假设针孔训练图像。我们提出了GOFish，这是第一个能够直接在鱼眼图像上进行训练的3D高斯点绘制表面重建方法，可栅格化超过180°的视场，并从同一场景中提取对齐的网格。经过校准的密化方法恢复了整个视场的表面，而校准的Mip滤波和广角外观建模将像素模糊和镜头衰减排除在场景和网格之外。在200°的FIORD和205°的FullCircle基准测试中，GOFish分别比之前最佳方法提升了2.28和1.97 dB的PSNR。它还在各自的视场内超越了鱼眼渲染基线，并在FIORD和主要的ScanNet++数据集上实现了最高的网格F-score。高斯场景和网格结合在一起，形成了无需LiDAR的、可用于仿真的数字孪生，能够实时渲染、对整个视场进行光线追踪，并提供碰撞几何信息。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37818 (HTTP 429)

Authors: Allen Tu, Meher Gitika Karumuri, Shifan Zhu, Hailing Wang, Yen-Liang Lin, David Held, Matthias Zwicker, Tom Goldstein

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.37818.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37818

Published: 2026-10-31T02:43:58.815Z

3. Convex Potential Learning for Optimal Transport Between Medical Images

Abstract:Computing optimal transport maps between medical images plays a fundamental role in morphological analysis. However, existing methods are either difficult to scale to large 3D medical images or rely on regularized formulations that provide only approximate transport maps. To improve the accuracy of downstream medical image analysis, we propose a novel approach for estimating optimal transport maps by parameterizing the convex potential function. By construction, the resulting transport map is guaranteed to be curl-free, consistent with Brenier’s theorem. We evaluate the proposed method on both 2D and 3D medical imaging datasets. Experimental results demonstrate that our approach achieves lower transport estimation errors and produces higher-quality image reconstructions than existing methods.

中文摘要

摘要：在医学图像之间计算最优传输映射在形态学分析中起着基础性作用。然而，现有方法要么难以扩展到大型三维医学图像，要么依赖于仅能提供近似传输映射的正则化形式。为了提高下游医学图像分析的准确性，我们提出了一种通过参数化凸势函数来估计最优传输映射的新方法。通过该构造，得到的传输映射保证无旋度，符合Brenier定理。我们在二维和三维医学影像数据集上评估了所提出的方法。实验结果表明，与现有方法相比，我们的方法在传输估计误差更低，同时生成的图像重建质量更高。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37844 (HTTP 429)

Authors: Hongyu Kan, Ivan Medri, Natasha Ironside, Kristofor Pas, Richard Corp, Shinjini Kundu, Gustavo Kunde Rohde

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.37844.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37844

Published: 2026-10-31T02:43:58.815Z

4. Effects of Deep Learning Architecture and Pretraining on Automated Vocal Fold Segmentation in Ex Vivo High-Speed Phonation Images

Abstract:Two-dimensional (2D) intraglottal particle image velocimetry (PIV) of ex vivo phonation experiments quantifies vocal fold medial surface dynamics and the intraglottal flow field. However, it relies on manual segmentation because residual glare, low tissue-to-flow contrast, and tracer-particle speckle obscure the vocal fold boundaries. This study assessed how deep learning architecture and ImageNet pretraining affect automated vocal fold segmentation in these high-speed images. Four established architectures from three categories were benchmarked: the convolutional neural network (CNN)-based UNet and UNet++, the hybrid CNN-transformer-based TransUNet, and the pure transformer-based Swin-UNet, the latter two trained from scratch and from ImageNet-pretrained weights. Using 16,344 expert-annotated images from 227 phonation trials of excised canine larynges, the six pipelines were trained with five seeds under a standardized protocol and compared on 50 held-out test trials using trial-level Wilcoxon signed-rank tests with Holm correction. Mean Dice similarity coefficient (DSC) spanned only 0.9734 to 0.9764, so the pipelines differed mainly in boundary fidelity. Pretraining significantly improved the boundary F1 score (BF1) and 95th percentile Hausdorff distance (HD95) of the TransUNet and Swin-UNet, about four times more for the Swin-UNet. Once pretrained, the two performed similarly, although the Swin-UNet has only about a quarter of the trainable parameters. Without pretraining, the TransUNet had a significantly higher BF1 and a lower HD95 than both CNN-based pipelines and about half their share of images with HD95 > 50 pixels. To our knowledge, this is the first study to apply deep learning to vocal fold segmentation in 2D intraglottal PIV, and it offers a reference framework for segmenting experimental flow-visualization images with noisy optical conditions or highly deformable objects.

中文摘要

摘要：二维（2D）声门内粒子图像测速（PIV）在离体发声实验中量化声带内侧表面动力学和声门内流场。然而，由于残余眩光、组织与流动对比度低以及示踪颗粒斑点掩盖了声带边界，它依赖于手动分割。本研究评估了深度学习架构和 ImageNet 预训练如何影响这些高速图像中自动声带分割的效果。来自三类的四种已建立架构进行了基准测试：基于卷积神经网络（CNN）的 UNet 和 UNet++，混合 CNN-Transformer 的 TransUNet，以及纯 Transformer 的 Swin-UNet，后两者分别从零训练和使用 ImageNet 预训练权重进行训练。使用来自 227 次去除犬喉声带试验的 16,344 张专家标注图像，这六个流水线在标准化协议下用五个随机种子进行训练，并在 50 个保留测试试验上通过采用 Holm 校正的试验级 Wilcoxon 符号秩检验进行比较。平均 Dice 相似系数（DSC）仅在 0.9734 到 0.9764 之间，因此各流水线的主要差异体现在边界精度上。预训练显著提高了 TransUNet 和 Swin-UNet 的边界 F1 分数（BF1）和 95 百分位 Hausdorff 距离（HD95），而 Swin-UNet 的提升约为其四倍。一旦经过预训练，这两者的表现相似，尽管 Swin-UNet 的可训练参数仅约为前者的四分之一。在未预训练的情况下，TransUNet 的 BF1 显著高于两种基于 CNN 的流水线，其 HD95 也更低，并且 HD95 > 50 像素的图像比例约为它们的一半。据我们所知，这是首个将深度学习应用于二维声门内 PIV 中声带分割的研究，并为在光学条件噪声较大或对象高度可变形的实验流动可视化图像分割提供了参考框架。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37860 (HTTP 429)

Authors: Varun Vinayak Kalaiarasan, Charles Farbos de Luzan, Khalid Malik, Ephraim Gutmark, Liran Oren

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.37860.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37860

Published: 2026-10-31T02:43:58.815Z

5. PairCLIP-SWA: Similarity-Guided Balancing and Stochastic Weight Averaging for Generalized Deepfake Image Detection with OpenCLIP

Abstract:Deepfake detectors are almost always deployed as the single checkpoint that scored best on a validation split, yet the reliability of that choice is rarely measured. We measure it with PairCLIP-SWA, a single-backbone detector that adapts an OpenCLIP ViT-L/14 visual tower by progressive block unfreezing on a class-balanced corpus built by CLIP-similarity-guided one-to-one subsampling, and averages the fine-tuning trajectory in weight space. Trained in-domain on the IEEE SP Cup 2025 (DF-Wild-Cup) benchmark under the same protocol as the strongest published three-backbone ensemble, the weight-averaged detector reaches 97.04% accuracy, 98.64% AU-ROC and a 4.07% equal error rate, against 94.46%, 97.60% and 5.00%. Its training log shows why averaging matters: over the averaged epochs the non-averaged checkpoint swings across 5.34 points of validation accuracy while its validation loss more than doubles, whereas the average is never worse and is up to 5.99 points better. On an OpenForensics-derived benchmark the same instability becomes a misranking: the checkpoint that validation accuracy selects (98.43%) scores 88.85% on a disjoint 10,905-image test partition, rejecting 21.1% of genuine photographs, while the terminal average, ranked last on validation, scores 94.76%. A study of twenty-two in-house configurations on the identical DF-Wild-Cup split, including preprocessing ablations, shows that CLIP pre-training rather than ensembling drives accuracy, that secondary face cropping costs about 34 points, and that frequency-domain inputs transfer poorly. We recommend per-class selection criteria measured off the early-stopping split.

中文摘要

摘要：深度伪造检测器几乎总是以在验证集上得分最高的单个检查点部署，但这种选择的可靠性却很少被衡量。我们使用 PairCLIP-SWA 进行衡量，这是一种单骨干检测器，通过在类平衡语料库上逐步解冻块来适应 OpenCLIP ViT-L/14 视觉塔，该语料库通过 CLIP 相似性指导的一对一子采样构建，并在权重空间中对微调轨迹进行平均。在 IEEE SP Cup 2025（DF-Wild-Cup）基准上使用与最强已发布三骨干集成相同的协议进行域内训练后，经权重平均的检测器达到了 97.04% 的准确率、98.64% 的 AU-ROC 和 4.07% 的等错误率，而未平均检测器分别为 94.46%、97.60% 和 5.00%。其训练日志显示了平均化的意义：在平均的 epoch 中，未平均的检查点在验证准确率上游动了 5.34 个百分点，而其验证损失增加了一倍多，而平均值从未更差，且最高可好 5.99 个百分点。在 OpenForensics 衍生基准上，同样的不稳定性导致错误排名：验证准确率选出的检查点（98.43%）在一个不相交的 10,905 张图像测试分区上的得分为 88.85%，拒绝了 21.1% 的真实照片，而在验证中排名最后的最终平均分为 94.76%。在相同的 DF-Wild-Cup 切分上对二十二种内部配置（包括预处理消融）进行的研究表明，是 CLIP 预训练而非集成推动了准确率，二次人脸裁剪损失约 34 个百分点，频域输入的迁移性较差。我们建议采用基于早停分割测量的每类选择标准。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37874 (HTTP 429)

Authors: Md. Nakib, Sudipto Sarkar, A. J. A. Jubayer Talukder, Mohammad Ariful Haque

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.37874.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37874

Published: 2026-10-31T02:43:58.815Z

6. Intersectional Invariant Learning for Fair Deepfake Detection

Abstract:Deepfake detection models achieve high overall accuracy but exhibit large performance disparities across demographic groups, causing some groups to experience substantially higher false-positive rates than others. Existing fairness-aware methods address this by constraining model outputs to equalize group predictions, but do not tackle the underlying feature-level bias. As a result, models may still rely on demographic characteristics, such as skin texture, as indirect cues for detecting manipulated content. To address this problem, we propose Intersectional Invariant Learning (IIL), a framework that improves feature learning rather than only adjusting model outputs. IIL alternates between identifying the most challenging demographic partitions and learning features that remain consistent across those partitions. The framework also includes a false-positive-rate fairness objective and group-balanced sampling across six gender x skin-tone intersectional groups. We evaluate IIL on a large-scale fairness benchmark containing over one million images, 10 baseline detectors, and 21 fairness and utility metrics. Results show that IIL achieves Pareto-optimal performance, simultaneously attaining an AUC above 99%, a false-positive rate under 2.5%, and the lowest intersectional fairness disparity (FMEO) among all compared methods. IIL outperforms prior fairness-aware methods on the large majority of the 21 evaluation metrics, while reducing both false-positive rates and intersectional fairness disparities by several-fold compared with the strongest competing method.

中文摘要

摘要：深度伪造检测模型总体上可以达到较高的准确率，但在不同人口群体之间表现出较大的性能差异，导致某些群体的假阳性率显著高于其他群体。现有的公平性感知方法通过限制模型输出以平衡群体预测来应对这一问题，但并未解决潜在的特征层偏差。因此，模型可能仍然依赖于人口特征，例如皮肤纹理，作为检测篡改内容的间接线索。为了解决这一问题，我们提出了交叉不变学习（Intersectional Invariant Learning, IIL）框架，该框架侧重于改进特征学习，而不仅仅是调整模型输出。IIL在识别最具挑战性的人口分区和学习在这些分区上保持一致的特征之间交替进行。该框架还包括假阳性率公平目标以及针对六个人口交叉群体（性别 x 肤色）的平衡采样。我们在一个包含超过一百万张图像、10个基准检测器以及21个公平性和效用指标的大规模公平性基准上对IIL进行了评估。结果显示，IIL实现了帕累托最优性能，同时达到AUC超过99%、假阳性率低于2.5%，以及所有比较方法中最低的交叉群体公平性差异（FMEO）。在21个评估指标中的绝大多数指标上，IIL都优于已有的公平性感知方法，同时与最强竞争方法相比，大幅降低了假阳性率和交叉群体公平性差异。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37881 (HTTP 429)

Authors: Uzoamaka Ezeakunne, Taiwo Olabamipe, Deevesh Jain, Xiuwen Liu

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.37881.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37881

Published: 2026-10-31T02:43:58.815Z

7. FYC: FPR-Disparity Weight Perturbation and Youden Calibration for Fair Deepfake Detection

Abstract:Deepfake detectors routinely achieve high aggregate accuracy while exhibiting substantial demographic disparities in their errors: false-positive rates differ by double digits across gender and skin tone. Most existing remedies require either full retraining with a fairness-aware objective or demographic-annotated training data, both impractical when only a pretrained checkpoint is available, while label-free post-hoc calibrators leave the detector’s weights unchanged. We introduce FYC (FPR-disparity Weight Perturbation and Youden Calibration), a two-stage post-hoc method that perturbs a detector’s weights under an FPR-disparity criterion, then calibrates a separate decision threshold for each demographic group - requiring no architectural changes and no full retraining. Evaluated against every detector benchmarked on AI-Face v2 across six intersectional gender x skin-tone groups, FYC outperforms the field on 18 of 21 reported fairness and utility metrics, including purpose-built fairness architectures, at a fraction of the training cost. We further identify why the choice of fairness criterion is critical: the original Demographic Parity objective used in prior weight-perturbation work is architecture-dependent and unpredictable, improving two of four evaluated architectures, leaving a third largely unchanged, and substantially degrading the fourth (PG-FDD), with no way to anticipate the outcome for a given checkpoint in advance. Conditioning the same objective on real images only - our FPR-disparity criterion - eliminates this unpredictability, improving fairness and accuracy jointly across all four architectures with no exceptions.

中文摘要

摘要：深度伪造检测器通常在整体准确率上表现很高，但在错误上存在显著的人口统计差异：不同性别和肤色的假阳性率相差达两位数。大多数现有的补救措施要么需要使用公平性意识目标进行完整的再训练，要么需要带有人口统计标注的训练数据，而在只有预训练检查点可用时，这两种方法都不现实；而无标签的事后校准方法则不会改变检测器的权重。我们引入了FYC（FPR差异权重扰动与Youden校准），这是一种两阶段的事后方法，首先根据FPR差异准则扰动检测器的权重，然后为每个人口统计组校准独立的决策阈值——无需改变架构，也无需完整再训练。在针对AI-Face v2上六个性别×肤色交叉群体的每个检测器基准测试中，FYC在报告的21个公平性和效用指标中有18项表现优于现有方法，包括专为公平性设计的架构，并且训练成本仅为其一小部分。我们进一步指出了为什么公平性准则的选择至关重要：先前权重扰动研究中使用的原始人口统计平等目标依赖于架构且不可预测，它改善了四个评估架构中的两个，使第三个基本保持不变，并显著降低了第四个（PG-FDD）的表现，且无法预先预测特定检查点的结果。仅在真实图像上调整相同目标——即我们的FPR差异准则——消除了这种不可预测性，使四个架构的公平性和准确性同时得到提升，没有例外。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37882 (HTTP 429)

Authors: Uzoamaka Ezeakunne, Taiwo Olabamipe, Deevesh Jain, Xiuwen Liu

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.37882.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37882

Published: 2026-10-31T02:43:58.815Z

8. Venus-DefakerOne2: A Vision-Forensics Foundation Model with Broader Coverage

Abstract:Real-world forgery is inherently heterogeneous: it manifests not only as digital manipulation but also as physical spoofing, and is carried not only by images but also by videos. Despite substantial progress, a unified anti-forgery foundation model covering such a broad threat spectrum remains absent. In this paper, we present Venus-DefakerOne2, which substantially extends the forensic capability boundary along both the forgery-form and media-type dimensions. Trained on tens of millions of open-source and in-house samples, it incorporates two key designs. Tide-Style-Tuning alternates LoRA and full-parameter fine-tuning to balance computational efficiency, generalization, and hard-sample fitting while mitigating catastrophic forgetting. Entropy-based Test Adaptation determines the optimal prediction modality by analyzing prediction entropy. Venus-DefakerOne2 achieves new state-of-the-art performance across a comprehensive suite of benchmarks, including DeepFake faces (96.7%), AI-generated images (93.3%), tampered documents (87.6%), manipulated natural images (89.6%), physical spoofs (93.1%), and AI-generated videos (96.0%). It significantly outperforms both specialized expert models and its predecessor, DefakerOne, establishing itself as a powerful and versatile foundation model for binary-decision forensic tasks.

中文摘要

摘要：现实世界中的伪造本质上是异质的：它不仅表现为数字篡改，还包括物理欺骗，而且既存在于图像中，也存在于视频中。尽管已有大量进展，但覆盖如此广泛威胁范围的统一反伪造基础模型仍然缺失。在本文中，我们提出了 Venus-DefakerOne2，它在伪造形式和媒体类型两个维度上显著扩展了取证能力边界。该模型在数千万开源和内部样本上训练，包含两个关键设计。潮汐风格调优（Tide-Style-Tuning）交替使用 LoRA 和全参数微调，以平衡计算效率、泛化能力以及困难样本拟合，同时缓解灾难性遗忘。基于熵的测试适应（Entropy-based Test Adaptation）通过分析预测熵确定最佳预测模式。Venus-DefakerOne2 在包括 DeepFake 面孔（96.7%）、AI 生成图像（93.3%）、篡改文档（87.6%）、操作过的自然图像（89.6%）、物理欺骗（93.1%）以及 AI 生成视频（96.0%）在内的综合基准测试中取得了新的最先进性能。它显著优于专门的专家模型和其前身 DefakerOne，确立了作为二元决策取证任务的强大且多功能基础模型的地位。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37918 (HTTP 429)

Authors: Chenfan Qu, Yachao Liang, Mingqi Fang, Jiangwei Xie, Zhenming Wang, Baoyue Zhang, Changjiang Jiang, Zheng Chang, Fanli Jin, Xuekang Zhu, Kaiwen Feng, Junchi Li, Weihong Zhang, Haoyue Wang, Chenfeng Zhang, Yinan Xia, Fangming Cui, Dong Fan, Jingjing Liu, Tianyi Zhang, Luobin Chen, Ji-Zhe Zhou, Sheng Li, Wei Wang, Bin Li, Ran He, Xinpeng Zhang, Linna Zhou, Jian Liu, Weiqiang Wang

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.37918.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37918

Published: 2026-10-31T02:43:58.815Z

9. You Don’t Need Aligned Representations: Knowledge Distillation via Random Prototype Spaces

Abstract:A widely held assumption in knowledge distillation is that teacher and student representations must occupy a compatible or explicitly aligned feature space. We propose RPKD (Random Prototype Knowledge Distillation), which discards this assumption by comparing both networks against a shared bank of randomly initialized, frozen prototype vectors. Alignment is therefore carried out in an external reference space rather than in the teacher’s native feature coordinates: the teacher side is projected by fixed random maps that carry no trainable parameters, and only the student adapts. Two branches handle the transfer: a logit branch aligning global prototype similarity distributions and a feature branch matching spatially-resolved prototype activation maps. Both operate within the same fixed prototype space, which can be interpreted as a random feature embedding that approximately preserves inter-sample similarity structure, so the prototype bank and the distillation objective stay unchanged across teacher-student pairs while only the input projections are instantiated per architecture. In low-category settings, decoupling the prototype vocabulary from the task label space yields richer supervision than the label space alone, an advantage absent from logit-based methods, whose supervisory signal collapses as class count shrinks. This advantage is especially relevant in real-world scenarios where the label space is typically constrained. Extensive experiments demonstrate the effectiveness of RPKD, achieving a maximum gain of +7.98% over OFA on CIFAR-100, +12.44% on ImageNet-100, and +1.28% on ImageNet-1K. Source code is available at this https URL

中文摘要

摘要：知识蒸馏中一个广泛持有的假设是教师和学生的表示必须占据兼容或明确对齐的特征空间。我们提出了 RPKD（随机原型知识蒸馏），通过将两个网络与一个共享的、随机初始化并冻结的原型向量库进行比较，从而舍弃了这一假设。因此，对齐是在外部参考空间中进行的，而不是在教师的原生特征坐标中：教师端通过不包含可训练参数的固定随机映射进行投影，只有学生端进行适应。两个分支处理知识迁移：一个是对数分支，对齐全局原型相似性分布；另一个是特征分支，匹配空间分辨的原型激活图。两者都在相同的固定原型空间内操作，该空间可以被解释为一个随机特征嵌入，大致保持了样本间相似性结构，因此原型库和蒸馏目标在教师-学生对之间保持不变，而只有输入投影根据架构实例化。在低类别设置下，将原型词汇与任务标签空间解耦，能够提供比单纯标签空间更丰富的监督，这是基于对数的方法所不具备的优势，因为随着类别数量的减少，其监督信号会崩塌。这一优势在标签空间通常受限的现实场景中特别重要。大量实验证明了 RPKD 的有效性，在 CIFAR-100 上比 OFA 提升最大 +7.98%，在 ImageNet-100 上提升 +12.44%，在 ImageNet-1K 上提升 +1.28%。源代码可在此 https URL 获得。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37937 (HTTP 429)

Authors: Hamza Etcibasi, Ramazan Gokberk Cinbis

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.37937.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37937

Published: 2026-10-31T02:43:58.815Z

10. Meta-Initialization of Implicit Neural Representations to Accelerate Deformable Medical Image Registration

Abstract:Implicit neural representation (INR) methods for deformable medical image registration require the network parameters to be optimized for each new image pair, making adaptation speed a practical limitation. This work investigates whether early adaptation can be accelerated using an initialization learned from a collection of previously solved registration tasks. We examine two schemes: Reptile meta-initialization and STRAINER transfer of the shared component of a coordinate network. KAN-IDIR is used as the base INR displacement-field model. Experiments are conducted on OASIS-1, with 364 consecutive pairs used for training and 49 pairs for test-time adaptation. Registration quality is evaluated using mean Dice and HD95, together with the fraction of voxels with a negative Jacobian determinant at iteration 1500. In the configurations studied, both schemes yield higher Dice and lower HD95 at early iterations than random initialization. Reptile numerically outperforms STRAINER under a small iteration budget, although their pretraining budgets are not matched. As the number of iterations increases, the differences between the schemes shrink. STRAINER pretraining takes 39-40 minutes, compared with 44.98 hours for Reptile.

中文摘要

摘要：用于可变形医学图像配准的隐式神经表示（INR）方法需要对每个新的图像对优化网络参数，使得适应速度成为实际限制。本文研究是否可以通过从一系列已解决的配准任务中学习到的初始化来加速早期适应。我们考察了两种方案：Reptile 元初始化和 STRAINER 共享坐标网络组件的迁移。KAN-IDIR 被用作基础的 INR 位移场模型。实验在 OASIS-1 上进行，使用 364 对连续图像用于训练，49 对图像用于测试时适应。配准质量通过平均 Dice 指数和 HD95 评估，同时计算第 1500 次迭代时负 Jacobian 行列式体素的比例。在研究的配置中，两种方案在早期迭代阶段均比随机初始化获得更高的 Dice 值和更低的 HD95。Reptile 在小迭代预算下在数值上优于 STRAINER，尽管它们的预训练预算不匹配。随着迭代次数的增加，两种方案之间的差异缩小。STRAINER 的预训练耗时 39-40 分钟，而 Reptile 为 44.98 小时。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.37965 (HTTP 429)

Authors: Marat O. Zinovev, Dmitry V. Sorokin

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.37965.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.37965

Published: 2026-10-31T02:43:58.815Z