ArXiv Domain 2026-10-07

数据来源：ArXiv Domain

LLM Domain Papers

1. Zero-Shot Visualization: Exploring Text Corpora with User-Prompted Axes

Abstract:We study the application of large language models (LLMs) to the visual exploration of textual corpora. We introduce zero-shot visualization (ZSV), a task in which users specify concepts in natural language and documents are mapped onto the corresponding concept axes for visualization. Building a ZSV system of practical value is non-trivial, as it requires choices at the intersection of feature functions, efficient implementation tradeoffs, and pre/post-processing decisions affecting visualization quality. To that end, we establish a benchmark that compares methods spanning embedding similarity, direct semantic judgments, and conditional likelihood estimation in this setting. Across multiple datasets and use cases we evaluate the properties of different scoring methods and design choices in terms of semantic faithfulness, score fidelity, and computational cost. Our results identify that scoring based on next-token probabilities offers the strongest practical trade-off among the evaluated methods. We further apply this approach to unlabeled corpora to examine its behavior in realistic exploratory settings. These experiments highlight additional design considerations, including the use of graded axes together with binary relevance filtering, and reveal a compositional sentiment bias in off-topic documents. Based on these findings, we provide practical guidelines for constructing end-to-end ZSV baselines.

中文摘要

摘要：我们研究了大型语言模型（LLM）在文本语料库可视化探索中的应用。我们引入了零样本可视化（ZSV），这是一个用户用自然语言指定概念，并将文档映射到相应概念轴进行可视化的任务。构建一个具有实际价值的ZSV系统并非易事，因为它需要在特征函数、有效实现权衡以及影响可视化质量的前/后处理决策之间做出选择。为此，我们建立了一个基准，比较在此设置下跨越嵌入相似性、直接语义判断和条件似然估计的方法。在多个数据集和使用场景中，我们评估了不同评分方法和设计选择在语义忠实度、评分保真度和计算成本方面的特性。我们的结果表明，基于下一个词概率的评分在所评估的方法中提供了最强的实际折衷。我们进一步将这一方法应用于无标签语料，以考察其在真实探索性场景中的表现。这些实验凸显了额外的设计考虑，包括将分级轴与二元相关性筛选结合使用，并揭示了离题文档中的复合情感偏差。基于这些发现，我们提供了构建端到端ZSV基线的实用指南。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06889

Authors: Arnau Bueno Tricas, Jose A. Rodríguez-Serrano

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.06889.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06889

Published: 2026-10-07T02:41:41.595Z

2. Capacity, Responsiveness and Alignment: What Makes a Latent Structure Actionable

Abstract:Localizing latent structures in the activation space of language models (LMs) is central to understanding and controlling their behavior. Yet, localized structures can differ substantially in their causal influence, raising the question of what makes a structure actionable. We tackle this question by casting causal influence as a product of three factors and showing empirically that they act as interpretable, distinct constraints: capacity, measuring the sensitivity of the model’s output to movement along the structure, responsiveness, capturing how promotable the concept is given the current context, and alignment, reflecting how well the structure aligns with the context-specific representation of the concept. Across 4 LM families and 50 concepts, we observe that causal effectiveness requires all factors to be high; low capacity and responsiveness reduce it by 84% and 95%, respectively, while low alignment can reverse it, suppressing concept expression. Moreover, we find that causality is context-dependent rather than an intrinsic property of the structure, with causally effective directions forming a low-dimensional subspace that varies across contexts. By restricting the training of linear probes to this subspace, we introduce causal probes that achieve 17%-118% improvement in steering across models, with only 3% reduction in concept detection.

中文摘要

摘要：在语言模型（LM）的激活空间中定位潜在结构，对于理解和控制其行为至关重要。然而，局部化结构的因果影响可能差异很大，这引出了一个问题：是什么使得某个结构具有可操作性。我们通过将因果影响作为三个因素的乘积来解决这个问题，并通过实验证明，它们作为可解释的、不同的约束起作用：容量，衡量沿结构移动时模型输出的敏感性；响应性，捕捉在当前上下文下概念可被激发的程度；对齐性，反映结构与概念在特定上下文表示之间的契合程度。在4个LM家族和50个概念中，我们观察到因果有效性要求所有因素都很高；低容量和低响应性分别将其降低84%和95%，而低对齐性可能会逆转因果作用，抑制概念表达。此外，我们发现因果性是依赖上下文的，而非结构的内在属性，因果有效的方向形成了一个随上下文变化的低维子空间。通过将线性探针的训练限制在这个子空间中，我们引入了因果探针，实现了在各模型中引导能力提升17%-118%，而概念检测仅降低3%.

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06897

Authors: Or Shafran, Mor Geva

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.06897.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06897

Published: 2026-10-07T02:41:41.595Z

Abstract:Retrieval-augmented generation grounds language models in external context, but for long documents flat top-$k$ retrieval can cluster on a single region and miss complementary evidence. RAPTOR-style summary trees address this by recursively clustering chunks and using a language model to summarize each cluster at indexing time, then ranking summary nodes alongside raw chunks at query time. We show the main benefit of summary trees in long-document QA can come from navigation rather than the generated summary content. We introduce NavTree, a leaves-only retriever that builds a deterministic balanced segment tree over chunks (zero language-model calls at indexing) and uses the tree purely as a navigation scaffold: a hybrid lexical-and-dense frontier walk, anchored on top retrieved leaves, descends from the root and emits only leaf chunks to the reader. On a matched-cost evaluation against flat retrievers and an extractive re-implementation of RAPTOR, NavTree is the strongest matched-cost hierarchical retriever in our evaluated grid and ties the strongest flat baseline. On long-document multi-hop QA, it is the only hierarchical method that significantly beats BM25 on a class-vs-class basis, corroborated by a reader-free retrieval-recall check. A matched-reader replication of the published abstractive RAPTOR variant, given strong cluster summaries, still loses to NavTree at every multi-chunk budget, at zero indexing cost. The ranking carries across stronger and open-weight readers, a stronger encoder, and a full factorial that isolates leaves-only emission as the structural lever.

中文摘要

摘要：检索增强生成使语言模型能够基于外部上下文，但对于长文档，平面 top-$k$ 检索可能会聚焦于单个区域，从而错过互补证据。RAPTOR 风格的摘要树通过递归聚类文档块，并在索引阶段使用语言模型对每个聚类进行摘要，在查询时将摘要节点与原始文档块一起排序，从而解决了这一问题。我们展示了在长文档问答中，摘要树的主要优势可能来自于导航，而不是生成的摘要内容。我们提出了 NavTree，一种仅使用叶节点的检索器，它在文档块上构建确定性平衡分段树（索引阶段无需调用语言模型），并将该树纯粹作为导航框架：从顶端检索叶节点出发，进行混合的词汇和密集边界遍历，从根节点向下，仅向阅读器输出叶节点文档块。在与平面检索器及 RAPTOR 的抽取式重实现的匹配成本评估中，NavTree 是我们评估网格中性能最强的匹配成本层级检索器，并与最强平面基线并列。在长文档多跳问答中，它是唯一在类别对类别层面显著优于 BM25 的层级方法，这一结果也经过无阅读器检索召回检查的验证。在匹配阅读器复现已发表的抽象 RAPTOR 变体时，即使提供强聚类摘要，NavTree 在每个多块预算下依然获胜，并且索引成本为零。这种排序优势在更强和开放权重的阅读器、更强的编码器以及隔离叶节点仅输出作为结构杠杆的全因子实验中依然有效。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06902

Authors: Priyank Jayraj, Poonam Goyal, Navneet Goyal

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.06902.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06902

Published: 2026-10-07T02:41:41.595Z

4. Component and Dimension Sparsity in Transformer Refusal Mechanisms

Abstract:Activation steering manipulates large language model behavior by intervening on internal activations, but the mechanistic basis of these interventions remains poorly understood. We decompose refusal steering into component-level interventions across four open-weight models, identifying the sparse subsets of attention and MLP components whose steering suffices to reproduce the full behavioral effect. We find that refusal directions concentrate in sparse component mechanisms comprising 28—48\% of upstream components, retaining 88—101\% of steering effectiveness. Within these mechanisms, effective steering further concentrates in approximately 50\% of residual stream dimensions, retaining 85—98\% of the component-mechanism baseline, consistent with a privileged basis structure. Sparsity thus operates at two levels: which components are steered, and which dimensions within those components carry the signal. Together these findings show that refusal is not diffusely encoded across a transformer but assembled by a structured, identifiable mechanism, providing a foundation for mechanistic understanding of how refusal behaviors are represented and steered. To facilitate reproducibility, we release all code and raw experimental results in this https URL.

中文摘要

摘要：激活引导通过干预内部激活来操纵大型语言模型的行为，但这些干预的机制基础仍然了解不足。我们将拒绝引导分解为跨四个开放权重模型的组件级干预，识别出那些能够充分再现全部行为效果的稀疏注意力和MLP组件子集。我们发现，拒绝方向集中在由28%到48%的上游组件组成的稀疏组件机制中，保留了88%到101%的引导效果。在这些机制中，有效引导进一步集中在大约50%的残差流维度中，保留了85%到98%的组件机制基线，这与特权基结构一致。因此，稀疏性在两个层面上起作用：哪些组件被引导，以及这些组件内哪些维度携带信号。综上，这些发现表明，拒绝行为不是在变换器中分散编码的，而是由一个结构化、可识别的机制组装而成，为理解拒绝行为的表示和引导提供了机制基础。为了便于重复实验，我们在此https URL发布了所有代码和原始实验结果。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06903

Authors: Vincent Siu, Glenn Grant-Richards, Vlad Pavlovich, Yizhou Sun, Dawn Song, Chenguang Wang

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.06903.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06903

Published: 2026-10-07T02:41:41.595Z

5. Stabilizing language models under continual learning via condition-anchored distillation

Abstract:Continual adaptation of language models can change their output distribution on prompts learned earlier, while retaining every old prompt-answer pair may be undesirable or impossible. We study condition-anchored generative distillation (CAGD): retain a small set of old prompts, use a frozen previous model to reconstruct completions and generation states, and match its predictive distributions while learning the next task. The formulation separates three roles that ordinary replay conflates: conditions select the behavior to protect, teacher generations locate relevant states, and soft targets specify how predictions may change. For autoregressive language generation, teacher-rollout distillation admits an exact chain-rule decomposition of sequence divergence. For masked-diffusion language modeling, our implementation directly controls local denoising drift on teacher-generated completions. In continual adaptation of a 219M masked diffusion language model, CAGD reduces four-task final held-out loss from 2.927 to 1.114 in one task order and from 2.168 to 0.891 in exact reverse. The same soft targets lower final average loss by 0.055 over hard replay when teacher-generated support is held identical. The direction persists on fresh facts and natural instructions across SMDM and Qwen3. On GSM8K, Qwen adaptation preserves answer-format compliance, but exact-match retention is seed-mixed at 0.6B and worsens at 1.7B. These results support condition-anchored functional preservation as a common design principle across the tested language-generation objectives.

中文摘要

摘要：语言模型的持续适应可能会改变其在之前学习的提示上的输出分布，而保留每一个旧的提示-回答对可能是不必要的或不可能的。我们研究了条件锚定生成蒸馏（CAGD）：保留一小部分旧提示，使用冻结的先前模型重构完成结果和生成状态，并在学习下一个任务时匹配其预测分布。该公式区分了普通重放混淆的三个角色：条件选择要保护的行为，教师生成定位相关状态，软目标指定预测可能的变化方式。对于自回归语言生成，教师展开蒸馏允许序列差异的精确链式规则分解。对于掩码扩散语言建模，我们的实现直接控制教师生成完成结果上的局部去噪漂移。在对一个 2.19 亿参数的掩码扩散语言模型进行持续适应时，CAGD 将四任务的最终保留损失从 2.927 降低到 1.114（按一个任务顺序）并从 2.168 降低到 0.891（完全相反的顺序）。当教师生成的支持保持相同时，相同的软目标将最终平均损失比硬重放降低了 0.055。这种效果在 SMDM 和 Qwen3 上对新事实和自然指令仍然存在。在 GSM8K 上，Qwen 的适应保持了答案格式的合规性，但在 6 亿参数下精确匹配的保留是种子混合的，在 17 亿参数下则恶化。这些结果支持条件锚定的功能保留作为所测试语言生成目标中的一个普遍设计原则。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06940

Authors: Huan Li, Zhe Cao, Qinlei Xie, Fushun Cui, Xuechen Liang

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.06940.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06940

Published: 2026-10-07T02:41:41.595Z

6. EMODE: Dynamic Para-Semantic Experts for Emotion-Aware Speech Language Modeling

Abstract:Large speech language models have demonstrated strong capabilities in unified cross-modal understanding and generation, yet paralinguistic cues, especially emotion, remain difficult to preserve. Existing systems typically rely on entangled acoustic representations, which allow the underlying language model to depend excessively on recovered lexical content instead of grounding its behavior in acoustic-prosodic evidence. We address this limitation with EMODE, an emotion-aware speech language model built around \textbf{Dynamic Para-Semantic Experts (DPSE)}. DPSE decomposes continuous speech features into semantic and paralinguistic pathways, routes them dynamically, and fuses them before integration into the language model. To turn this structural decomposition into functional specialization, EMODE is trained with a three-stage curriculum consisting of semantic warm-up, paralinguistic activation, and joint refinement, guided by Orthogonal Expert Guidance (OEG), Semantic-to-Acoustic Alignment (SAA), and Gating Diversity Regularization (GDR). Experiments on SER test, empathetic response evaluation, and the newly constructed bilingual MEPA benchmark show that EMODE improves the balance between lexical fidelity and emotional sensitivity, strengthens affect-grounded response generation, and exposes the value of explicit para-semantic factorization for robust cross-corpus emotion understanding.

中文摘要

摘要：大型语音语言模型在统一的跨模态理解与生成方面展现出强大的能力，但旁语言线索，尤其是情感，仍然难以保留。现有系统通常依赖纠缠的声学表示，这使得底层语言模型过度依赖恢复的词汇内容，而不是基于声学-韵律证据来指导其行为。我们通过 EMODE 解决了这一限制，它是一种面向情感的语音语言模型，围绕\textbf{动态旁语义专家（DPSE）}构建。DPSE 将连续语音特征分解为语义路径和旁语言路径，对其进行动态路由，并在整合到语言模型之前进行融合。为了将这种结构分解转化为功能专门化，EMODE 采用由三个阶段组成的课程训练，包括语义预热、旁语激活和联合优化，并由正交专家指导（OEG）、语义到声学对齐（SAA）以及门控多样性正则化（GDR）引导。SER 测试、共情响应评估，以及新构建的双语 MEPA 基准实验表明，EMODE 提高了词汇保真度与情感敏感性之间的平衡，加强了情感驱动的响应生成，并揭示了显式旁语义分解在稳健的跨语料库情感理解中的价值。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06956

Authors: Jianan Pan, Yiwen Gu, Xinze Li, Rui Wang, Kejie Huang

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.06956.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06956

Published: 2026-10-07T02:41:41.595Z

7. Verdicts Without Annotated Evidence: Rejection Sampling or Label-Only Post-Training for Evidence Recovery?

Abstract:In many review workflows the verdict is the only thing retained. The passages behind it are not marked, because that annotation costs far more than recording the decision. We measure how much of that evidence a small language model can recover when it is post-trained on the verdicts alone, with no human evidence labels at any stage. On ContractNLI the human evidence spans are held out until evaluation. Matching the recorded verdict and agreeing with those spans are not the same thing: across six systems the two scores are only weakly related and rank the systems differently, so accuracy is a poor guide when the citations have to be reviewable. Label-only training on the bare verdict reaches accuracy 0.896 and span F1 0.564. Rejection sampling, which keeps a generated trace only when its verdict matches the record and then picks one by an automatic source-grounding score, reaches 0.797 and 0.556, against 0.747 and 0.493 before training. Verbatim citation rises from 0.597 to 0.729 under label-only training and to 0.701 under rejection sampling. One seed on one corpus cannot say which method is better, but both improve the evidence without anyone annotating it.

中文摘要

摘要：在许多审查工作流中，裁决是唯一被保留的内容。其背后的段落没有被标记，因为标注这些段落的成本远高于记录裁决的成本。我们衡量一个小型语言模型在仅对裁决进行后训练时，在没有任何人工证据标签的情况下，能恢复多少证据。在 ContractNLI 上，人工证据片段在评估前是保留的。匹配记录的裁决与同意这些片段并不相同：在六个系统中，这两个分数的相关性很弱，并且对系统的排序不同，因此在必须审核引用内容时，准确率并不是一个可靠的指标。仅对裁决进行标签训练在裸裁决上达到 0.896 的准确率和 0.564 的片段 F1。拒绝采样（仅在生成的轨迹裁决与记录匹配时保留该轨迹，然后通过自动源定位分数选择一个）达到 0.797 和 0.556，而训练前分别为 0.747 和 0.493。在仅标签训练下，逐字引用从 0.597 提升到 0.729，在拒绝采样下提升到 0.701。在一个语料库上使用一个随机种子不能说明哪种方法更好，但两者都在无人标注证据的情况下改善了证据质量。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06962

Authors: Nishanth Nayakanti, Prasang Gupta, Ashutosh Bilthare, Kevin Paul

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.06962.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06962

Published: 2026-10-07T02:41:41.595Z

8. WavePrune: One period is often enough for RoPE

Abstract:Rotary Position Embedding (RoPE) encodes token positions by rotating each two-dimensional channel of the query and key vectors at a channel-specific frequency, making the attention logits invariant to a common shift of positions. However, this rotation is periodic, and it leads to position aliasing where relative positions separated by a full rotation period become hard to tell apart. To address this, we propose WavePrune, which restricts each channel to its first rotation period. We show that it removes the distractions in attention maps created by position aliasing and improves overall long-context performance. Specifically, WavePrune raises the HELMET score on four of five models we test without any extra tuning (e.g., 35.7 -> 40.0 on Qwen3-8B). When pretraining models from scratch, WavePrune also achieves lower validation loss at extrapolated lengths than pretraining without it. Because WavePrune restricts each channel to a sliding window, it induces a fine-grained sparsity that our hardware-aligned CUDA kernels exploit for 1.15x prefill and 1.24x decoding speedups over FlashAttention-2 at 32K context. Together, these results show that RoPE’s periodic structure, widely regarded as essential, is largely redundant beyond the first rotation period.

中文摘要

摘要：旋转位置嵌入（Rotary Position Embedding, RoPE）通过以每个通道特定的频率旋转查询和键向量的每个二维通道来编码标记位置，使注意力对数在位置的共同偏移下保持不变。然而，这种旋转是周期性的，这会导致位置混叠，即相隔一个完整旋转周期的相对位置变得难以区分。为了解决这一问题，我们提出了 WavePrune，它将每个通道限制在其第一个旋转周期内。我们证明它消除了位置混叠在注意力图中造成的干扰，并提高了整体长上下文性能。具体而言，WavePrune 在我们测试的五个模型中的四个模型上提升了 HELMET 分数，而无需任何额外调优（例如，Qwen3-8B 从 35.7 提升到 40.0）。在从零开始进行模型预训练时，WavePrune 在外推长度下也能实现比不使用它的预训练更低的验证损失。由于 WavePrune 将每个通道限制在一个滑动窗口内，它引入了细粒度的稀疏性，我们的硬件优化 CUDA 内核利用这一点，在 32K 上下文中相比 FlashAttention-2 实现了 1.15 倍的预填充和 1.24 倍的解码速度提升。综合来看，这些结果表明，RoPE 的周期结构，虽然被广泛认为是必不可少的，在超过第一个旋转周期后基本上是多余的。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06963 (HTTP 429)

Authors: Guancheng Du, Luotian Huang, Shaowen Wang, Si Li, Kaifeng Lyu

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.06963.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06963

Published: 2026-10-07T02:41:41.595Z

9. Calibrated Answers About Randomized Trials From a 4-Billion-Parameter Open Model: A Registered Test and a License-Clean Release

Abstract:Fiorillo v0.5 is an open model that answers typed questions with a probability for each answer. Its main specialist reads a randomized trial’s article, cut to 6,144 tokens, and answers whether an intervention significantly increased, significantly decreased or did not significantly change an outcome against a comparator (Evidence Inference 2.0, EI). It is Qwen3-4B-Base with low-rank adapters and a decision head, fine-tuned for EI only on the 1,431 of 2,657 training articles whose own license allows reuse. Four criteria registered on the Open Science Framework before this version’s test predictions decided its release, the second bar judged on EI’s test split, whose labels are public. On that split (1,218 prompts in 333 articles), the expected calibration error was 0.0168 against a limit of 0.05; log loss was below the prior’s by 0.8603 (95 percent interval 0.8104 to 0.9078) and below that of Gemma 4 31B-it, reading the same input, by 0.1829 (0.1164 to 0.2598); and macro-F1 was 0.9248 against 0.8668, so all four criteria passed. Training the same recipe on clean articles alone cost 0.0123 in accuracy (0.0034 to 0.0207; descriptive). With no article, macro-F1 fell to 0.4384; the title alone raised it by 0.0939 (0.0655 to 0.1234), which a title stating the result or recall of the trial could explain; exchanging intervention and comparator reversed 0.6652 of its direction answers. Run as released, the files matched the evaluated predictions within limits set in advance. The release is under the Apache License 2.0 (digital object identifier https://doi.org/10.57967/hf/10722).

中文摘要

摘要：Fiorillo v0.5 是一个开放模型，它通过为每个答案计算概率来回答输入的问题。其主要专家阅读了一篇随机试验文章，将其裁剪至 6,144 个标记，并回答某一干预措施相对于对照组是否显著增加、显著减少或未显著改变结果（证据推断 2.0，EI）。它基于 Qwen3-4B-Base，配有低秩适配器和决策头，仅针对 EI 进行微调，训练数据为 2,657 篇文章中许可证允许重用的 1,431 篇文章。在该版本测试预测之前，在开放科学框架上登记的四个标准决定了其发布，其中第二项标准基于 EI 的测试集进行评估，其标签是公开的。在该测试集上（333 篇文章中的 1,218 个提示），期望校准误差为 0.0168，低于 0.05 的限制；对数损失比先验低 0.8603（95% 区间：0.8104 至 0.9078），比阅读相同输入的 Gemma 4 31B-it 低 0.1829（0.1164 至 0.2598）；宏平均 F1 为 0.9248，而标准为 0.8668，因此四个标准均已通过。在仅使用干净文章训练相同方案时，准确率损失为 0.0123（0.0034 至 0.0207；描述性指标）。没有文章时，宏平均 F1 降至 0.4384；仅使用标题可提高 0.0939（0.0655 至 0.1234），这可能由标题中说明结果或试验回顾解释；交换干预与对照则使 0.6652 的方向性答案反向。按发布版本运行时，文件在预先设定的范围内匹配评估预测结果。该版本发布在 Apache License 2.0 下（数字对象标识符 https://doi.org/10.57967/hf/10722）。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.07019

Authors: Johann Emmanuel Li

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.07019.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.07019

Published: 2026-10-07T02:41:41.595Z

10. Investigating Model Compression for Neural Machine Translation in the Biomedical Domain

Abstract:Large-scale pretrained transformer models have achieved state-of-the-art performance across diverse machine translation tasks, including multilingual settings. Knowledge distillation has emerged as a sustainable approach for model compression, transferring knowledge from large teacher models to smaller, more efficient student models. Similarly, quantization, which reduces the numerical precision of model weights and activations (e.g., from 32-bit to 8-bit representations) is widely used to accelerate inference, enabling models to run several times faster during deployment. However, both techniques face limitations when applied to specialized domain data, particularly under low-resource conditions. In knowledge distillation, the effectiveness of transfer is often constrained by the scarcity of domain-specific parallel data, while quantization can lead to performance degradation as bit precision decreases. In this work, we investigate the combined application of knowledge distillation and quantization for French-to-English biomedical translation, a domain characterized by specialized terminology and limited parallel resources. We develop and compare multiple fine-tuning strategies to adapt compressed student models to this challenging setting. Our experiments demonstrate that a collaboratively distilled and quantized student model achieves a 69% reduction in size, a 98.21% increase in inference speed, and a 98.46% reduction in CO2 emissions compared to the original baseline all without sacrificing translation quality. These results indicate that jointly optimized compression techniques can yield efficient, high-performance models suitable for translation service providers operating under resource constraints.

中文摘要

摘要：大规模预训练的变换器（transformer）模型在多种机器翻译任务中，包括多语言环境，已经达到了最先进的性能。知识蒸馏（Knowledge Distillation）作为一种可持续的模型压缩方法，通过将大规模教师模型的知识转移到更小、更高效的学生模型中得到了广泛应用。类似地，量化（Quantization）通过降低模型权重和激活值的数值精度（例如，从32位降到8位表示）被广泛用于加速推理，使模型在部署时运行速度提高数倍。然而，这两种技术在应用于专门领域的数据时面临局限性，尤其是在低资源条件下。在知识蒸馏中，知识转移的有效性通常受限于领域特定平行数据的稀缺，而量化则可能随着位精度降低而导致性能下降。在本工作中，我们研究了知识蒸馏与量化在法语到英语生物医学翻译中的联合应用，该领域具有专业术语多且平行资源有限的特点。我们开发并比较了多种微调策略，以适应这种具有挑战性的压缩学生模型。实验结果表明，经过协同蒸馏与量化的学生模型相比原始基线模型，在不牺牲翻译质量的前提下，实现了模型尺寸减少69%、推理速度提升98.21%、二氧化碳排放减少98.46%。这些结果表明，联合优化的压缩技术能够产生高效、高性能的模型，适用于在资源受限条件下运营的翻译服务提供商。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.07032

Authors: Maria Zafar, Souhail Bakkali, Rejwanul Haque

Categories: cs.CL

PDF URL: https://arxiv.org/pdf/2610.07032.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.07032

Published: 2026-10-07T02:41:41.595Z

Agent Domain Papers

1. GAMEGO: Training Game-Dev Agents with Synthetic Trajectories Anchored in Real-World Assets

Abstract:Recent advances in Large Language Models (LLMs) have demonstrated remarkable capabilities in web front-end execution, with browser-based game generation emerging as a particularly prominent frontier. While previous efforts frequently rely on complex multi-turn workflows or focus on static game evaluation benchmarks, this work targets direct end-to-end real-world game synthesis driven by coding agents. However, generating complex games directly from sparse user queries often forces coding agents to make underspecified assumptions, yielding incomplete mechanics, disconnected gameplay flows, and limited visual aesthetics. To resolve this issue, this paper presents GameGo, a scalable framework that systematically transforms brief game seeds into comprehensive Product Requirements Documents grounded in industry game-development practices. To retain core gameplay constraints without restricting design exploration, GameGo uses task-specific dynamic compression to maximize information density while preserving instruction following. Based on this pipeline, GameGoData is constructed with 55,060 development trajectories across 2D, 2.5D, and 3D games, alongside GameGoBench, a benchmark comprising 124 diverse game queries. Training GameGoCoder on GameGoData yields a model that outperforms matched baselines and is comparable to frontier models across gamedev benchmarks. All code, datasets, and models will be made publicly available.

中文摘要

摘要：近期大型语言模型（LLMs）的进展展示了其在网页前端执行方面的显著能力，其中基于浏览器的游戏生成成为一个特别突出的前沿领域。尽管以往的工作通常依赖复杂的多轮工作流程或专注于静态游戏评估基准，但本研究的目标是由编码代理驱动的直接端到端真实世界游戏合成。然而，从稀疏的用户查询直接生成复杂游戏往往迫使编码代理做出不充分指定的假设，导致机制不完整、游戏流程脱节以及视觉美感有限。为了解决这一问题，本文提出了GameGo，一个可扩展框架，系统地将简短的游戏种子转化为基于行业游戏开发实践的完整产品需求文档。为了在不限制设计探索的情况下保留核心游戏玩法约束，GameGo使用任务特定的动态压缩来最大化信息密度，同时保持对指令的遵循。在此流程基础上，构建了GameGoData，涵盖2D、2.5D和3D游戏的55060条开发轨迹，以及GameGoBench，一个包含124个多样化游戏查询的基准。在GameGoData上训练GameGoCoder，得到的模型在游戏开发基准上超越了匹配的基线，并且与前沿模型相当。所有代码、数据集和模型将公开提供。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06910

Authors: Haoyue Yang, Jingyao Li, Zhengfan Wu, Jing Liu, Xuanle Zhao, Kang Liu

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06910.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06910

Published: 2026-10-07T02:42:08.971Z

2. Text2Dashboard: A Governed Agent Architecture for Natural-Language Dashboard Generation over Enterprise DataBrain

Abstract:Text2Dashboard is a DataBrain-specific prototype that turns natural-language analytic requests into inspectable dashboards. An installable Codex plugin and standalone Agent Runtime combine schema-constrained model decisions with typed tools, persistent state, and deterministic Hooks for approval, audit, checkpointing, recovery, and failure handling. The pipeline resolves entities, discovers metadata, enforces read-only SQL, composes dashboards, and applies static checks, dynamic preflight, and browser inspection. The model proposes actions while deterministic software controls execution and records state transitions. We evaluate the workflow on frozen real-DataBrain tasks and controlled Hook faults. Strict success was 6/8 on metadata and SQL tasks: metadata selection passed 4/4, all four SQL tasks met semantic criteria, and 2/4 met the exact output-column contract. The final release passed 4/4 single-panel dashboard tasks, one two-panel task, and one existing-dashboard refinement; a parameterised task exceeded its step limit. All ten fault scenarios met their specified outcomes without unapproved external side effects. Model inference accounted for over 97\% of observed runtime in every reported group. These small, DataBrain-specific results do not establish production readiness, general text-to-SQL accuracy, or an efficiency advantage over manual dashboard construction.

中文摘要

摘要：Text2Dashboard 是一个专门针对 DataBrain 的原型，它将自然语言分析请求转化为可检查的仪表板。一个可安装的 Codex 插件和独立的 Agent Runtime 将模式约束的模型决策与类型化工具、持久状态以及用于审批、审计、检查点、恢复和故障处理的确定性 Hooks 结合起来。该流程解决实体解析、元数据发现、只读 SQL 执行、仪表板组合，并应用静态检查、动态预检和浏览器检查。模型提议操作，而确定性软件控制执行并记录状态转换。我们在冻结的真实 DataBrain 任务和受控 Hook 故障上评估了该工作流。在元数据和 SQL 任务上严格成功率为 6/8：元数据选择全部通过 4/4，所有四个 SQL 任务符合语义标准，其中 2/4 符合精确输出列约定。最终版本完成了 4/4 单面板仪表板任务、一个双面板任务以及一个已有仪表板的改进；一个参数化任务超出了其步骤限制。所有十个故障场景均在未发生未批准的外部副作用的情况下实现了指定结果。模型推理占据了每个报告组超过 97% 的观察运行时。这些小型、专门针对 DataBrain 的结果并不意味着具备生产就绪性、通用文本到 SQL 的准确性，或相比手动构建仪表板的效率优势。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06914

Authors: Yiou Wu, Zezhi Tang, Ningwei Bai, Liuhaichen Yang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06914.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06914

Published: 2026-10-07T02:42:08.971Z

3. FluidPD: In-Place Elasticity for SLO-Aware Prefill-Decode Disaggregated LLM Serving

Abstract:Prefill-decode disaggregation is becoming a common architecture for LLM serving because it separates two phases with distinct execution patterns and SLO objectives. Existing systems typically combine a fixed prefill/decode worker ratio with request routing across workers. However, real-world workloads exhibit both short bursts and sustained shifts in the prefill-to-decode demand ratio. As a result, a configuration that is well provisioned at one time may quickly become mismatched, causing latency SLO violations even when idle capacity exists elsewhere. Existing autoscaling mechanisms can add capacity, but they react slowly, require spare GPUs, and do not directly address short-timescale phase imbalance. We present FluidPD, a P/D-disaggregated serving system that provides SLO-aware in-place elasticity. FluidPD introduces two complementary mechanisms. FluidToken handles transient imbalance by offloading a bounded portion of prefill computation to decode workers when decode-side slack is available. FluidRole handles sustained imbalance by reassigning running workers between prefill and decode roles in place, avoiding model reload and engine restart. Both mechanisms are guided by lightweight pressure indices that expose prefill and decode-side resource pressure before they appear as SLO violations. Across production Azure trace workloads, FluidPD improves overall SLO attainment over static SGLang by up to 94.6 percentage points, demonstrating that SLO-aware in-place P/D elasticity improves service quality without provisioning additional workers.

中文摘要

摘要：预填充-解码分离正在成为大型语言模型（LLM）服务的常见架构，因为它将具有不同执行模式和服务等级目标（SLO）的两个阶段分开。现有系统通常结合固定的预填充/解码工作者比例与工作者间的请求路由。然而，现实世界的工作负载表现出预填充与解码需求比例的短时突发和持续变化。因此，在某一时刻配置良好的系统可能很快变得不匹配，即使在其他地方存在空闲容量，也会导致延迟SLO违规。现有的自动扩展机制可以增加容量，但其反应缓慢，需要备用GPU，并且不能直接处理短时间尺度的阶段失衡。我们提出了FluidPD，一种P/D分离的服务系统，提供SLO感知的原地弹性。FluidPD引入了两种互补机制。FluidToken通过在解码侧有空余时将有限的预填充计算卸载到解码工作者，从而处理短暂性失衡。FluidRole通过在原地重新分配运行工作者的预填充和解码角色来处理持续失衡，避免模型重新加载和引擎重启。两种机制均通过轻量级压力指标进行引导，该指标在预填充和解码侧资源压力表现为SLO违规之前就予以暴露。在生产的Azure追踪工作负载中，FluidPD相比静态SGLang提高了整体SLO达成率高达94.6个百分点，表明SLO感知的原地P/D弹性能够在无需额外配置工作者的情况下提升服务质量。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06917

Authors: Kartik Ramesh, Kaidi Fu, Zihan Zheng, Jiahuan Yu, Fabio Oliveira, Carlos Costa, Minjia Zhang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06917.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06917

Published: 2026-10-07T02:42:08.971Z

4. Anchor Divergence for Semantic Geometry in Contrastive Learning

Abstract:This paper concerns how semantic context determines geometry in learned vector representations. Similarity is typically measured using cosine similarity, which provides a single fixed geometry. Semantic similarity, however, is inherently context dependent: two images may be similar because they depict the same object, share a visual style, or are relevant to the same clinical finding. We show that contrastive representations naturally encompass a family of geometries that can be specialized to particular semantic structure. The key idea is to use an interplay between contrastive learning, exponential families, and information geometry to establish a correspondence between probability distributions over “anchors” and Bregman geometries on the representation space. We use this correspondence to define “Anchor Divergences”, a method for specifying context-specific semantic geometries on fixed representations. Under this correspondence, modeling the anchor distribution models the geometry itself. Experiments on retrieval show that anchor divergences provide an effective and efficient way to specify context-specific semantic similarity.

中文摘要

摘要：本文讨论了语义上下文如何在学习到的向量表示中决定几何结构。相似性通常使用余弦相似度来衡量，这提供了一种固定的几何结构。然而，语义相似性本质上是依赖于上下文的：两张图像可能相似，因为它们描绘了相同的对象、具有相同的视觉风格，或与相同的临床发现相关。我们表明，对比表示自然包含一族几何结构，这些几何结构可以专门用于特定的语义结构。关键思想是利用对比学习、指数族和信息几何之间的相互作用，在“锚点”的概率分布与表示空间上的Bregman几何之间建立对应关系。我们利用这种对应关系定义了“锚点散度”，这是一种在固定表示上指定上下文特定语义几何结构的方法。在这种对应关系下，对锚点分布建模即对几何本身建模。检索实验表明，锚点散度提供了一种有效且高效的方式来指定上下文特定的语义相似性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06919

Authors: Akash Kannan, Kiho Park, Victor Veitch

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06919.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06919

Published: 2026-10-07T02:42:08.971Z

5. RadOnc-Agent: An LLM-Orchestrated Framework for AI Workflows Across the Radiotherapy Care Pathway

Abstract:Artificial intelligence has advanced individual radiotherapy tasks, yet these capabilities remain separated across clinical stages, software environments and data modalities. This fragmentation contrasts with the longitudinal radiotherapy workflow from treatment decision-making through follow-up. Here we present RadOnc-Agent, an agentic artificial-intelligence framework that formalizes radiotherapy into four clinical phases and provides 26 callable functions through a conversational interface. A large-language-model controller maps clinical intent to schema-constrained calls, preserves patient and workflow context, and routes requests to specialist services. We evaluated system execution using 2,600 single-function requests (7,800 repeat executions), 200 prespecified synthetic cross-stage scenarios spanning four phases (600 executions), and 120 workflow instances from 60 de-identified patient records (360 clean executions) representing decision-to-planning and planning-to-adaptation. RadOnc-Agent selected the intended function in 98.79% of single-function executions, completed 96.50% of scripted cross-stage workflows, and completed 96.67% of real-patient workflow executions. In comparative ablations, removing longitudinal state reduced cross-stage completion from 96.50% to 84.00%, while disabling schema and identity validation increased mismatched backend dispatch from 0% to 95.28% in a replay/test evaluation. These findings establish the technical feasibility of an LLM-orchestrated architecture for coordinating heterogeneous radiotherapy capabilities and information across longitudinal workflows; they do not establish clinical correctness, clinical utility or prospective benefit.

中文摘要

摘要：人工智能已经在个体放射治疗任务中取得进展，但这些能力仍然分散在不同的临床阶段、软件环境和数据模式中。这种分散性与从治疗决策到随访的纵向放射治疗工作流程形成对比。在这里，我们提出了 RadOnc-Agent，这是一种具代理性的人工智能框架，将放射治疗形式化为四个临床阶段，并通过对话界面提供 26 个可调用功能。大型语言模型控制器将临床意图映射到模式约束的调用，保留患者和工作流程的上下文，并将请求路由到专业服务。我们使用 2,600 个单功能请求（7,800 次重复执行）、200 个跨四阶段预设的合成场景（600 次执行）和来自 60 份去标识化患者记录的 120 个工作流程实例（360 次有效执行）评估了系统执行情况，涵盖从决策到计划以及从计划到调整。RadOnc-Agent 在单功能执行中以 98.79% 的选择率选择了预期功能，在脚本化跨阶段工作流程中完成率为 96.50%，在真实患者工作流程执行中完成率为 96.67%。在比较消融实验中，删除纵向状态使跨阶段完成率从 96.50% 降至 84.00%，而禁用模式和身份验证会在回放/测试评估中将不匹配的后端调度从 0% 提高到 95.28%。这些结果确立了由大型语言模型协调架构在纵向工作流程中整合异构放射治疗能力和信息的技术可行性；但它们并未确立临床正确性、临床实用性或前瞻性益处。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06923

Authors: Caiwen Jiang, Shuoyang Wei, Songlin Zhao, Junyu Li, Jingyuan Chen, Wei Liu

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06923.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06923

Published: 2026-10-07T02:42:08.971Z

6. Metonymic Circuits for Abstract Concept Grounding in Vision Transformers

Abstract:We study how Vision Transformers ground abstract concepts (e.g., angry) when training data provide limited direct referential evidence. We hypothesize a metonymic grounding mechanism in which abstract predictions are driven by concrete, interpretable anchor concepts (e.g., fire) that bridge visual signals to abstract semantics. By applying Transcoders on CLIP and DINO vision encoders, we recover intermediate features that can be associated with semantic labels for more concrete concepts, and trace their contributions in circuits underlying abstract concept recognition. Experiments on a carefully curated icon dataset reveal structured metonymic circuits, in which perceptual primitives dominate early layers and object-like anchors precede abstract targets. Images containing rendered text instead recruit a distinct perceptual-to-textual route. Causal interventions further validate that metonymic intermediates are functionally involved in grounding abstract concepts.

中文摘要

摘要：我们研究了当训练数据提供的直接指称证据有限时，视觉变换器如何对抽象概念（例如“愤怒”）进行落地。我们提出了一种转喻式落地机制假设，即抽象预测是由具体的、可解释的锚定概念（例如“火”）驱动的，这些概念将视觉信号与抽象语义联系起来。通过在CLIP和DINO视觉编码器上应用Transcoders，我们恢复了可以与更具体概念的语义标签相关联的中间特征，并追踪它们在抽象概念识别相关电路中的贡献。在精心策划的图标数据集上的实验显示了结构化的转喻电路，其中感知原语主导了早期层，而类对象的锚点出现在抽象目标之前。包含渲染文本的图像则使用了不同的感知到文本路径。因果干预进一步验证了转喻中间体在抽象概念落地中的功能性作用。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06928

Authors: Jing Ding, Ziqiao Ma, Jiayuan Mao, Joyce Chai, Freda Shi

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06928.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06928

Published: 2026-10-07T02:42:08.971Z

7. Principles that Guide, Actions that Inform: Agent Evolution via Knowledge Abstraction

Abstract:Large language model (LLM) agents have demonstrated strong capabilities in interactive environments, yet their ability to continually evolve from experience remains limited. Although fine-tuning enables adaptation, its dependence on parameter access and high computational costs restrict its flexibility, especially for large-scale and closed-source LLMs. External memory offers an alternative by allowing agents to accumulate experience without modifying model parameters. However, existing methods mainly focus on experience representation and organization, while the acquired knowledge remains tightly coupled with specific tasks and contexts, limiting generalization. A key challenge is how to transform concrete interactions into abstract and reusable knowledge that guides future decisions beyond individual experiences. To address this challenge, we propose SAGA (\underline{\textbf{S}}elf-evolving \underline{\textbf{A}}gents through Experience-\underline{\textbf{G}}rounded \underline{\textbf{A}}bstraction), a framework for experience-grounded knowledge abstraction and utilization in LLM agents. SAGA progressively transforms interaction trajectories into episodic descriptions, reusable procedures, and principles with explicit applicability conditions, while maintaining links to execution evidence. Retrieved principles are instantiated into task-specific guidance and used to refine candidate actions through corrective feedback and resampling. This creates an execution—abstraction feedback loop, where accumulated knowledge guides future interactions and new experiences continuously update hierarchical memory. Experiments on ScienceWorld and ALFWorld demonstrate improved task performance, with ablation studies highlighting the importance of contextual instantiation and action regulation for leveraging principle-level knowledge.

中文摘要

摘要：大型语言模型（LLM）代理在交互环境中表现出强大的能力，但其从经验中持续进化的能力仍然有限。虽然微调能够实现适应，但其对参数访问的依赖和高计算成本限制了灵活性，尤其是对于大规模和封闭源代码的LLM。外部记忆提供了一种替代方案，使代理能够在不修改模型参数的情况下积累经验。然而，现有方法主要关注经验的表示和组织，而获得的知识仍紧密依赖于特定任务和环境，限制了泛化能力。一个关键挑战是如何将具体的交互转化为抽象且可重复使用的知识，从而指导未来决策，而不仅仅依赖于个体经验。为解决这一挑战，我们提出了SAGA（通过经验驱动抽象进行自我进化的代理，Self-evolving Agents through Experience-Grounded Abstraction），这是一个用于LLM代理的经验驱动知识抽象与利用的框架。SAGA逐步将交互轨迹转化为情节描述、可重复使用的程序和具有明确适用条件的原则，同时保持与执行证据的关联。检索到的原则被实例化为特定任务指导，并通过纠正反馈和重新采样用于优化候选动作。由此形成了执行—抽象反馈循环，累积的知识指导未来交互，而新经验持续更新分层记忆。在ScienceWorld和ALFWorld上的实验显示了任务性能的提升，消融研究强调了上下文实例化和动作调控对于利用原则级知识的重要性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06964

Authors: Bowen Ye, Yongchao Xu, Junkai Ma, Xiang Yin, Wenzhao Li

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06964.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06964

Published: 2026-10-07T02:42:08.971Z

8. AegisFlow: A Multi-Agent Agentic AI Framework for Autonomous Remediation and Self-Healing in Fragile Data Ecosystems

Abstract:Traditional data pipelines are notoriously brittle, often failing due to upstream schema drift, API contract changes, or website DOM modifications. Present observability tools only raise alerts but for human engineers, resulting in a high Mean Time to Repair (MTTR) and operational fatigue. In this paper we propose AegisFlow (Agentic Engine for Intelligent Self-healing and Graph-driven Operations for Workload remediation), a novel agentic framework that closes the loop between detection and resolution. AegisFlow uses a Watchdog agent to collect runtime telemetry and has a Repair agent to automatically create, test and deploy code patches based on Large Language Models (LLMs). The framework presents the non-intrusive execution model called Parallel Shadow Patching, a non-intrusive execution model based on the Monitor, Analyze, Plan, Execute, Knowledge (MAPE-K) loop to generate and verify patches in digital twin environments. Through experimental testing, we have evaluated AegisFlow across five common failure scenarios, and see 98.1 percent improvement in MTTR (from an average of 170 minutes per patch to 3.2 minutes) and a patch success rate of 92 percent . In particular, the system is successful in dealing with changes in the JSON schema (96 percent ) and punctuation drift (98 percent ), and is least successful in Shadow DOM cases (85 percent ). AegisFlow frees up about 98 percent of data engineering on-call time from firefighting and reallocates it towards innovation. The framework is deployment agnostic consisting of a system that can be deployed in a plugin fashion into an existing pipeline orchestration system with minimal uplift to the existing system.

中文摘要

摘要：传统数据管道众所周知易出问题，常因上游模式漂移、API 合同变更或网站 DOM 修改而失败。现有的可观测性工具仅向人工工程师发出警报，导致平均修复时间（MTTR）长且运维疲劳。在本文中，我们提出了 AegisFlow（用于智能自愈和基于图的工作负载修复操作的自主引擎），一种新型自主框架，能闭合检测与解决之间的循环。AegisFlow 使用 Watchdog 代理收集运行时遥测，并通过 Repair 代理基于大型语言模型（LLM）自动创建、测试和部署代码补丁。该框架提出了名为平行影子修补（Parallel Shadow Patching）的非侵入式执行模型，这是一种基于监控、分析、计划、执行与知识（MAPE-K）循环的非侵入式执行模型，用于在数字孪生环境中生成和验证补丁。通过实验测试，我们在五种常见失败场景中评估了 AegisFlow，MTTR 提升了 98.1%（平均每个补丁从 170 分钟降至 3.2 分钟），补丁成功率为 92%。其中，系统在处理 JSON 模式变化（96%）和标点漂移（98%）方面最为成功，而在 Shadow DOM 情况下成功率最低（85%）。AegisFlow 释放了约 98% 数据工程值班时间，使其免于应急处理，并重新分配用于创新。该框架对部署环境无依赖，由一个可以以插件方式部署到现有管道编排系统中的系统组成，对现有系统的改动最小。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06971

Authors: Muhammad Bilal Awan, Zubair Hussain, Abdul Shahid

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06971.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06971

Published: 2026-10-07T02:42:08.971Z

9. EPOCH: Reliable Discovery through Evidence-Governed Search

Abstract:AI research agents are increasingly used to search over programs, mathematical constructions, and proofs. However, existing systems typically optimize evaluator feedback without adequately governing how that feedback is interpreted, challenged, and reused. As a result, promising but fragile candidates can be promoted as discoveries, while benchmark improvements, finite certificates, and theorem-level claims are too easily conflated. We introduce EPOCH, an evidence-governed architecture designed to close this gap. EPOCH implements an evidence-governed discovery loop by combining explicit task contracts, typed memory, active falsification, admission checks, and independent replay, so that each candidate is evaluated against the strength and scope of the claim it supports. EPOCH achieves state-of-the-art aggregate performance on AlgoTune, substantially exceeding the strongest baseline in mean normalized score (0.65 vs. 0.53), and attains the highest mean score on the internal Math14 suite (0.57). It further shows favorable held-out behavior under official-test replay and leads the descriptive aggregate on AgentHPO. Across ten discovery problems, EPOCH delivers substantial task-specific advances, including improved executable constructions, optimized algorithms, counterexamples, and proof-supported results. These advances demonstrate its ability to convert search into concrete progress across mathematical and computational domains. Together, the results suggest that evidence governance is a necessary step toward AI research agents that produce not only stronger solutions, but also more trustworthy scientific discoveries.

中文摘要

摘要：AI研究代理越来越多地用于搜索程序、数学构造和证明。然而，现有系统通常优化评估者的反馈，却没有充分管理这些反馈是如何被解读、质疑和再利用的。因此，有前景但脆弱的候选结果可能被误认为是发现，而基准改进、有限证书和定理级声明则过于容易混淆。我们提出了EPOCH，一种由证据管理的架构，旨在弥补这一差距。EPOCH通过结合明确的任务契约、类型化存储、主动反证、准入检查和独立重放，实现了由证据驱动的发现循环，使每个候选结果都根据其支持的声明的强度和范围进行评估。EPOCH在AlgoTune上实现了最先进的综合性能，平均归一化分数远超最强基线（0.65对0.53），并在内部Math14套件上获得最高平均分（0.57）。它在官方测试重放下也表现出良好的保持行为，并在AgentHPO上引领描述性综合指标。在十个发现问题中，EPOCH带来了显著的特定任务改进，包括改进的可执行构造、优化算法、反例和有证明支持的结果。这些进展展示了其将搜索转化为数学和计算领域具体进展的能力。总体结果表明，证据管理是实现AI研究代理不仅产生更强解决方案，也能生成更可信科学发现的必要步骤。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06986

Authors: Binjie Guo, Aisheng Mo, Ruitong Li, Xinle Deng

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06986.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06986

Published: 2026-10-07T02:42:08.971Z

10. When better traffic forecasts fail to improve signal control: a layered diagnostic study of forecast-to-decision value

Abstract:Improved traffic forecasts do not necessarily yield better signal-control decisions. We investigate this gap through a layered diagnostic study using 29 days of reconstructed demand from Xuancheng, China, with seven dates reserved for testing. The framework evaluates point forecasts, conformal intervals, dependence-aware scenarios, and matched closed-loop controllers. Entry-level and movement-level forecasts reduce mean absolute error by 4.03% and 3.92%, respectively, relative to historical means. A nominal 90% conformal interval achieves 90.72% marginal coverage but only 75.66% on an ex-post high-demand subset. Interface audits identify decision-time leakage and reveal that only two of nine controlled intersections offer multiple effective actions. We correct the temporal interface and compare causal forecasts with a five-second event oracle using exhaustive joint-action search. A synthetic positive control demonstrates that future information can reduce the internal rollout cost by 61.5%. On the frozen test dates, however, causal forecasts and the event oracle increase queue vehicle?seconds by 6.09% and 3.39% relative to the matched no-future rollout, while the oracle reduces spillback exposure by 3.78%; paired-day bootstrap intervals cross zero. These findings indicate that forecast value depends on temporal observability, action identifiability, dynamics consistency, and objective alignment. The proposed protocol provides a practical way to diagnose where predictive improvements fail to translate into operational benefits.

中文摘要

摘要：改进的交通预测不一定能产生更好的信号控制决策。我们通过一项分层诊断研究来调查这一差距，使用来自中国宣城的29天重建需求数据，其中七天用于测试。该框架评估点预测、保形区间、依赖感知情景以及匹配的闭环控制器。与历史平均值相比，入门级和动作级预测将平均绝对误差分别降低了4.03%和3.92%。名义上90%的保形区间实现了90.72%的边际覆盖率，但在事后高需求子集上仅为75.66%。界面审计识别出决策时间泄漏，并显示在九个受控交叉口中，只有两个提供多个有效动作。我们校正了时间界面，并使用全面的联合动作搜索，将因果预测与五秒事件预言器进行比较。一种合成正控制演示表明，未来信息可以将内部展开成本降低61.5%。然而，在冻结测试日期上，相对于匹配的无未来展开，因果预测和事件预言器分别增加了队列车辆·秒数6.09%和3.39%，而预言器将溢出暴露减少了3.78%；配对日自助区间跨越零。这些发现表明，预测价值取决于时间可观测性、动作可识别性、动态一致性和目标对齐。所提出的协议提供了一种实用的方法，用于诊断预测改进未能转化为运营收益的原因。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06992

Authors: Jianing Long, Xiaobin Li, Wuming Lei, Weiguang Wang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06992.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06992

Published: 2026-10-07T02:42:08.971Z

Evaluation Domain Papers

1. GAMEGO: Training Game-Dev Agents with Synthetic Trajectories Anchored in Real-World Assets

中文摘要

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06910

Authors: Haoyue Yang, Jingyao Li, Zhengfan Wu, Jing Liu, Xuanle Zhao, Kang Liu

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06910.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06910

Published: 2026-10-07T02:42:34.884Z

2. Text2Dashboard: A Governed Agent Architecture for Natural-Language Dashboard Generation over Enterprise DataBrain

中文摘要

摘要：Text2Dashboard 是一个专门针对 DataBrain 的原型，它将自然语言分析请求转化为可检查的仪表板。一个可安装的 Codex 插件和独立的 Agent Runtime 将模式约束的模型决策与类型化工具、持久状态以及用于审批、审计、检查点、恢复和故障处理的确定性 Hooks 结合起来。该流程解决实体解析、发现元数据、执行只读 SQL、生成仪表板，并应用静态检查、动态预检和浏览器检查。模型提出操作建议，而确定性软件控制执行并记录状态转换。我们在冻结的真实 DataBrain 任务和受控 Hook 故障上评估了该工作流程。严格成功率在元数据和 SQL 任务上为 6/8：元数据选择均通过 4/4，四个 SQL 任务全部满足语义标准，其中 2/4 满足精确输出列契约。最终版本通过了 4/4 单面板仪表板任务、一个双面板任务和一个现有仪表板优化任务；一个参数化任务超出了其步骤限制。所有十个故障场景均按预期结果执行且未产生未经批准的外部副作用。模型推理在每个报告组中占观测运行时超过 97%。这些小规模、DataBrain 专属的结果并不能证明其具备生产准备性、通用文本到 SQL 的准确性，或在效率上优于手动构建仪表板。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06914

Authors: Yiou Wu, Zezhi Tang, Ningwei Bai, Liuhaichen Yang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06914.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06914

Published: 2026-10-07T02:42:34.884Z

3. FluidPD: In-Place Elasticity for SLO-Aware Prefill-Decode Disaggregated LLM Serving

中文摘要

摘要：预填充-解码分离正在成为大型语言模型（LLM）服务的常见架构，因为它将具有不同执行模式和服务等级目标（SLO）的两个阶段分开。现有系统通常结合固定的预填充/解码工作者比例与工作者间的请求路由。然而，现实世界的工作负载表现出预填充与解码需求比例的短时突发和持续变化。因此，在某一时刻配置良好的系统可能很快变得不匹配，即使在其他地方存在空闲容量，也会导致延迟SLO违规。现有的自动扩展机制可以增加容量，但其反应缓慢，需要备用GPU，并且不能直接处理短时间尺度的阶段失衡。我们提出了FluidPD，一种P/D分离的服务系统，提供SLO感知的原地弹性。FluidPD引入了两种互补机制。FluidToken通过在解码侧有空余时将有限的预填充计算卸载到解码工作者，从而处理短暂性失衡。FluidRole通过在原地重新分配运行工作者的预填充和解码角色来处理持续失衡，避免模型重新加载和引擎重启。两种机制均通过轻量级压力指标进行引导，该指标在预填充和解码侧资源压力表现为SLO违规之前就予以暴露。在生产的Azure跟踪工作负载中，FluidPD相比静态SGLang提高了整体SLO达成率高达94.6个百分点，表明SLO感知的原地P/D弹性能够在无需额外配置工作者的情况下提升服务质量。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06917

Authors: Kartik Ramesh, Kaidi Fu, Zihan Zheng, Jiahuan Yu, Fabio Oliveira, Carlos Costa, Minjia Zhang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06917.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06917

Published: 2026-10-07T02:42:34.884Z

4. Anchor Divergence for Semantic Geometry in Contrastive Learning

中文摘要

摘要：本文讨论了语义上下文如何在学习到的向量表示中决定几何结构。相似性通常使用余弦相似度来衡量，这提供了一种固定的几何结构。然而，语义相似性本质上依赖于上下文：两幅图像可能相似，因为它们描绘了相同的对象、具有相似的视觉风格，或与相同的临床发现相关。我们展示了对比表示自然包含一族可以针对特定语义结构进行专门化的几何结构。关键思想是利用对比学习、指数族和信息几何之间的相互作用，在“锚点”的概率分布与表示空间上的Bregman几何之间建立对应关系。我们使用这种对应关系定义了“锚点散度”，这是一种在固定表示上指定上下文特定语义几何的方法。在这种对应关系下，对锚点分布进行建模即对几何结构本身进行建模。检索实验表明，锚点散度提供了一种有效且高效的方法来指定上下文特定的语义相似性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06919

Authors: Akash Kannan, Kiho Park, Victor Veitch

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06919.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06919

Published: 2026-10-07T02:42:34.884Z

5. RadOnc-Agent: An LLM-Orchestrated Framework for AI Workflows Across the Radiotherapy Care Pathway

中文摘要

摘要：人工智能已经在个体放射治疗任务中取得进展，但这些能力仍然分散在不同的临床阶段、软件环境和数据模式中。这种分散性与从治疗决策到随访的纵向放射治疗工作流程形成对比。在这里，我们提出了 RadOnc-Agent，这是一种具代理性的人工智能框架，将放射治疗形式化为四个临床阶段，并通过对话界面提供 26 个可调用函数。大型语言模型控制器将临床意图映射到模式约束的调用，保留患者和工作流程的上下文，并将请求路由到专业服务。我们使用 2,600 个单功能请求（7,800 次重复执行）、200 个跨四阶段预设的合成场景（600 次执行）和来自 60 份去标识化患者记录的 120 个工作流程实例（360 次有效执行）评估了系统执行情况，涵盖从决策到计划以及从计划到调整。RadOnc-Agent 在单功能执行中以 98.79% 的选择率选择了预期功能，在脚本化跨阶段工作流程中完成率为 96.50%，在真实患者工作流程执行中完成率为 96.67%。在比较消融实验中，删除纵向状态使跨阶段完成率从 96.50% 降至 84.00%，而禁用模式和身份验证会在回放/测试评估中将不匹配的后端调度从 0% 提高到 95.28%。这些结果确立了由大型语言模型协调架构在纵向工作流程中整合异构放射治疗能力和信息的技术可行性；但它们并未确立临床正确性、临床实用性或前瞻性益处。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06923

Authors: Caiwen Jiang, Shuoyang Wei, Songlin Zhao, Junyu Li, Jingyuan Chen, Wei Liu

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06923.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06923

Published: 2026-10-07T02:42:34.884Z

6. Metonymic Circuits for Abstract Concept Grounding in Vision Transformers

中文摘要

摘要：我们研究了在训练数据提供的直接指称证据有限的情况下，视觉变换器（Vision Transformers）如何建立抽象概念（例如“愤怒”）的基础。我们假设存在一种转喻性（metonymic）基础机制，其中抽象预测由具体的、可解释的锚定概念（例如“火”）驱动，这些概念将视觉信号与抽象语义联系起来。通过在 CLIP 和 DINO 视觉编码器上应用转码器（Transcoders），我们恢复了可以与更具体概念的语义标签相关联的中间特征，并追踪它们在支持抽象概念识别的神经回路中的贡献。在一个精心策划的图标数据集上的实验揭示了结构化的转喻回路，其中感知原语在早期层占主导，而类对象的锚定概念在抽象目标之前出现。包含渲染文本的图像则招募了另一条独特的感知到文本的路径。因果干预进一步验证了转喻中间体在抽象概念基础中具有功能性作用。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06928

Authors: Jing Ding, Ziqiao Ma, Jiayuan Mao, Joyce Chai, Freda Shi

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06928.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06928

Published: 2026-10-07T02:42:34.884Z

7. Principles that Guide, Actions that Inform: Agent Evolution via Knowledge Abstraction

中文摘要

摘要：大型语言模型（LLM）智能体在交互环境中展现出强大能力，但其持续从经验中演进的能力仍然有限。虽然微调能够适应，但其对参数访问的依赖和高计算成本限制了其灵活性，尤其适用于大规模和闭源LLMs。外部内存提供了一种替代方案，允许智能体在不修改模型参数的情况下积累经验。然而，现有方法主要关注经验的表示和组织，而获得的知识仍紧密结合特定任务和上下文，限制了泛化。一个关键挑战是如何将具体互动转化为抽象且可重用的知识，指导未来决策，超越个人体验。为应对这一挑战，我们提出了SAGA（\下划线{\textbf{S}}通过经验\下划线{\textbf{A}}绅士们通过体验-\下划线{\textbf{G}}圆角\下划线{\textbf{A}}抽象化）提出一个基于经验的知识抽象与利用框架，用于LLM代理。SAGA逐步将交互轨迹转化为情节描述、可重用程序和带有明确适用条件的原则，同时保持与执行证据的链接。检索到的原则被实例化为任务特定指导，并用于通过修正反馈和重采样来优化候选动作。这形成了执行-抽象反馈循环，积累的知识引导未来的交互，新体验不断更新层级记忆。ScienceWorld和ALFWorld上的实验显示任务性能有所提升，消融研究强调了情境实例化和动作调控在利用原理层面知识中的重要性。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06964

Authors: Bowen Ye, Yongchao Xu, Junkai Ma, Xiang Yin, Wenzhao Li

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06964.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06964

Published: 2026-10-07T02:42:34.884Z

8. AegisFlow: A Multi-Agent Agentic AI Framework for Autonomous Remediation and Self-Healing in Fragile Data Ecosystems

中文摘要

摘要：传统数据管道众所周知易出问题，常因上游模式漂移、API 合同变更或网站 DOM 修改而失败。现有的可观测性工具仅向人工工程师发出警报，导致平均修复时间（MTTR）长且运维疲劳。在本文中，我们提出了 AegisFlow（用于智能自愈和基于图的工作负载修复操作的自主引擎），一种新型自主框架，能闭合检测与解决之间的循环。AegisFlow 使用 Watchdog 代理收集运行时遥测，并通过 Repair 代理基于大型语言模型（LLM）自动创建、测试和部署代码补丁。该框架提出了名为“并行影子修补”的非侵入式执行模型，该模型基于监控-分析-计划-执行-知识（MAPE-K）循环，在数字孪生环境中生成并验证补丁。通过实验测试，我们在五种常见故障场景中评估了 AegisFlow，MTTR 提升了 98.1%（每个补丁平均从 170 分钟降至 3.2 分钟），补丁成功率达到 92%。特别地，系统在处理 JSON 模式变化（96%）和标点漂移（98%）方面非常成功，Shadow DOM 情况下成功率最低（85%）。AegisFlow 释放了约 98% 的数据工程值班时间，使其从应急处理转向创新。该框架与部署无关，由一个可以以插件方式部署到现有管道编排系统中、对现有系统几乎没有改动的系统组成。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06971

Authors: Muhammad Bilal Awan, Zubair Hussain, Abdul Shahid

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06971.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06971

Published: 2026-10-07T02:42:34.884Z

9. EPOCH: Reliable Discovery through Evidence-Governed Search

中文摘要

摘要：AI 研究代理越来越多地用于搜索程序、数学结构和证明。然而，现有系统通常优化评估者的反馈，却没有充分管理这些反馈是如何被解读、质疑和再利用的。因此，有前景但脆弱的候选结果可能被误认为是发现，而基准改进、有限证书和定理级声明则过于容易混淆。我们提出了 EPOCH，一种由证据驱动的架构，旨在弥补这一差距。EPOCH 通过结合明确的任务合同、类型化内存、主动证伪、审核检查和独立重放，实现了证据驱动的发现循环，使每个候选结果都根据其支持的声明的强度和范围进行评估。EPOCH 在 AlgoTune 上实现了最先进的综合性能，平均归一化分数远超最强基线（0.65 对 0.53），并在内部 Math14 套件上获得最高平均分（0.57）。它在官方测试重放下也表现出良好的保持行为，并在 AgentHPO 上引领描述性综合指标。在十个发现问题中，EPOCH 提供了显著的特定任务进展，包括改进的可执行构造、优化算法、反例和有证明支持的结果。这些进展展示了其将搜索转化为数学和计算领域具体成果的能力。总体而言，结果表明证据治理是迈向不仅产生更强解决方案，而且产生更可信科学发现的 AI 研究代理的必要步骤。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06986

Authors: Binjie Guo, Aisheng Mo, Ruitong Li, Xinle Deng

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06986.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06986

Published: 2026-10-07T02:42:34.884Z

10. When better traffic forecasts fail to improve signal control: a layered diagnostic study of forecast-to-decision value

中文摘要

摘要：改进的交通预测不一定能产生更好的信号控制决策。我们通过一项分层诊断研究来调查这一差距，使用来自中国宣城的29天重建需求数据，其中七天用于测试。该框架评估点预测、保形区间、依赖感知情景以及匹配的闭环控制器。与历史平均值相比，入门级和动作级预测将平均绝对误差分别降低了4.03%和3.92%。名义90%的保形区间实现了90.72%的边际覆盖率，但在事后高需求子集上仅为75.66%。界面审计发现决策时间泄漏，并揭示九个受控交叉口中仅有两个提供多个有效行动。我们修正了时间界面，并将因果预测与使用穷尽联合行动搜索的五秒事件信号进行比较。一种合成正控制显示未来信息可以将内部展开成本降低61.5%。然而，在冻结测试日期上，因果预测和事件信号分别较匹配的无未来展开增加了6.09%和3.39%的排队车辆?秒，而事件信号减少了3.78%的溢流暴露；配对日自助区间穿越零。这些发现表明，预测的价值取决于时间可观测性、行动可识别性、动态一致性以及目标一致性。所提出的协议提供了一种实用方法来诊断预测改进为何未能转化为操作效益。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06992

Authors: Jianing Long, Xiaobin Li, Wuming Lei, Weiguang Wang

Categories: cs.AI

PDF URL: https://arxiv.org/pdf/2610.06992.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06992

Published: 2026-10-07T02:42:34.884Z

VLM Domain Papers

1. Medical Image Alignment Assessment as a Test of Generalist Visual Reasoning in Frontier Multimodal Models

Abstract:Frontier multimodal large language models (MLLMs) are increasingly positioned as general purpose visual reasoners as part of the quest for artificial general intelligence. A key test of this generality is whether they can perform novel visual judgments that humans can make reliably from visual evidence and task instructions, without task-specific parameter optimisation. We investigate this question through the task of medical image alignment assessment, where the goal is to establish whether there is anatomical correspondence between two images. Human visual assessment of image alignment is still the gold standard and most common approach; however, it requires trained operators and is impractical to scale for large datasets. We evaluate recent generations of MLLMs on two exemplar medical image alignment tasks, varying both prompting strategies and image-presentation methods. We compare against a locally fine-tuned MLLM and a task-specific CNN to examine the trade-off between frontier general purpose models and smaller models that require specific task optimisation but can be used locally. We show that are reaching an inflection point, where frontier MLLMs can now perform effective visual assessment of medical image alignment. Models released only a few months ago generalise poorly and, in some settings, perform barely above chance, whereas GPT-6 achieves over 85% across almost all scenarios tested. Fine-tuned local models can match or exceed frontier-model performance on the tasks on which they are trained, but transfer substantially less effectively to unseen settings. These findings identify medical image alignment as a useful test bed for generalist visual reasoning and suggest that frontier multimodal models are beginning to acquire capabilities that could support a common quality-control mechanism across heterogeneous medical-imaging pipelines.

中文摘要

摘要：前沿多模态大型语言模型（MLLMs）正越来越多地被定位为通用视觉推理器，这是实现人工通用智能的一部分。检验其通用性的重要方法是它们是否能够在没有任务特定参数优化的情况下，依靠视觉证据和任务指令执行人类可以可靠完成的新颖视觉判断。我们通过医学影像对齐评估任务来探讨这一问题，其目标是确定两幅图像之间是否存在解剖学对应关系。人类对图像对齐的视觉评估仍然是黄金标准和最常用的方法；然而，这需要受过训练的操作员，并且对于大数据集来说不具备可扩展性。我们评估了最近几代 MLLMs 在两个医学影像对齐示例任务上的表现，同时变化提示策略和图像展示方法。我们将其与本地微调的 MLLM 和任务特定的卷积神经网络（CNN）进行比较，以研究前沿通用模型与需要特定任务优化但可本地使用的小型模型之间的权衡。我们显示出 MLLMs 正达到一个拐点，前沿 MLLMs 现在可以有效地进行医学影像对齐的视觉评估。几个月前发布的模型泛化能力较差，在某些场景下，其表现几乎仅略高于随机水平，而 GPT-6 在几乎所有测试场景中准确率超过 85%。经过微调的本地模型可以在其训练的任务上匹配或超过前沿模型的表现，但在未知场景中的迁移能力显著较弱。这些发现表明，医学影像对齐是通用视觉推理的有益测试平台，并暗示前沿多模态模型开始具备能够支持异构医学影像处理流程中通用质量控制机制的能力。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06896

Authors: Ross Callaghan, Niannu Gao, Hojjat Azadbakht, Hui Zhang

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.06896.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06896

Published: 2026-10-07T02:42:59.827Z

2. RADC: Risk-Aware Dual Caching for Vision-Language Test-Time Adaptation

Abstract:Cache-based test-time adaptation (TTA) for vision-language models is often hindered by background bias in global representations and unreliable entropy-based cache admission under representation variations. To address these limitations, we propose RADC, which enhances prototype learning through reliable dual caching. RADC introduces a Semantic Foreground Cache that aggregates category-consistent spatial evidence from CLIP representations, yielding foreground prototypes that complement the global cache while mitigating background interference. To reliably manage both caches, Gaussian Risk Admission models multi-view representations as diagonal Gaussian distributions and jointly considers class separation and feature uncertainty to prioritize reliable cache candidates. RADC integrates zero-shot logits with complementary global- and foreground-cache predictions for robust inference. Extensive experiments on cross-domain and out-of-distribution benchmarks demonstrate consistent state-of-the-art performance.

中文摘要

摘要：基于缓存的视觉-语言模型测试时自适应（TTA）经常受到全局表示中的背景偏差和表示变化下基于熵的缓存接纳不可靠的影响。为了解决这些限制，我们提出了RADC，它通过可靠的双缓存增强原型学习。RADC引入了语义前景缓存（Semantic Foreground Cache），从CLIP表示中聚合类别一致的空间证据，生成前景原型，以补充全局缓存，同时减轻背景干扰。为了可靠地管理两个缓存，高斯风险接纳（Gaussian Risk Admission）将多视图表示建模为对角高斯分布，并联合考虑类别分离和特征不确定性，以优先选择可靠的缓存候选项。RADC将零样本logits与全局缓存和前景缓存的互补预测集成，用于稳健推理。在跨领域和分布外基准上的大量实验表明其性能在各方面均为最新最优。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06932

Authors: Siyu Huang, Yueyong Chen, Xuejiao Li, Jun Zhou

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.06932.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06932

Published: 2026-10-07T02:42:59.827Z

3. UniPro: Unified Multi-Mode Medical Image Segmentation from 2D Images to 3D Volumes via Propagation

Abstract:Medical image segmentation remains fragmented along two axes: segmentation paradigms and data dimensionality. Existing methods are typically developed separately for semantic, in-context, and interactive segmentation, and are further specialized to either native 2D images or 3D volumetric data. In clinical practice, however, segmentation workflows take many forms: a case may be initialized by semantic prediction, reference-guided segmentation, or user interaction. Regardless of how it begins, fine-grained refinement is naturally performed on 2D views; for volumetric scans, such 2D edits must propagate coherently to the rest of the volume. We present UniPro, a unified model that bridges segmentation paradigms and data dimensionality, using propagation to extend 2D segmentation to 3D volumes. Our key insight is that volumetric propagation and in-context segmentation share the same reference-conditioned prediction mechanism, differing only in whether the reference image-mask pairs come from other cases or from previously segmented neighboring slices. Building on this view, UniPro supports semantic, in-context, interactive, and propagation-based segmentation within a single slice-based framework, using class priors, reference exemplars, user clicks, and neighboring-slice predictions as mode-specific conditioning inputs. To improve propagation reliability, UniPro further incorporates bidirectional and 3D supervision to regularize slice-wise propagation beyond per-slice losses. Extensive experiments across diverse modalities and anatomies show that UniPro achieves strong performance across all segmentation settings, enabling annotation-efficient 3D segmentation from sparse 2D initialization and reducing slice-by-slice correction effort.

中文摘要

摘要：医学图像分割在两个方面仍然存在分裂：分割范式和数据维度。现有方法通常分别为语义分割、上下文分割和交互式分割开发，并进一步专门针对原生二维图像或三维体数据。然而，在临床实践中，分割工作流程形式多样：一个案例可能由语义预测、参考引导的分割或用户交互初始化。无论如何开始，细粒度的精细化通常在二维视图上进行；对于体积扫描，这种二维编辑必须一致地传播到整个体积。我们提出了UniPro，一个统一模型，桥接分割范式和数据维度，通过传播将二维分割扩展到三维体积。我们的关键见解是，体积传播和上下文分割共享相同的参考条件预测机制，仅在参考图像-掩膜对来自其他案例或来自先前分割的相邻切片时有所不同。在此视角基础上，UniPro在单个切片框架内支持语义分割、上下文分割、交互式分割和基于传播的分割，使用类别先验、参考示例、用户点击和相邻切片预测作为模式特定的条件输入。为提高传播的可靠性，UniPro进一步引入双向和三维监督，以正则化切片间传播，超越每片切片的损失。在多种模态和解剖结构上进行的大量实验表明，UniPro在所有分割设置中均表现出强劲的性能，实现了从稀疏二维初始化到高效注释的三维分割，并减少了逐片修正的工作量。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06938

Authors: Bangwei Guo, Yunhe Gao, Meng Ye, Yang Zhou, Difei Gu, Guoning Zhang, Leon Axel, Dimitris Metaxas

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.06938.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06938

Published: 2026-10-07T02:42:59.827Z

4. Beyond the Linear Representation Hypothesis: Non-Linear Activation Steering in Text-to-Image Models

Abstract:Mechanistic interpretability often relies on the Linear Representation Hypothesis (LRH), which assumes that high-level concepts are encoded as linear directions in activation space. Yet a natural visual concept does not necessarily require a linear visual transition: between sunny and stormy lies an intermediate weather state such as a sky with a few white clouds, not simply a weaker storm; between a caterpillar and a butterfly, the progression is not a caterpillar with continuously growing wings. This raises the question of whether such true intermediate states are also represented nonlinearly by the model. Indeed, when we prompt text-to-image models directly for intermediate attributes, their activations rarely fall along the straight direction connecting the endpoints. Therefore, we propose KANSteer, which models concept traversal as a curve passing through its intermediate states. Seeking a representation that is both simple and interpretable, we propose to use Kolmogorov-Arnold Networks (KANs), which provide a one-dimensional coordinate whose learned functions define the trajectory. This allows the steering direction to vary along the concept while preserving an interpretable representation. Across several concepts and text-to-image diffusion transformers, we find that their activation trajectories substantially deviate from straight lines, and that KANSteer provide a closer fit and smoother traversal of intermediate attributes than linear steering.

中文摘要

摘要：机械可解释性通常依赖于线性表示假设（LRH），该假设认为高层概念在激活空间中以线性方向进行编码。然而，自然的视觉概念并不一定需要线性的视觉过渡：在晴天和暴风雨之间存在一个中间的天气状态，例如有几朵白云的天空，而不仅仅是一个较弱的暴风雨；在毛毛虫和蝴蝶之间，过渡也不是毛毛虫的翅膀不断生长。这就引出了一个问题：这些真实的中间状态是否也被模型以非线性方式表示。事实上，当我们直接提示文本到图像模型生成中间属性时，它们的激活很少沿着连接端点的直线方向变化。因此，我们提出了KANSteer，它将概念遍历建模为经过中间状态的一条曲线。为了寻求既简单又可解释的表示，我们提出使用Kolmogorov-Arnold网络（KANs），它们提供一个一维坐标，其学习函数定义轨迹。这允许在保持可解释表示的同时，使引导方向随着概念而变化。在多个概念和文本到图像扩散变换器中，我们发现它们的激活轨迹明显偏离直线，同时KANSteer相比线性引导提供了更接近的拟合和更平滑的中间属性遍历。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06945

Authors: Muhammad Atif Butt, Paweł Skierś, Joost Van De Weijer, Kamil Deja

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.06945.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06945

Published: 2026-10-07T02:42:59.827Z

5. DistScene: Object-to-Scene Distillation for 3D Scene Generation

Abstract:We present DistScene, a framework for single-image compositional 3D scene generation by jointly modeling the environment and individual objects. Unlike existing methods that represent scenes primarily as collections of objects, we model the environment as an explicit scene component to provide geometric context for object placement. Specifically, we introduce Scene-Frame Generation, which jointly generates separate environment and object components in a shared coordinate frame, allowing their geometry and relative placement to be learned together. Then we introduce Object-Centric Refinement to refine each object in a local frame with scene context. Finally, we develop Object-to-Scene Distillation to transfer pretrained object-generation priors to scene generation through automatically composed and rendered synthetic scenes. Evaluations on indoor and outdoor benchmarks demonstrate improved scene-level spatial coherence over the evaluated baselines. Project page: this https URL

中文摘要

摘要：我们提出了 DistScene，这是一个通过联合建模环境和单个物体的单图像组合式三维场景生成框架。与现有主要将场景表示为物体集合的方法不同，我们将环境建模为显式场景组件，以提供物体摆放的几何上下文。具体来说，我们引入了场景框架生成（Scene-Frame Generation），在共享坐标系中联合生成独立的环境和物体组件，从而使它们的几何形状和相对位置能够被共同学习。然后，我们提出了以物体为中心的优化（Object-Centric Refinement），在局部坐标系中结合场景上下文来优化每个物体。最后，我们开发了物体到场景蒸馏（Object-to-Scene Distillation），通过自动组合和渲染的合成场景将预训练物体生成先验转移到场景生成中。在室内和室外基准数据集上的评估表明，该方法在场景级空间一致性上优于被评估的基线方法。项目页面：该 https URL

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06960

Authors: Kunming Luo, Hongyu Yan, Ken Deng, Chengcheng Zhou, Tianyu Liu, Haipeng Li, Haibin Huang, Xuelong Li, Ping Tan

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.06960.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06960

Published: 2026-10-07T02:42:59.827Z

6. BoT-Feedback: Grounding Multimodal Reasoning in Biomechanical Evidence for Explainable Human Action Feedback

Abstract:Multimodal Large Language Models (MLLMs) have demonstrated impressive capabilities in visual understanding and multimodal reasoning, yet they remain fundamentally limited in Human Action Feedback Generation. Existing methods infer coaching feedback directly from visual observations, producing generic advice, limited interpretability, and physically implausible hallucinations. In contrast, expert human coaches diagnose performance through explicit biomechanical reasoning over joint kinematics, posture, and body dynamics. We introduce BoT-Feedback, a framework that grounds MLLM reasoning in structured biomechanical evidence. Our key contribution is Biomechanics of Thought (BoT), a four-stage reasoning framework that progressively identifies the action, localises the critical body regions, analyses quantitative biomechanical differences between expert and student performances, and synthesises interpretable coaching feedback. To support this reasoning process, we develop a plug-and-play Biomechanical Data Parser (BDP) that converts videos into structured biomechanical descriptors and an alignment strategy that temporally matches expert and student motions. We further introduce BiomAF, a benchmark containing paired teacher-student videos, 3D skeletons, biomechanical attributes, and expert-coaching annotations. Experiments across twelve open- and closed-source MLLMs demonstrate that grounding reasoning in biomechanical evidence consistently improves feedback quality, interpretability, and robustness while substantially reducing biomechanical hallucinations. BoT-Feedback improves the average expert evaluation score from 2.07 to 2.95 (+40%), enabling compact open-source MLLMs to approach the performance of substantially larger proprietary systems for explainable action feedback generation.

中文摘要

摘要：多模态大语言模型（MLLMs）在视觉理解和多模态推理方面展现了令人印象深刻的能力，但在人体动作反馈生成方面仍存在根本性局限。现有方法直接从视觉观察中推断训练反馈，产生通用建议、有限的可解释性以及物理上不可能的幻觉。相比之下，专业人体教练通过对关节运动学、姿势和身体动力学进行明确的生物力学推理来诊断表现。我们提出了 BoT-Feedback，一种将 MLLM 推理建立在结构化生物力学证据上的框架。我们的核心贡献是思维的生物力学（Biomechanics of Thought, BoT），这是一个四阶段推理框架，逐步识别动作、定位关键身体部位、分析专家与学生之间的量化生物力学差异，并综合生成可解释的教学反馈。为了支持这一推理过程，我们开发了一个即插即用的生物力学数据解析器（BDP），将视频转换为结构化生物力学描述符，并设计了一个时间上匹配专家与学生动作的对齐策略。我们进一步引入了 BiomAF，这是一套包含成对的师生视频、3D 骨架、生物力学属性及专家指导注释的基准数据集。在十二种开源和闭源 MLLM 上的实验表明，将推理建立在生物力学证据上能够持续提高反馈的质量、可解释性和稳健性，同时显著减少生物力学幻觉。BoT-Feedback 将专家平均评估得分从 2.07 提升至 2.95（+40%），使紧凑型开源 MLLM 能够在可解释动作反馈生成方面接近大规模专有系统的性能。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06972

Authors: Xu Dong, Wanqing Li, Anthony Adeyemi-Ejeye, Andrew Gilbert

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.06972.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06972

Published: 2026-10-07T02:42:59.827Z

7. Event Cameras for Melt-Pool Monitoring in Additive Manufacturing: A Benchmark and a Cross-Machine Transfer Analysis

Abstract:Melt-pool monitoring is central to qualifying metal additive manufacturing (AM), yet no public event-camera benchmark exists for this domain. Event cameras report per-pixel brightness changes with microsecond timing instead of reading full frames, giving the temporal resolution AM transients demand at a fraction of the data rate. We present SynAM-E (Synthetic AM Events), the first public multi-source simulated event-camera benchmark for metal-AM melt-pool monitoring: 85 physics-calibrated event shards from 15 sources across 8 institutions, with public baselines and fixed cross-machine evaluation splits. On a single-machine case study, event-spatial monitoring matches dense-frame accuracy (0.874 versus 0.863 macro-F1), and the absolute intensity that events discard adds only +0.006 under fusion. On the NIST Additive Manufacturing Metrology Testbed (AMMT) build, a near-sensor event-rate counter recovers a raw-frame-confirmed 528.7 Hz intensity oscillation at ~380 times less sensor readout than the frame stream requires. A compact 93 k-parameter spiking model runs at 15 times lower modeled inference energy for a 0.073 macro-F1 cost. Every cross-source task includes a built-in trust test against camera identity shortcuts: process-type classification passes while material classification remains confounded by camera band, a corpus-structural limitation the release documents and the trust test exposes.

中文摘要

摘要：熔池监测是金属增材制造（AM）验证的核心，但在该领域尚不存在公开的事件相机基准。事件相机以微秒级时间报告每个像素的亮度变化，而不是读取完整帧，这为 AM 瞬态提供了所需的时间分辨率，同时数据量仅为全帧的一小部分。我们提出了 SynAM-E（Synthetic AM Events），这是第一个用于金属 AM 熔池监测的公开多源模拟事件相机基准：来自 8 个机构、15 个来源的 85 个经过物理校准的事件片段，提供公开基线和固定的跨机器评估分割。在单机器案例研究中，事件空间监测的准确性与密集帧匹配（宏 F1 指数 0.874 对 0.863），事件丢失的绝对强度在融合下仅增加 +0.006。在 NIST 增材制造计量测试平台（AMMT）构建中，近传感器事件率计数器以大约帧流所需传感器读取量的 1/380 检测到了经原始帧确认的 528.7 Hz 强度振荡。一个紧凑的 93k 参数脉冲模型在推理能耗降低 15 倍的情况下，仅以 0.073 的宏 F1 损失运行。每个跨源任务都包括对相机身份捷径的内置信任测试：工艺类型分类通过，而材料分类仍受相机频段干扰，这是语料结构的限制，发布说明对此进行了记录，并且信任测试也揭示了这一点。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06973

Authors: Mohamad Yazan Sadoun, Sarah Sharif, Yingtao Liu, Zahed Siddique, Yaser Mike Banad

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.06973.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06973

Published: 2026-10-07T02:42:59.827Z

8. Visual-Invariance-Augmented Feature Optimal Alignment for Transferable Adversarial Attacks against Closed-Source MLLMs

Abstract:Multimodal large language models (MLLMs) remain vulnerable to transferable adversarial examples, especially in black-box settings where only open-source surrogate models are accessible. Existing targeted transfer attacks mainly align adversarial and target samples using global image-level features, such as encoder [CLS] embeddings. However, such coarse alignment insufficiently exploits patch-level visual structures, limiting transferability across heterogeneous closed-source MLLMs. We propose IAU-FOA, a visual-invariance-augmented feature optimal alignment attack with adaptive unbalanced transport, to improve targeted transferability against closed-source MLLMs. IAU-FOA aligns adversarial and target samples at both global and local levels: a cosine-based objective narrows their global semantic gap, while patch tokens are clustered into compact local patterns and matched through optimal transport for fine-grained feature alignment. Balanced optimal transport enforces fixed marginal masses even for local clusters without reliable counterparts, potentially introducing misleading alignment gradients. We therefore introduce confidence-adaptive unbalanced transport to relax these constraints for weakly matched clusters, aiming to reduce unreliable local alignment and improve adversarial transferability. We further study the effect of input transformations and propose visual-invariance augmentation, which applies bidirectional pixel-intensity rescaling and per-channel white-balance adjustment to simulate exposure, contrast, illumination, and color-temperature variations. This strategy encourages adversarial perturbations to generalize across different visual encoders. Extensive experiments on open-source and closed-source MLLMs show that IAU-FOA consistently outperforms state-of-the-art transferable attack methods. Code is available at this https URL.

中文摘要

摘要：多模态大语言模型（MLLMs）仍然容易受到可迁移的对抗样本攻击，尤其是在仅能访问开源替代模型的黑箱环境下。现有的目标迁移攻击主要使用全局图像级特征（如编码器的 [CLS] 嵌入）来对齐对抗样本和目标样本。然而，这种粗粒度对齐未能充分利用局部图块级视觉结构，从而限制了在异构闭源 MLLMs 上的迁移能力。我们提出 IAU-FOA，一种结合自适应非平衡传输的视觉不变增强特征最优对齐攻击，以提升对闭源 MLLMs 的目标迁移能力。IAU-FOA 在全局和局部层面对对抗样本和目标样本进行对齐：基于余弦的目标函数缩小它们的全局语义差距，而将图块令牌聚成紧凑的局部模式，并通过最优传输进行精细特征对齐。即使对于没有可靠对应的局部簇，平衡最优传输也会强制固定边际质量，可能引入误导的对齐梯度。因此，我们引入信心自适应非平衡传输，以放宽对弱匹配簇的约束，旨在减少不可靠的局部对齐并提高对抗迁移性。我们进一步研究输入变换的影响，并提出视觉不变增强，通过双向像素强度重缩放和每通道白平衡调整来模拟曝光、对比度、照明和色温变化。这一策略鼓励对抗扰动在不同视觉编码器上具有泛化能力。在开源和闭源 MLLMs 上的大量实验表明，IAU-FOA 一直优于现有最先进的可迁移攻击方法。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: 2610.06977

Authors: Xiaojun Jia, Simeng Qin, Yiming Li, Jie Liao, Sensen Gao, Ke Ma, Yang Liu, Xiaochun Cao

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.06977.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06977

Published: 2026-10-07T02:42:59.827Z

9. Hierarchy-GBP: Accelerating Factor Graph Inference via Abstraction and Recovery

Abstract:Gaussian Belief Propagation (GBP) is a distributed inference algorithm that passes messages in graphical models, making it attractive for scalable spatial intelligence. However, we find GBP most effective locally: it rapidly smooths message errors that vary sharply between neighbor variables, but corrects global errors across distant graph regions incrementally through long-range message propagations. We propose Hierarchy-GBP (H-GBP), an iterative, two-stage framework that accelerates GBP by first solving these global errors with a coarse graph approximation (abstraction) and projecting the results back to the original graph (recovery), then refining the remaining local errors with GBP. We prove H-GBP convergence to the optimum by deriving the combined matrix operator of our abstraction and recovery steps and analyzing its spectral radius. Experiments on linear sparse graphs show that H-GBP converges fundamentally faster than standard GBP. Moreover, we validate H-GBP on two important spatial problems: Pose Graph Optimization (PGO) and Bundle Adjustment (BA). H-GBP markedly accelerates large-scale PGO and achieves state-of-the-art runtime across all tested BA scales.

中文摘要

摘要：高斯信念传播（Gaussian Belief Propagation, GBP）是一种在图模型中传递信息的分布式推理算法，使其在可扩展空间智能中具有吸引力。然而，我们发现GBP在局部最为有效：它能够快速平滑邻近变量之间急剧变化的信息误差，但通过远程信息传播仅能逐步纠正远距离图区域的全局误差。我们提出了分层高斯信念传播（Hierarchy-GBP, H-GBP），这是一种迭代的两阶段框架，通过先使用粗略图近似（抽象）解决这些全局误差并将结果投影回原始图（恢复），然后用GBP细化剩余的局部误差，从而加速GBP。我们通过推导抽象和恢复步骤的组合矩阵算子并分析其谱半径，证明了H-GBP收敛到最优解。在线性稀疏图上的实验表明，H-GBP的收敛速度明显快于标准GBP。此外，我们在两个重要的空间问题上验证了H-GBP：姿态图优化（Pose Graph Optimization, PGO）和束调整（Bundle Adjustment, BA）。H-GBP显著加速了大规模PGO，并在所有测试的BA规模中实现了最先进的运行时间。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06978

Authors: Yuzhou Cheng, Tom Yates, Ignacio Alzugaray, Danyal Akarca, Pedro A. M. Mediano, Andrew J. Davison

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.06978.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06978

Published: 2026-10-07T02:42:59.827Z

10. State-Aware Interaction MIL for Rare Joint Molecular Phenotype Prediction in Colorectal Cancer and Lung Adenocarcinoma

Abstract:Joint molecular phenotype prediction is complicated by small joint-positive populations and overlapping histological features across alternative molecular states. Existing computational pathology approaches typically predict biomarkers independently or formulate the joint-positive phenotype as a binary endpoint. Independent prediction does not model interactions between biomarker-specific histological representations, whereas binary joint prediction collapses the double-negative and two single-positive configurations into a single negative class. We propose State-Aware Interaction MIL, a weakly supervised method that preserves biomarker-specific histological representations, models their interaction, and supervises the complete four-state molecular configuration. We evaluate the proposed approach for joint BRAF+/MSI+ prediction in colorectal cancer and EGFR+/TP53+ prediction in lung adenocarcinoma using frozen UNI2-h and CONCH pathology foundation-model representations. With UNI2-h, State-Aware Interaction MIL achieved an average precision of 0.5566 in colorectal cancer (joint-positive prevalence 6.8%) compared with 0.5161 for NaiveMTL, and 0.2784 in lung adenocarcinoma (joint-positive prevalence 8.6%) compared with 0.2525 for IndependentPair. With CONCH, State-Aware achieved an average precision of 0.4410 compared with 0.3932 for DirectJoint in colorectal cancer and 0.1659 compared with 0.1226 for DirectJoint in lung adenocarcinoma. These results indicate that pathology foundation-model representations contain predictive information for rare joint molecular phenotypes and that preserving biomarker-specific representations within a structured molecular-state formulation can improve prediction of these phenotypes from histopathology.

中文摘要

摘要：联合分子表型预测因联合阳性群体小及不同分子状态间组织学特征重叠而变得复杂。现有的计算病理学方法通常独立预测生物标志物或将联合阳性表型制定为二元终点。独立预测无法对生物标志物特异性组织学表示之间的相互作用建模，而二元联合预测将双阴性和两个单阳性配置合并为单一的阴性类。我们提出了状态感知交互MIL（State-Aware Interaction MIL），这是一种弱监督方法，能够保留生物标志物特异性组织学表示，对其相互作用建模，并监督完整的四状态分子配置。我们使用冻结的UNI2-h和CONCH病理基础模型表示，评估了该方法在结直肠癌的BRAF+/MSI+联合预测以及肺腺癌的EGFR+/TP53+联合预测中的效果。在UNI2-h下，状态感知交互MIL在结直肠癌中取得了平均精度0.5566（联合阳性率6.8%），相比之下，NaiveMTL的平均精度为0.5161；在肺腺癌中取得了平均精度0.2784（联合阳性率8.6%），相比之下，IndependentPair的平均精度为0.2525。在CONCH下，状态感知MIL在结直肠癌中取得了平均精度0.4410，相比DirectJoint的0.3932，在肺腺癌中取得了平均精度0.1659，相比DirectJoint的0.1226。这些结果表明，病理基础模型表示包含对罕见联合分子表型的预测信息，并且在结构化分子状态框架内保留生物标志物特异性表示可以提高从组织病理学中预测这些表型的能力。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.06991

Authors: Dasari Naga Raju, Tripti Bameta

Categories: cs.CV

PDF URL: https://arxiv.org/pdf/2610.06991.pdf

CoolPaper URL: https://papers.cool/arxiv/2610.06991

Published: 2026-10-07T02:42:59.827Z

ArXiv Domain 2026-10-07

LLM Domain Papers

1. Zero-Shot Visualization: Exploring Text Corpora with User-Prompted Axes

2. Capacity, Responsiveness and Alignment: What Makes a Latent Structure Actionable

3. Tree Navigation Without LLM Summaries: A Matched-Cost Study of Hierarchical Retrieval for Long-Document QA

4. Component and Dimension Sparsity in Transformer Refusal Mechanisms

5. Stabilizing language models under continual learning via condition-anchored distillation

6. EMODE: Dynamic Para-Semantic Experts for Emotion-Aware Speech Language Modeling

7. Verdicts Without Annotated Evidence: Rejection Sampling or Label-Only Post-Training for Evidence Recovery?

8. WavePrune: One period is often enough for RoPE

9. Calibrated Answers About Randomized Trials From a 4-Billion-Parameter Open Model: A Registered Test and a License-Clean Release

10. Investigating Model Compression for Neural Machine Translation in the Biomedical Domain

Agent Domain Papers

1. GAMEGO: Training Game-Dev Agents with Synthetic Trajectories Anchored in Real-World Assets

2. Text2Dashboard: A Governed Agent Architecture for Natural-Language Dashboard Generation over Enterprise DataBrain

3. FluidPD: In-Place Elasticity for SLO-Aware Prefill-Decode Disaggregated LLM Serving

4. Anchor Divergence for Semantic Geometry in Contrastive Learning

5. RadOnc-Agent: An LLM-Orchestrated Framework for AI Workflows Across the Radiotherapy Care Pathway

6. Metonymic Circuits for Abstract Concept Grounding in Vision Transformers

7. Principles that Guide, Actions that Inform: Agent Evolution via Knowledge Abstraction

8. AegisFlow: A Multi-Agent Agentic AI Framework for Autonomous Remediation and Self-Healing in Fragile Data Ecosystems

9. EPOCH: Reliable Discovery through Evidence-Governed Search

10. When better traffic forecasts fail to improve signal control: a layered diagnostic study of forecast-to-decision value

Evaluation Domain Papers

1. GAMEGO: Training Game-Dev Agents with Synthetic Trajectories Anchored in Real-World Assets

2. Text2Dashboard: A Governed Agent Architecture for Natural-Language Dashboard Generation over Enterprise DataBrain

3. FluidPD: In-Place Elasticity for SLO-Aware Prefill-Decode Disaggregated LLM Serving

4. Anchor Divergence for Semantic Geometry in Contrastive Learning

5. RadOnc-Agent: An LLM-Orchestrated Framework for AI Workflows Across the Radiotherapy Care Pathway

6. Metonymic Circuits for Abstract Concept Grounding in Vision Transformers

7. Principles that Guide, Actions that Inform: Agent Evolution via Knowledge Abstraction

8. AegisFlow: A Multi-Agent Agentic AI Framework for Autonomous Remediation and Self-Healing in Fragile Data Ecosystems

9. EPOCH: Reliable Discovery through Evidence-Governed Search

10. When better traffic forecasts fail to improve signal control: a layered diagnostic study of forecast-to-decision value

VLM Domain Papers

1. Medical Image Alignment Assessment as a Test of Generalist Visual Reasoning in Frontier Multimodal Models

2. RADC: Risk-Aware Dual Caching for Vision-Language Test-Time Adaptation

3. UniPro: Unified Multi-Mode Medical Image Segmentation from 2D Images to 3D Volumes via Propagation

4. Beyond the Linear Representation Hypothesis: Non-Linear Activation Steering in Text-to-Image Models

5. DistScene: Object-to-Scene Distillation for 3D Scene Generation

6. BoT-Feedback: Grounding Multimodal Reasoning in Biomechanical Evidence for Explainable Human Action Feedback

7. Event Cameras for Melt-Pool Monitoring in Additive Manufacturing: A Benchmark and a Cross-Machine Transfer Analysis

8. Visual-Invariance-Augmented Feature Optimal Alignment for Transferable Adversarial Attacks against Closed-Source MLLMs

9. Hierarchy-GBP: Accelerating Factor Graph Inference via Abstraction and Recovery

10. State-Aware Interaction MIL for Rare Joint Molecular Phenotype Prediction in Colorectal Cancer and Lung Adenocarcinoma