Daily Research Digest
arXiv Papers
2026-08-17
272
Papers
9
Categories
53
Translated
收藏清单 0
精选 · Favorites
53
cs.AI / 1 / 2608.13767
Simulation-Aware In-Context Policy Improvement for LLM-Aided Analog Layout Refinement
面向LLM辅助模拟版图优化的仿真感知上下文策略改进
large language model
大语言模型相关
Abstract
Analog IC layout design remains a labor-intensive iterative process dominated by simulation-driven refinement. Although end-to-end layout generators accelerate initial placement and routing, they still require experts to manually tune layout optimization parameters with repeated post-layout simulations for stringent design specifications. While Bayesian Optimization (BO) is widely adopted for parameter tuning in analog IC design, at the layout level it typically requires hundreds to thousands of evaluations, each involving costly parasitic extraction and post-layout simulation, which makes it impractical. Recently, Large Language Models (LLMs) have demonstrated potential in improving the sample efficiency of such simulation-driven tuning. However, their restricted access to geometric layout context and design-specific heuristics limits their ability to manipulate the layout optimization process. In this paper, we propose a simulation-aware LLM multi-agent framework that performs in-context policy improvement (ICPI) by iteratively updating layout optimization parameters exposed by an analog layout generator through an act-observe-reflect loop on compact structured layout representations. Experiments on real-world analog circuits show that, with only tens of post-layout simulations, our approach improves post-layout performance over the generator's built-in heuristics and BO-based tuning method.
Chinese Translation
模拟IC版图设计仍然是一个劳动密集型的迭代过程，以仿真驱动的优化为主。虽然端到端版图生成器加速了初始布局和布线，但对于严格的设计规格，它们仍然需要专家通过重复的后仿真来手动调整版图优化参数。尽管贝叶斯优化（BO）在模拟IC设计中广泛用于参数调整，但在版图层面，它通常需要数百到数千次评估，每次评估都涉及昂贵的寄生提取和后仿真，这使得它不切实际。最近，大语言模型（LLMs）在提高此类仿真驱动调优的样本效率方面展现出潜力。然而，它们对几何版图上下文和设计特定启发式方法的访问受限，限制了其操纵版图优化过程的能力。在本文中，我们提出了一种仿真感知的LLM多智能体框架，该框架通过在一个紧凑的结构化版图表示上执行“行动-观察-反思”循环，迭代更新模拟版图生成器所暴露的版图优化参数，从而实现上下文策略改进（ICPI）。在真实模拟电路上的实验表明，仅需数十次后仿真，我们的方法就能在版图后性能上优于生成器的内置启发式方法和基于BO的调优方法。
cs.AI / 2 / 2608.13820
SDO: Subspace Deconflicting Operator for Multi-Adapter Composition
SDO：用于多适配器组合的子空间去冲突算子
diffusion
扩散模型相关
Abstract
Composing independently trained adapters within a shared diffusion backbone provides a modular approach to multi-character generation, but naive joint deployment often causes identity mixing, cross-character attribute leakage, and unstable scene composition. We study this interference from a parameter-space perspective and hypothesize that it arises partly from conflicts between overlapping dominant subspaces in shared layers. To address this issue, we propose \textbf{SDO}, a \textbf{S}ubspace \textbf{D}econflicting \textbf{O}perator for multi-adapter composition. SDO reconstructs layer-wise low-rank updates from the selected adapters, extracts compact subspace signatures, measures pairwise conflict through output-subspace overlap, and applies a permutation-equivariant transformation that suppresses harmful shared directions while retaining identity-specific characteristics. The resulting representations are mapped back to standard adapter updates and can be directly incorporated into existing diffusion inference pipelines. Experiments demonstrate that SDO consistently improves identity fidelity and compositional stability, with particularly clear gains as the number of jointly composed adapters increases.
Chinese Translation
在共享扩散骨干网络内组合独立训练的适配器，为多角色生成提供了一种模块化方法，但简单的联合部署常常导致身份混合、跨角色属性泄漏以及不稳定的场景组合。我们从参数空间的角度研究这种干扰，并假设其部分源于共享层中重叠主导子空间之间的冲突。为解决这一问题，我们提出了\textbf{SDO}，一种用于多适配器组合的\textbf{子}空间\textbf{去冲}突\textbf{算}子。SDO 从选定的适配器中重构逐层低秩更新，提取紧凑的子空间签名，通过输出子空间重叠度量两两冲突，并应用置换等变变换，在保留身份特定特征的同时抑制有害的共享方向。得到的表示被映射回标准适配器更新，并可直接集成到现有的扩散推理流程中。实验表明，SDO 持续提高了身份保真度和组合稳定性，并且随着联合组合适配器数量的增加，其增益尤为明显。
cs.AI / 3 / 2608.13926
Never the Number: Structural Abstention for AI Systems Whose Answers Are Consumed as Fact
绝非数字：答案被当作事实消费的AI系统的结构性弃权
large language model
大语言模型相关
Abstract
Large language models have made natural language interfaces to databases (NLIDB) newly credible, but LLM text-to-SQL systems fail in a way that matters for deployment: a hallucinated column or a mis-aggregated total yields a fluent wrong answer, indistinguishable at the point of use from a right one. Where the consumer cannot inspect the generated query, as in enterprise AI deployments and operational dashboards, and increasingly where the consumer is a tool-using agent rather than a person, accuracy alone is insufficient: nothing marks which answers to distrust. This is a reliability problem before it is an accuracy problem. We propose an architectural pattern for such systems, a trusted kernel with a generative shell, resting on one invariant: a component that can fabricate may influence which question the system answers, never which value it returns. A generative shell interprets underspecified input and phrases replies; a deterministic kernel matches fully specified questions against a bounded set of answerable question shapes and compiles them to queries by deterministic execution. The two meet at a confirmation the user reads before any value is computed, and requests the kernel cannot express are declined rather than approximated. We call this structural abstention, and distinguish it from the statistical abstention of selective prediction and calibrated confidence: refusal here needs no confidence estimate, because unanswerable requests are unrepresentable. We specify the pattern implementation-independently, give a five-decision recipe and work it across three domains, extend the invariant from returned values to the actions of agentic systems, and report a two-year production case study alongside two generative alternatives, a fine-tuned parser and a tool-retrieval agent. We close against enterprise and reliability benchmarks published since.
Chinese Translation
大型语言模型使数据库的自然语言接口（NLIDB）重新变得可信，但LLM文本到SQL系统以一种对部署至关重要的方式失败：一个幻觉列或一个错误汇总的总数会产生流利但错误的答案，在使用的点上与正确答案无法区分。当消费者无法检查生成的查询时，例如在企业AI部署和运营仪表盘中，并且越来越多的情况是消费者是使用工具的智能体而非人类，仅靠准确性是不够的：没有任何标记表明哪些答案不可信。在成为准确性问题之前，这首先是一个可靠性问题。我们为这类系统提出了一种架构模式——一个带生成外壳的可信内核，它基于一个不变式：能够编造的组件可以影响系统回答哪个问题，但绝不能影响它返回哪个值。生成外壳解释未充分指定的输入并组织回复的措辞；确定性内核将完全指定的问题与一组有限的可回答问题形状进行匹配，并通过确定性执行将它们编译为查询。两者在用户计算任何值之前读取的确认处汇合，内核无法表达的请求会被拒绝，而不是被近似处理。我们将此称为结构性弃权，并将其与选择性预测和校准置信度的统计性弃权区分开来：这里的拒绝不需要置信度估计，因为无法回答的请求是不可表示的。我们以与实现无关的方式指定该模式，给出一个五决策配方，并在三个领域中进行实践，将不变式从返回值扩展到智能体系统的行动，并报告一个为期两年的生产案例研究，同时提供两种生成式替代方案：一个微调解析器和一个工具检索智能体。我们最后针对此后发布的企业和可靠性基准进行验证。
cs.AI / 4 / 2608.14014
Buy the Rumor, Sell the News: When Is News Priced In?
买传闻，卖新闻：新闻何时被计入价格？
large language model
大语言模型相关
Abstract
Two old market sayings hold that news is already priced in by the time it is published, and that the rumor is bought while the news is sold. Both place the price move associated with a piece of news before and at publication rather than after it. Whether the claims hold, for which kinds of news, and by how much are basic questions about how fast markets absorb public information. We test them on 4.57 million financial news articles covering roughly 3,000 US stocks (2023-2026). A large language model teacher, distilled into a compact classifier through active learning, assigns each article one of 17 event tags and five attributes; articles are clustered into stories to separate first reports from follow-up coverage; and beta-adjusted abnormal returns are measured around the resulting 1.68 million stock-day events, with 364,405 neutral-sentiment events as a placebo group. Three results follow. First, the price move associated with news concentrates before and at publication: pooled across all signed events, the cumulative move in the news direction by the close of publication day is 2.8 times its value 20 days later, and for rumor-flagged events the rumor day captures the entire move while the subsequent confirmation contributes nothing. Second, measured against the placebo of comparable stocks, markets underreact to numbers and overreact to stories: quantified fundamental news (earnings, dividends, guidance, analyst actions) keeps drifting in the direction of the news for weeks, while soft story-driven news (launches, macro commentary, leadership) gives back its move. Third, news carries width as well as direction: publicity raises volatility before the publication day, and volatility declines once the news is out, because publication resolves uncertainty. The study also produces a table of measured drift for each event tag, usable as a prior in news-conditioned forecasting models.
Chinese Translation
两句古老的市场谚语认为，新闻在发布之时已被计入价格，且谣言被买入、新闻被卖出。两者都将与特定新闻相关的价格变动放在发布之前和发布之时，而不是发布之后。这些说法是否成立、对哪些类型的新闻成立、以及在多大程度上成立，是关于市场吸收公共信息速度的基本问题。我们在覆盖约3,000只美国股票（2023-2026年）的457万篇金融新闻文章上对它们进行了检验。一个通过主动学习蒸馏为紧凑分类器的大型语言模型教师，为每篇文章分配17个事件标签之一和五个属性；文章被聚类为新闻报道，以区分首次报道与后续报道；并围绕由此产生的168万个股票-日事件测量β调整后的异常收益，其中364,405个中性情绪事件作为安慰剂对照组。接下来是三个结果。第一，与新闻相关的价格变动集中在发布之前和发布之时：将所有有方向的事件汇总后，在发布日收盘时沿新闻方向的累计变动是其20天后数值的2.8倍；对于被标记为传闻的事件，传闻日捕捉了全部变动，而随后的确认没有任何贡献。第二，以可比股票的安慰剂为对照，市场对数字反应不足，对故事反应过度：量化的基本面新闻（盈利、股息、指引、分析师行动）在数周内持续沿新闻方向漂移，而软性故事驱动新闻（发布、宏观评论、领导层变动）则回吐其变动。第三，新闻不仅具有方向性，也具有宽度：在发布日之前，公开度提高波动性；一旦新闻发布，波动性下降，因为发布消除了不确定性。该研究还为每个事件标签产出了一张实测漂移表，可用作新闻条件预测模型中的先验。
cs.AI / 5 / 2608.14035
Agent-Orchestration in Autonomous Chip Design
自主芯片设计中的智能体编排
large language model
大语言模型相关
Abstract
Recent developments in large language models (LLMs) and tool-using agents encourage people to explore the potential of using agents in chip design. The core question is what kind of AI we really need in such a sophisticated industry. To this end, we bring the idea of modeling a chip-design superintelligence as an enormous \textit{AI-organization}.
Chinese Translation
大语言模型（LLM）和工具使用智能体的最新发展鼓励人们探索在芯片设计中使用智能体的潜力。核心问题是，在这样一个高度复杂的行业中，我们究竟需要什么样的人工智能。为此，我们提出了将芯片设计超级智能建模为一个庞大的“AI-组织”的想法。
cs.AI / 6 / 2608.14071
Scaling Domain Data Repetition in LLM Pretraining
扩展LLM预训练中的领域数据重复
large language model
大语言模型相关
Abstract
As large language models scale, their training-token budgets must also increase to maintain an appropriate tokens-per-parameter ratio (\(\mathrm{TPP}\)). However, high-quality domain data is much harder to scale than general web data. As model size and the training-token budget increase, its fraction in the training mixture tends to decrease. Repeating the available high-quality data provides an effective way to counteract this dilution, but excessive repetition may lead to overfitting. We study this trade-off under practical LLM scaling, where the training-token budget grows proportionally with model size. For a fixed domain, we first find that, surprisingly at a fixed \(\mathrm{TPP}\), the optimal repetition count mildly increases with model size. Across different domains, we find that the optimal repetition count is strongly negatively correlated with the final validation loss of a domain: domains with lower loss can generally benefit from more repetitions. In contrast, the amount of unique domain data is only weakly related to the optimal repetition count. These findings suggest that repetition counts tuned on smaller proxy models with the same \(\mathrm{TPP}\) can provide a practical estimate for larger models.
Chinese Translation
随着大语言模型的扩展，其训练token预算也必须增加，以维持适当的每参数token数比例（\(\mathrm{TPP}\)）。然而，高质量领域数据比通用网页数据更难扩展。随着模型规模和训练token预算的增加，其在训练混合数据中的占比往往趋于下降。重复利用现有的高质量数据是抵消这种稀释效应的有效方法，但过度重复可能导致过拟合。我们在实际的LLM扩展场景下研究这一权衡，其中训练token预算随模型规模成比例增长。对于固定领域，我们首先发现，在固定的\(\mathrm{TPP}\)下，最优重复次数随模型规模温和增加，这令人意外。在不同领域之间，我们发现最优重复次数与某一领域的最终验证损失强烈负相关：损失较低的领域通常可以从更多次重复中受益。相比之下，唯一领域数据的数量与最优重复次数的关系仅较弱。这些发现表明，在具有相同\(\mathrm{TPP}\)的较小代理模型上调优的重复次数，可以为更大模型提供实用的估计。
cs.AI / 7 / 2608.14089
Regime-Conditional Verification: Correctness Estimation for Adapting and Monitoring Safety Classifiers
条件式验证：用于适配和监控安全分类器的正确性估计
large language model
大语言模型相关
Abstract
Safety classifiers deployed with large language models often fail for two reasons: their decisions reflect the policy learned during training rather than the deployer's desired policy, and their performance degrades as deployment traffic evolves. We present Regime-Conditional Verification (RCV), a lightweight wrapper that adapts an off-the-shelf safety classifier without retraining it. RCV estimates, from the classifier's internal representations, the probability that each prediction disagrees with the deployer's policy, and selectively corrects predictions likely to be wrong. The same correctness estimates also provide a label-free signal for detecting distribution shift, enabling a maintenance loop that updates the correctness estimation layer and resorts to classifier fine-tuning only when necessary. Across three off-the-shelf safety classifiers and two benchmark datasets, RCV improves adherence to the deployer's policy in every classifier-dataset combination, catching up to 0.81 of previously missed unsafe content without modifying the underlying classifier. In a deployment study with ten attack campaigns, each a harm category held out of RCV's training, RCV detects every campaign in a dedicated injection panel; in the maintenance census most drift episodes are repaired without updating the classifier, and the fine-tune is reserved for the residual episodes that repair does not restore.
Chinese Translation
随大型语言模型部署的安全分类器常因两个原因失效：其决策反映的是训练期间学习的策略，而非部署者期望的策略；并且随着部署流量演变，其性能会下降。我们提出条件式验证（RCV），这是一种轻量级包装器，无需重新训练即可适配现成的安全分类器。RCV从分类器的内部表征中估计每个预测与部署者策略不一致的概率，并选择性地纠正可能错误的预测。相同的正确性估计还提供了用于检测分布漂移的无标签信号，从而支持一个维护循环：更新正确性估计层，并仅在必要时才诉诸于分类器微调。在三个现成安全分类器和两个基准数据集上，RCV在每个分类器-数据集组合中均提升了对部署者策略的遵守度，在不修改底层分类器的情况下，捕获了此前遗漏的不安全内容中高达0.81的比例。在一项包含十个攻击活动的部署研究中，每个活动都是RCV训练中未包含的危害类别，RCV在专门的注入面板中检测到了每一个活动；在维护普查中，大多数漂移事件无需更新分类器即可修复，而微调仅保留用于修复无法恢复的残余事件。
cs.AI / 8 / 2608.14152
Towards Efficient Multimodal and Multilingual Opinion Extraction for STI: A QLoRA-Based Fine-Tuning Approach
迈向高效的STI多模态与多语言意见抽取：一种基于QLoRA的微调方法
large language model
大语言模型相关
Abstract
Recent advances in large language models (LLMs) have reshaped semantic analysis. Opinion Extraction (OE) for Science and Technology Intelligence (STI) requires concise core opinions from large information streams. Off-the-shelf models struggle to filter noise from these streams and show limited structured-output reliability in zero-shot multilingual and multi-modal settings. To address information overload and extraction defocus, this study proposes a multimodal core-opinion extraction framework in which visual evidence serves as a contextual anchor for textual judgment. Using VideoLLaMA2 (VL2) and VideoLLaMA2.1 (VL2.1) as the base models, we apply Quantized Low-Rank Adaptation (QLoRA) fine-tuning on a curated dataset of 2,194 multilingual and multimodal samples. Under the selected Image-Augmented setting, fine-tuned VL2.1 generates structured JSON core-opinion outputs, achieving 64.98% Precision, 42.15% Recall, 51.14% F1-score, and 74.00% sample-level accuracy. Relative to the zero-shot VL2.1 setting, it raises the F1-scores of Spanish and Russian from 4.83% and 0.45% to 46.05% and 51.93%, respectively. The framework further incorporates a Fuzzy Cumulative Prospect Theory-based post-extraction triage module for case-level value assessment, providing a case-level value signal for downstream STI screening.
Chinese Translation
大型语言模型（LLM）的最新进展重塑了语义分析。面向科技情报（STI）的意见抽取（OE）需要从大量信息流中提炼简洁的核心意见。现成模型难以过滤这些信息流中的噪声，并且在零样本多语言和多模态场景下表现出有限的结构化输出可靠性。为了解决信息过载和抽取失焦问题，本研究提出了一种多模态核心意见抽取框架，其中视觉证据作为文本判断的上下文锚点。以VideoLLaMA2（VL2）和VideoLLaMA2.1（VL2.1）作为基础模型，我们在一个包含2,194个多语言和多模态样本的精选数据集上应用量化低秩适配（QLoRA）微调。在选定的图像增强设置下，微调后的VL2.1生成结构化JSON核心意见输出，实现了64.98%的精确率、42.15%的召回率、51.14%的F1分数以及74.00%的样本级准确率。相对于零样本VL2.1设置，它将西班牙语和俄语的F1分数分别从4.83%和0.45%提升至46.05%和51.93%。该框架进一步引入了一个基于模糊累积前景理论的后抽取分诊模块，用于案例级价值评估，为下游STI筛选提供案例级价值信号。
cs.AI / 9 / 2608.14179
Can Language Models Understand mmWave Data? Benchmarking Large Language Models for mmWave Radar-Based Human Understanding
语言模型能理解毫米波数据吗？面向基于毫米波雷达的人类理解的大语言模型基准测试
large language model
大语言模型相关
Abstract
Large language models (LLMs) have shown remarkable reasoning and generative capabilities, motivating their use as universal reasoning engines for perception. While modern approaches such as vision-language models (VLMs) have attempted to incorporate reasoning capabilities into visual sensing, the integration of LLMs with the millimeter-wave (mmWave) modality-despite its unique advantages under low light and occlusion-remains largely unexplored. The principal bottlenecks stem from the scarcity of radar language pairs, severe cross-dataset heterogeneity, and the absence of a foundational mmWave encoder. We address this gap through a minimal textualization interface that serializes each mmWave point cloud into concise natural language, allowing off-the-shelf LLMs to operate in a question answering (QA) setting. Building on this, we present mmWave-QA, the first benchmark for language-conditioned mmWave human perception. mmWave-QA aggregates heterogeneous public mmWave datasets and harmonizes them via calibration-aware preprocessing and global taxonomy alignment, while providing natural language QA. Spanning six scenarios and five QA tasks, the benchmark enables standardized evaluation across diverse mmWave hardware and experimental conditions, establishing a foundation for scalable research on mmWave-LLM integration. We further evaluate and analyze LLMs on our mmWave-QA, highlighting their zero-shot reasoning potential for radar perception, as well as their robustness under visual degradation.
Chinese Translation
大语言模型（LLMs）展现了卓越的推理和生成能力，促使它们被用作感知领域的通用推理引擎。尽管诸如视觉-语言模型（VLMs）等现代方法已尝试将推理能力融入视觉感知，但大语言模型与毫米波（mmWave）模态的集成——尽管该模态在低光照和遮挡条件下具有独特优势——在很大程度上仍未得到探索。主要瓶颈源于雷达-语言对的稀缺、跨数据集的高度异质性以及基础毫米波编码器的缺失。我们通过一个极简文本化接口来弥合这一差距，该接口将每个毫米波点云序列化为简洁的自然语言，从而使现成的大语言模型能够在问答（QA）场景中运行。在此基础上，我们提出了 mmWave-QA，这是首个面向语言条件化的毫米波人体感知的基准。mmWave-QA 聚合了异构的公开毫米波数据集，并通过考虑标定的预处理和全局分类对齐来协调它们，同时提供自然语言问答。该基准涵盖六个场景和五个问答任务，能够在多样化的毫米波硬件和实验条件下实现标准化评估，为毫米波-大语言模型集成的可扩展研究奠定基础。我们进一步在 mmWave-QA 上评估和分析了大语言模型，突出了它们在雷达感知方面的零样本推理潜力，以及在视觉退化条件下的鲁棒性。
cs.AI / 10 / 2608.14212
APTER: Adaptive Post-Training with Expert-Grounded Rubrics
APTER：基于专家依据的评分标准的自适应后训练
large language model
大语言模型相关
Abstract
As large language models enter professional domains, they must satisfy domain constraints, include critical evidence, and provide complete reasoning rather than merely produce fluent responses. Existing post-training methods often rely on holistic preferences or outcome-level verification, while recent rubric-based methods usually generate rubrics independently for each query. In specialized domains, such unconstrained rubrics may omit critical requirements and vary across samples, hindering the diagnosis and targeted repair of persistent capability deficiencies. We propose APTER (Adaptive Post-Training with Expert-Grounded Rubrics), a framework that integrates structured domain knowledge into fine-grained evaluation, optimization, and diagnosis for specialized complex reasoning. First, expert-grounded rubric construction starts from an expert criteria framework built by domain experts, where each criterion represents a stable professional capability. For each query, APTER selects relevant criteria and instantiates them into query-level rubrics linked to their source criteria, turning reusable expert criteria into executable query-level supervision without reference answers. Second, adaptive post-training uses rubric verdicts as both optimization and criterion-level diagnostic signals. Aggregating low-scoring verdicts by criterion ID reveals persistent deficiencies and triggers targeted supervised fine-tuning updates during reinforcement learning. Experiments on mathematical reasoning and medical question answering show consistent gains across both domains. Across three model generations, APTER improves the mathematics and medical averages over the corresponding base models by up to 15.86 and 8.04 points, respectively. Code and rubric datasets are available at https://github.com/AntDT-APTER/APTER.
Chinese Translation
随着大型语言模型进入专业领域，它们必须满足领域约束、包含关键证据并提供完整的推理，而不仅仅是产生流畅的回复。现有的后训练方法通常依赖于整体偏好或结果级验证，而最近基于评分标准的方法通常为每个查询独立生成评分标准。在专业领域中，这种不受约束的评分标准可能会遗漏关键要求，并在不同样本之间产生差异，从而阻碍对持续性能力缺陷的诊断和针对性修复。我们提出了APTER（基于专家依据的评分标准的自适应后训练），这是一个将结构化领域知识整合到专业复杂推理的细粒度评估、优化和诊断中的框架。首先，基于专家依据的评分标准构建始于由领域专家构建的专家准则框架，其中每个准则代表一种稳定的专业能力。对于每个查询，APTER选择相关准则并将其实例化为与其来源准则相关联的查询级评分标准，从而将可复用的专家准则转化为无需参考答案的可执行查询级监督。其次，自适应后训练将评分标准判定同时用作优化信号和准则级诊断信号。按准则ID聚合低分判定可揭示持续性缺陷，并在强化学习期间触发有针对性的监督微调更新。在数学推理和医学问答上的实验表明，在这两个领域中都取得了一致的提升。在三代模型中，APTER相对于相应基础模型的数学和医学平均分分别提升了高达15.86分和8.04分。代码和评分标准数据集可在 https://github.com/AntDT-APTER/APTER 获取。
cs.AI / 11 / 2608.14252
Grounding Without Corrective Control: Truth-Tracking Profiles for Large Language Models
无纠正控制的接地：大型语言模型的真值追踪概况
large language model
大语言模型相关
Abstract
Recent work suggests that some large language model representations have content or reference. Grounding can secure either without supplying live routes for correction. This paper asks what follows from that gap. An output is answerable when discrepancies can affect what a target- and task-specific arrangement produces, accepts, or withdraws. The arrangement has corrective control only when live, sufficiently independent routes can detect and repair fresh discrepancies. A route profile records which routes constrain the arrangement and how they are related. Those profiles support analysis of truth-tracking: patterned support for representational success. Language models are the pressure case; text-only arrangements provide a task-relative limiting case. Text-trained models inherit patterns of testimony, coherence, and prior correction. Where target-sensitive correction survives training, these can supply derivative answerability (inherited constraint); live answerability is the relation supplied by a current route for fresh discrepancies. Fluent failures should follow when a task requires independently informative access to the facts. Self-consistency, retrieval, tools, code execution, multimodal input, and feedback should help selectively. Route-by-task interactions test the distinctions. The decomposition's empirical burden is to predict held-out route--task combinations or improve intervention choice without conceptual refitting. Surface improvement and truth-tracking improvement can come apart.
Chinese Translation
近期研究表明，一些大型语言模型的表征具有内容或指称。接地可以在不提供实时纠正路径的情况下确保其中任一者。本文追问这一差距的后果。当一个针对目标和任务的特定配置所产生、接受或撤回的内容能受差异影响时，该输出便是可应答的。只有当实时且足够独立的路径能够检测并修复新差异时，该配置才具有纠正控制。路径概况记录了哪些路径约束该配置以及它们如何关联。这些概况支持对真值追踪的分析：即对表征成功的模式化支持。语言模型是压力案例；纯文本配置提供了一个与任务相关的极限案例。经文本训练的模型继承了证言、连贯性和先前纠正的模式。在目标敏感纠正于训练后仍然存续之处，这些模式能够提供衍生的可应答性（继承的约束）；实时可应答性则是由当前路径为新差异提供的那种关系。当一项任务要求对事实进行独立且信息充分的访问时，流利的失败应当随之出现。自我一致性、检索、工具、代码执行、多模态输入和反馈应有选择地提供帮助。路径与任务的交互作用检验这些区分。该分解的经验负担在于预测留出的路径-任务组合，或在不进行概念重塑的情况下改进干预选择。表面改进与真值追踪改进可能相互背离。
cs.AI / 12 / 2608.14270
TimeSage-EV: A Live Benchmark for Agentic Time Series Analysis in Evolving Environments
TimeSage-EV：演化环境中智能体时间序列分析的动态基准
large language model
大语言模型相关
Abstract
Time series analysis in high-stakes domains relies on recurring data releases, where new observations can alter the evidence base and the validity of later conclusions. Existing time series QA benchmarks mostly rely on fixed snapshots, leaving temporal validity and cutoff-aware evidence use unevaluated. We introduce TimeSage-EV, a live benchmark for agentic time series analysis in evolving environments. It tracks 60 real institutional scenarios across 6 domains, comprising 1,485 scenario-period QA pairs from Feb 2023 to May 2026 and spanning monthly, weekly, daily, and irregular release cadences. At each period, large language model (LLM) agents receive time series data and source reports, while the withheld target release provides ground truth. TimeSage-EV evaluates state identification, data summarization, and outlook reasoning. Experiments with frontier LLM agents and TimeSage-1.0, a novel self-evolving agent with a reusable analytical skill library, reveal significant performance gaps across model tiers and recurring failures in temporal validity, exogenous context use, and adaptation. We release TimeSage-EV as a research resource with monthly updates, code, a leaderboard, and failure-mode analyses.
Chinese Translation
高风险领域中的时间序列分析依赖于周期性的数据发布，新的观测可能改变证据基础以及后续结论的有效性。现有时间序列问答基准大多依赖固定快照，使得时间有效性和考虑截止时间的证据使用未得到评估。我们提出 TimeSage-EV，一个用于演化环境中智能体时间序列分析的动态基准。它追踪6个领域的60个真实机构场景，包含从2023年2月到2026年5月的1,485个场景-时段问答对，涵盖月度、周度、每日和不规则的发布节奏。在每个时段，大语言模型（LLM）智能体接收时间序列数据和源报告，而被保留的目标发布则提供真实值。TimeSage-EV 评估状态识别、数据总结和展望推理。使用前沿LLM智能体和TimeSage-1.0（一个具有可重用分析技能库的新型自演化智能体）进行的实验揭示了不同模型层级之间的显著性能差距，以及在时间有效性、外部情境使用和适应性方面的反复失败。我们发布 TimeSage-EV 作为研究资源，包含月度更新、代码、排行榜和失败模式分析。
cs.AI / 13 / 2608.14320
AnchorBench: A Multi-Pathway Benchmark for the Anchoring Effect in LLMs
AnchorBench：一个面向大语言模型中锚定效应的多路径基准
large language model
大语言模型相关
Abstract
The anchoring effect is a cognitive bias in which an initial reference value shifts a later judgment toward itself. This effect is well established in human judgment and decision-making, and recent work suggests that large language models (LLMs) exhibit similar behavior. However, existing work on anchoring in LLMs typically evaluates only a narrow set of anchor pathways and rarely distinguishes irrelevant from plausible anchors. We introduce AnchorBench, a benchmark for the anchoring effect in LLMs that evaluates multiple anchor pathways under an explicit anchor relevance axis. Across fourteen models, including ten open-weight models and four frontier API models, and a large set of controlled prompts, we find that (1) anchoring is strongly pathway-dependent, (2) plausible anchors usually induce larger shifts than irrelevant ones when introduced through stronger pathways, (3) anchor influence generally weakens as the anchor moves farther from the evidence-supported answer, most clearly on External and RAG, and (4) high task accuracy on the anchor-free control condition (Acc$_{10}$: answers within 10 points of gold) does not guarantee robustness: even frontier API models above 95% control accuracy remain susceptible to plausible anchors.
Chinese Translation
锚定效应是一种认知偏差，即初始参考值会使后续判断向该参考值偏移。这一效应在人类判断与决策中已被充分证实，而近期研究表明，大语言模型（LLMs）也表现出类似行为。然而，现有关于大语言模型中锚定效应的研究通常仅评估狭窄的锚定路径集合，且很少区分无关锚定与看似合理的锚定。我们提出了 AnchorBench，这是一个面向大语言模型中锚定效应的基准，它在明确的锚定相关性轴上评估多种锚定路径。在十四个模型上——包括十个开放权重模型和四个前沿 API 模型——以及大量受控提示下，我们发现：（1）锚定效应强烈依赖于路径；（2）当通过更强的路径引入时，看似合理的锚定通常比无关锚定引起更大的偏移；（3）随着锚定远离证据支持的答案，锚定影响通常减弱，这在外部路径（External）和检索增强生成路径（RAG）上最为明显；（4）在无锚定对照条件下较高的任务准确率（Acc$_{10}$：答案与金标准相差 10 分以内的比例）并不能保证鲁棒性：即使控制准确率高于 95% 的前沿 API 模型，仍然容易受到看似合理的锚定的影响。
cs.AI / 14 / 2608.14330
Program-space Diffusion for Morphology-to-Transcriptomics Prediction
面向形态学到转录组学预测的程序空间扩散
diffusion
扩散模型相关
Abstract
Spatial transcriptomics (ST) enables genome-wide gene expression profiling while preserving tissue architecture, but its cost and limited scalability remain major bottlenecks. This has motivated models that predict spatial expression directly from routine histology. Despite promising results, most existing approaches operate at the gene level without leveraging established transcriptomic modeling practices and rely on heterogeneous gene selection strategies, which complicates fair comparison across methods. We propose to reformulate morphology-to-transcriptomics prediction as conditional generation in transcriptional program space, thereby exploiting coordinated transcriptional variation instead of predicting genes independently. Using consensus non-negative matrix factorization (cNMF), we extract a low-dimensional set of transcriptional programs capturing coordinated expression variation in the training data, and train a conditional diffusion model to generate program activations from histology. This formulation exploits coordinated transcriptional variation and substantially lowers the dimensionality of the conditional generative task.
Chinese Translation
空间转录组学（ST）能够在保留组织结构的同时进行全基因组基因表达谱分析，但其成本和有限的可扩展性仍是主要瓶颈。这促使了直接从常规组织学预测空间表达的模型的发展。尽管已有令人鼓舞的结果，但大多数现有方法在基因水平上运作，没有利用成熟的转录组学建模实践，并且依赖异质的基因选择策略，这使得不同方法之间的公平比较变得复杂。我们提出将形态学到转录组学的预测重新构建为转录程序空间中的条件生成，从而利用协调的转录变异，而不是独立预测基因。我们使用共识非负矩阵分解（cNMF）提取一组低维转录程序，捕获训练数据中的协调表达变异，并训练一个条件扩散模型，从组织学生成程序激活。这种表述利用了协调的转录变异，并大幅降低了条件生成任务的维度。
cs.AI / 15 / 2608.14392
Tripwire: Triggering Aligned Refusal via Statistically Certified Safety Neurons
Tripwire：通过统计认证的安全神经元触发对齐拒绝
large language model
大语言模型相关
Abstract
Neuron- and path-level interventions offer the finest-grained route to defending large language models (LLMs) against jailbreak attacks, yet existing methods fall short of this promise, i.e., they often compromise model utility significantly. Specifically, one line of work suppresses toxic neurons to erase harmful semantics, but since such semantics are distributed across the network, blocking every pathway forces a large intervention footprint. An alternative line of research focus on identify safety neurons using external classifiers. While promising, the existing approaches suffer from compromising neurons that are important for the model utility as well. Moreover, both approaches remain always on and thus perturb every benign request even when no attack is present. To address these limitations, we present \ours{}, a training-free defense that first identifies safety-specific neurons through per-neuron hypothesis tests under false-discovery-rate control together with a utility-specificity filter. Based on this identification, a trigger-style clamp holds the selected neurons at their harmful-conditional mean activations, injecting an internal harmful-input signal that triggers the refusal behavior learned during alignment. The clamp is then realized by two provably equivalent deployment modes, namely a detector-gated inference-time intervention and an offline bias-patch weight edit. Extensive experiments across four safety-aligned LLMs and four representative attacks demonstrate that \ours{} reduces the average attack success rate to at most 2.0\% while incurring a utility drop of only 0.5\% to 5.3\% on MT-Bench, the smallest among all defenses. Code is available at https://anonymous.4open.science/r/Tripwire-65C4.
Chinese Translation
神经元和路径层面的干预为防御大型语言模型（LLMs）免受越狱攻击提供了最细粒度的途径，然而现有方法未能兑现这一承诺，即它们常常严重损害模型的实用性。具体地，一条工作线通过抑制有毒神经元来消除有害语义，但由于此类语义分布在整个网络中，阻断每条通路会导致巨大的干预足迹。另一条研究路线侧重于使用外部分类器来识别安全神经元。尽管前景可观，现有方法也会损害对模型实用性同样重要的神经元。此外，这两种方法都保持常开状态，因此在没有攻击时也会干扰每一个良性请求。为解决这些局限性，我们提出了Tripwire，一种无需训练的防御方法，它首先通过错误发现率控制下的逐神经元假设检验以及实用性特异性过滤器来识别安全特异性神经元。基于这一识别，一种触发器式钳制将选定的神经元保持在其有害条件下的平均激活水平，从而注入一个内部有害输入信号，触发对齐过程中学习到的拒绝行为。该钳制随后通过两种可证明等价的部署模式实现，即检测器门控的推理时干预和离线偏置补丁权重编辑。在四个安全对齐的LLMs和四种代表性攻击上进行的大量实验表明，Tripwire将平均攻击成功率降至最多2.0%，同时在MT-Bench上仅造成0.5%至5.3%的实用性下降，这是所有防御方法中最低的。代码可在 https://anonymous.4open.science/r/Tripwire-65C4 获取。
cs.AI / 16 / 2608.14397
LLMs Don't Pay for the Jump
大语言模型不为跳跃付出代价
large language model
大语言模型相关
Abstract
Zahavy [2026] argues that Large Language Models, despite their capabilities in induction and deduction, cannot perform the abductive "Jump" that produced Einstein's equivalence principle, and attributes this limitation to the absence of embodied simulation. Zheng-Xin [2026] and Farmer [2026] question whether embodiment is necessary for abduction, pointing to alternative routes to General Relativity and forms of abduction that require no sensorimotor grounding. Max Planck resolved the blackbody radiation problem in 1900. Planck's move to E = hν required no embodied simulation. It was motivated by a mathematical consequence of classical theory, an infinite predicted energy for a finite measured quantity, that could not be physically accepted. We show that neither induction nor deduction could have produced the postulate and argue that its adoption required a coupling between epistemic error and physical cost. We formalize this distinction through thermodynamic coupling and show that fixed-weight transformer inference lacks such coupling, regardless of model scale. This is consistent with empirical results showing that output entropy remains nearly unchanged across tasks with sharply increasing causal difficulty, even as accuracy falls from 100% to 17%. We therefore argue that the missing ingredient in machine abduction may lie deeper than embodiment: a system must have some physical mechanism through which epistemic error becomes costly enough to force revision.
Chinese Translation
Zahavy [2026] 认为，尽管大语言模型在归纳和演绎方面具备能力，但它们无法执行产生爱因斯坦等效原理的溯因“跳跃”，并将这一限制归因于缺乏具身模拟。Zheng-Xin [2026] 和 Farmer [2026] 质疑具身对于溯因是否必要，指出了通往广义相对论的其他路径以及不需要感觉运动基础的溯因形式。马克斯·普朗克于1900年解决了黑体辐射问题。普朗克转向 E = hν 这一步不需要具身模拟。它的动机来自经典理论的一个数学推论：对于有限测量量，预测能量为无穷大，这在物理上无法接受。我们表明，归纳和演绎都无法产生该假设，并论证采用该假设需要认知错误与物理成本之间的耦合。我们通过热力学耦合将这种区别形式化，并表明固定权重的Transformer推理缺乏这种耦合，无论模型规模如何。这与实证结果一致：在因果难度急剧增加的任务中，输出熵几乎保持不变，即使准确率从100%下降到17%。因此，我们认为机器溯因中缺失的成分可能比具身更深层：系统必须具有某种物理机制，通过该机制认知错误变得代价足够高，从而迫使修正。
cs.AI / 17 / 2608.14452
SheetCompass: Hierarchical Relation Graphs for Agentic Spreadsheet Reasoning
SheetCompass：用于智能体电子表格推理的层级关系图
large language model
大语言模型相关
Abstract
Spreadsheets are widely used to organize, analyze, and manipulate semi-structured data, yet automated spreadsheet reasoning remains challenging for large language models (LLMs). Real-world workbooks often contain implicit cross-table associations, fine-grained column dependencies, and complex spatial layouts. Existing methods typically flatten these multidimensional structures into sequential strings, losing important intra-sheet boundaries and inter-sheet semantics. Consequently, LLMs cannot exploit the global spatial context that human experts naturally use when inspecting spreadsheets. We propose SheetCompass, a graph-guided and memory-driven agentic framework for spreadsheet reasoning and automation. SheetCompass explicitly models structural relationships within and across worksheets while maintaining task-relevant information in memory, enabling agents to reason more effectively over complex workbooks.
Chinese Translation
电子表格被广泛用于组织、分析和操作半结构化数据，然而对于大型语言模型（LLMs）而言，自动化电子表格推理仍然具有挑战性。真实世界的工作簿通常包含隐式的跨表关联、细粒度的列依赖关系以及复杂的空间布局。现有方法通常将这些多维结构展平为顺序字符串，从而丢失了重要的工作表内边界和工作表间语义。因此，大型语言模型无法利用人类专家在检查电子表格时自然使用的全局空间上下文。我们提出了SheetCompass，一种图引导且记忆驱动的智能体框架，用于电子表格推理与自动化。SheetCompass显式地对工作表内部及跨工作表的结构关系进行建模，同时在记忆中维护任务相关信息，使智能体能够更有效地对复杂工作簿进行推理。
cs.AI / 18 / 2608.14528
Handover of In-Context Learning State Across Session Boundaries
跨会话边界的上下文学习状态交接
large language model
大语言模型相关
Abstract
This study investigates the methodological and theoretical properties of session handover in applications that use large language models. A task may continue in a new session when the context reaches the model's input limit, when the application restarts, or when another agent is asked to finish the task. The application must then decide which information from the earlier session to pass on. We formulate handover as the transfer of a task-relative in-context learning (ICL) state and distinguish exact recovery of earlier material from preservation of the target distribution. Under an exogeneity condition, predictive equivalence characterizes the coarsest deterministic sufficient handover and gives a fixed-length bit requirement. The analysis isolates the effects of the memory constraint, the writer, and the continuation procedure, and quantifies the cost of writing before the realized downstream query is known. We propose a three-part record that stores decisions and constraints exactly, uses task-justified statistics for repeated evidence, and retains original observations whose effect is not preserved by those statistics. Gaussian linear regression gives an exact finite-dimensional handover and finite-bit perturbation bounds, while nonparametric regression gives upper and lower bounds that relate memory to squared prediction error. These results provide a theory and method for deciding what a handover must retain and how its memory requirement depends on the continuation task.
Chinese Translation
本研究探讨了使用大语言模型的应用中会话交接的方法论与理论性质。当上下文达到模型的输入限制、应用重新启动，或要求另一个智能体完成任务时，任务可能会在新会话中继续。此时，应用必须决定从较早会话中传递哪些信息。我们将交接形式化为任务相关上下文学习（ICL）状态的转移，并区分对早期材料的精确恢复与对目标分布的保护。在一种外生性条件下，预测等价性刻画了最粗糙的确定性充分交接，并给出了固定长度的比特需求。该分析分离了记忆约束、写入者和延续过程的影响，并量化了在已知实现的下游查询之前进行写入的成本。我们提出了一种三部分记录，它精确存储决策和约束，对重复证据使用任务合理的统计量，并保留那些未被这些统计量保留其效应的原始观测。高斯线性回归给出了精确的有限维交接和有限比特扰动界，而非参数回归给出了将记忆与平方预测误差联系起来的上下界。这些结果为决定交接必须保留什么、以及其记忆需求如何依赖于延续任务提供了一种理论和方法。
cs.AR / 19 / 2608.13868
Exploring High-Bandwidth Flash for Modern LLM Inference: Opportunities and Challenges
探索面向现代 LLM 推理的高带宽闪存：机遇与挑战
large language model
大语言模型相关
Abstract
This work investigates the potential benefits and technical challenges of using high-bandwidth flash (HBF) for large language model (LLM) inference. HBF has gained increasing attention as a promising solution to mitigate memory-capacity bottlenecks in modern LLM-serving systems, but its benefits and challenges remain largely uninvestigated. To address this gap, we thoroughly analyze HBF-based LLM-serving systems under diverse system configurations and operating scenarios in which HBF serves as a main GPU-memory component to handle both reads and writes. Our analysis shows that, despite its limited write performance, HBF can significantly improve the batch size, throughput, and flexibility of LLM-serving systems while reducing the minimum GPU requirements, but realizing these benefits critically depends on sustaining HBM-comparable read bandwidth and requires significant endurance improvements.
Chinese Translation
本文研究了将高带宽闪存（HBF）用于大型语言模型（LLM）推理的潜在优势和技术挑战。HBF作为一种有前景的解决方案，在缓解现代LLM服务系统中的内存容量瓶颈方面日益受到关注，但其优势与挑战在很大程度上仍未得到充分研究。为了填补这一空白，我们在多种系统配置和运行场景下全面分析了基于HBF的LLM服务系统，其中HBF作为主要的GPU内存组件来处理读取和写入。我们的分析表明，尽管写入性能有限，HBF能够显著提升LLM服务系统的批处理大小、吞吐量和灵活性，同时降低最低GPU要求，但实现这些优势关键依赖于保持与HBM相当的读取带宽，并且需要显著的耐久性改进。
cs.CL / 20 / 2608.13947
Scaling Creative Writing Beyond Story-Centric Data with Attribute-Guided Genre Expansion
超越以故事为中心的数据：基于属性引导的体裁扩展实现创意写作规模化
large language model
大语言模型相关
Abstract
High-quality creative writing data for large language models (LLMs) remains dominated by story-centric data, limiting models' ability to follow the structural and functional conventions of diverse creative formats. We propose an attribute-guided genre expansion framework for scaling creative writing data beyond story generation. By separating thematic breadth from genre-form control, our framework leverages human-authored story prompts as diverse creative seeds, while utilizing manually curated genre attributes to enforce distinct structural, stylistic, and formatting conventions. We combine these to prompt strong LLMs for genre-faithful query-response pairs, which are then quality-filtered. Applying this framework, we construct the Multi-Genre Collection, a 50K-example corpus spanning 13 creative genres, including story, rap, lyrics, scripts, game design, character design, and other creative formats. Experiments across out-of-distribution writing benchmarks and held-out genre diagnostics demonstrate that models fine-tuned on our data consistently surpass not only base models and writing-specialized baselines, but also models trained on existing writing corpora. Genre-count ablations further indicate that controlled genre expansion, rather than story-centric scaling alone, is a key driver of robust creative writing capability.
Chinese Translation
用于大型语言模型（LLM）的高质量创意写作数据仍以故事为中心的数据为主导，这限制了模型遵循多样化创意格式的结构与功能惯例的能力。我们提出了一种属性引导的体裁扩展框架，用于将创意写作数据的规模化扩展至故事生成之外。通过将主题广度与体裁形式控制相分离，我们的框架利用人类撰写的故事提示作为多样化的创意种子，同时利用人工整理的体裁属性来强制执行不同的结构、风格和格式惯例。我们将这两者结合，以提示强大的LLM生成符合体裁要求的查询-响应对，随后对这些数据进行质量过滤。应用该框架，我们构建了多体裁语料库（Multi-Genre Collection），这是一个包含5万个示例的语料库，涵盖13种创意体裁，包括故事、说唱、歌词、剧本、游戏设计、角色设计以及其他创意格式。在分布外写作基准和保留体裁诊断上的实验表明，基于我们的数据微调的模型不仅持续超越基础模型和写作专项基线，也超越了在现有写作语料库上训练的模型。体裁数量消融实验进一步表明，受控的体裁扩展——而非仅靠以故事为中心的规模化——是稳健创意写作能力的关键驱动因素。
cs.CL / 21 / 2608.14055
HERMES: a multi-agent framework for structured knowledge extraction from ultra-long documents in geoscience
HERMES：一种面向地球科学超长文档结构化知识抽取的多智能体框架
large language model
大语言模型相关
Abstract
Authoritative scientific knowledge in geoscience remains largely trapped in legacy monographs and historical literature, where unstructured text and complex layouts hinder computational access. We introduce HERMES, a scalable multi-agent framework that extracts structured data from ultra-long scientific documents. Using a coordinating large language model, HERMES integrates domain constraints, validation rules and evidence tracing within a unified document-level extraction process that incorporates parsed text, tables, figures and captions. Applied to the 55-volume Treatise on Invertebrate Paleontology, the system produced a structured database of 32,277 fossil taxonomic entities and 451,878 attributes, released online at https://treatise.geolex.org. Extraction performance remained stable across fossil groups (average F1 scores of approximately 0.90 for entities and 0.91 for attributes), improving per-volume efficiency approximately sixfold relative to the tested fully manual baseline. Evaluation in palaeomagnetism and geochemistry, conducted without additional model training, demonstrated transfer across distinct geoscience domains. This work provides a practical pathway to transform historical scientific literature into FAIR-oriented structured data, offering a sustainable infrastructure for data-intensive disciplines and large-scale knowledge integration.
Chinese Translation
地球科学中的权威科学知识在很大程度上仍被困在传统专著和历史文献中，其中非结构化的文本和复杂的版式阻碍了计算访问。我们引入了 HERMES，一个可扩展的多智能体框架，用于从超长科学文档中提取结构化数据。通过使用一个协调性大语言模型，HERMES 在统一的文档级提取过程中整合了领域约束、验证规则和证据追踪，并结合了解析后的文本、表格、图形和标题。将该系统应用于 55 卷的《无脊椎古生物学论丛》，系统生成了一个包含 32,277 个化石分类实体和 451,878 个属性的结构化数据库，并已在 https://treatise.geolex.org 在线发布。提取性能在不同化石类群中保持稳定（实体的平均 F1 分数约为 0.90，属性的平均 F1 分数约为 0.91），与所测试的完全人工基线相比，每卷效率提高了约六倍。在古地磁学和地球化学中的评估，在没有额外模型训练的情况下进行，展示了其在不同地球科学领域间的可迁移性。这项工作提供了一条将历史科学文献转化为面向 FAIR 的结构化数据的实用途径，为数据密集型学科和大规模知识集成提供了可持续的基础设施。
cs.CL / 22 / 2608.14377
A Survey of Large Models in Sports
大模型在体育领域的综述
large language model
大语言模型相关
Abstract
Sports have witnessed growing global enthusiasm in recent years, serving as a vital force for physical health, cultural exchange, social connection, and economic growth. The rapid advancement of large models, particularly (multimodal) large language models (M)LLMs, has demonstrated transformative potential to reshape sports understanding, analysis, and interaction across diverse domains. This paper presents a comprehensive survey of large models in sports, including (i) an overview of tasks and applications across different participant groups; (ii) a detailed analysis of sports-related datasets and benchmarks; and (iii) a critical discussion of current challenges and future directions. Our goal is to establish a foundation for advancing research and practical development of large-model-driven sports intelligence. An open-source GitHub repository is maintained at: https://github.com/Road2Redemption/Awesome_Large_Models_In_Sports1.
Chinese Translation
近年来，体育在全球范围内展现出日益高涨的热情，已成为促进身体健康、文化交流、社会联系和经济增长的重要力量。大模型的快速发展，尤其是（多模态）大语言模型（(M)LLMs），已展现出重塑体育理解、分析和跨领域互动的变革性潜力。本文对体育领域的大模型进行了全面综述，包括：（i）不同参与群体中任务与应用的概述；（ii）对体育相关数据集和基准的详细分析；以及（iii）对当前挑战和未来方向的关键性讨论。我们的目标是为推动大模型驱动的体育智能的研究和实践发展奠定基础。我们维护了一个开源GitHub仓库，地址为：https://github.com/Road2Redemption/Awesome_Large_Models_In_Sports1。
cs.CR / 23 / 2608.13928
CoSA: Context-Aware Severity Assessment via Context Analysis with Large Language Models
CoSA：通过大型语言模型的上下文分析实现上下文感知的严重性评估
large language model
大语言模型相关
Abstract
Accurate vulnerability severity assessment is essential for prioritizing remediation, yet manually assessing Common Vulnerability Scoring System (CVSS) base metrics remains labor-intensive. Existing automated approaches often fail to capture the repository-level evidence required for assessing many CVSS base metrics. Such repository-aware assessment is challenging because relevant evidence is scattered across the entire repository under heavy noise. To address these challenges, we present CoSA, a Context-aware vulnerability Severity Assessment approach that infers CVSS base metrics from repository artifacts. CoSA constructs a code property graph (CPG) and applies a two-stage repository-pruning strategy: lightweight static pruning to preserve structurally proximal context, followed by an agentic large language model (LLM)-guided pruning step to retain CVSS-relevant context while collecting supporting evidence. The LLM then consolidates the retrieved repository context into compact, CVSS metric-wise textual summaries, which are fed into a lightweight transformer predictor. We also construct a higher-quality repository-level dataset comprising 6,816 CVSS labeled instances spanning 90 Common Weakness Enumeration (CWE) types. Experiments on real-world vulnerabilities show that CoSA consistently outperforms function-level and pure-LLM baselines. It improves prediction accuracy by 14.4% and Macro-F1 by 15.3% over the best-performing baseline, suggesting that explicit, metric-oriented repository context retrieval is crucial for practical and reliable automated severity assessment.
Chinese Translation
准确的漏洞严重性评估对于确定修复优先级至关重要，然而手动评估通用漏洞评分系统（CVSS）基础指标仍然需要大量人力。现有的自动化方法通常无法捕获评估许多CVSS基础指标所需的仓库级证据。这种具备仓库感知的评估具有挑战性，因为相关证据散布在整个仓库中，且伴有大量噪声。为应对这些挑战，我们提出了CoSA，一种上下文感知的漏洞严重性评估方法，它从仓库工件中推断CVSS基础指标。CoSA构建代码属性图（CPG），并采用两阶段仓库剪枝策略：首先是轻量级静态剪枝以保留结构上邻近的上下文，其次是智能体式大型语言模型（LLM）引导的剪枝步骤，以在收集支持证据的同时保留与CVSS相关的上下文。随后，LLM将检索到的仓库上下文整合为紧凑的、按CVSS指标划分的文本摘要，并将其输入轻量级Transformer预测器。我们还构建了一个更高质量的仓库级数据集，其中包含6,816个CVSS标记实例，涵盖90种常见弱点枚举（CWE）类型。在真实漏洞上的实验表明，CoSA始终优于函数级和纯LLM基线。与表现最佳的基线相比，它将预测准确率提高了14.4%，宏F1分数提高了15.3%，这表明面向指标的显式仓库上下文检索对于实用且可靠的自动化严重性评估至关重要。
cs.CR / 24 / 2608.14216
MazeRunner: Nonlinear Task and Clue Orchestration for LLM-driven Black-Box Automated Penetration Testing
MazeRunner：面向LLM驱动的黑盒自动化渗透测试的非线性任务与线索编排
large language model
大语言模型相关
Abstract
Penetration testing is essential yet resource-intensive. Although large language models (LLMs) show promise for automating security auditing, existing agents mainly execute end-to-end workflows in simplified linear scenarios. Real-world black-box testing is fundamentally nonlinear: the attack graph is initially unknown and must be incrementally inferred from environmental feedback. Observations may reveal multiple attack branches, failures are often ambiguous, and critical clues may span long action horizons. Existing agents therefore tend to become trapped in depth-first exploration, misdiagnose failures, and forget prior evidence. We present MazeRunner, an autonomous penetration testing system built on a three-agent task-and-clue orchestration framework. It separates global orchestration, context-intensive execution, and failure-oriented review while persistently maintaining task states and environmental evidence. This design supports action revision, prerequisite recovery, branch switching, and long-range clue correlation. We evaluate MazeRunner on 10 recently released HTB targets, limiting each system-target run to 20 million LLM tokens and preventing target-specific solution leakage. With Claude Sonnet 4.5, MazeRunner completes 47.7% of annotated subtasks, compared with 36.2% for PentestGPT-V2 and 34.2% for Claude Code. It achieves user-level or higher access on six targets, including root access on two; each same-model baseline reaches user-level access on only two targets and never obtains root access. Execution-trace analysis further shows that MazeRunner explores more attack branches and acquires shells more efficiently.
Chinese Translation
渗透测试至关重要，但资源消耗巨大。尽管大型语言模型（LLM）在自动化安全审计方面展现出潜力，但现有智能体主要是在简化的线性场景中执行端到端工作流。真实世界的黑盒测试本质上是非线性的：攻击图最初是未知的，必须根据环境反馈逐步推断。观察结果可能揭示多个攻击分支，失败往往具有歧义，而关键线索可能跨越较长的行动跨度。因此，现有智能体容易陷入深度优先探索、误判失败，并遗忘先前的证据。我们提出了MazeRunner，一个基于三智能体任务与线索编排框架的自主渗透测试系统。它将全局编排、上下文密集型执行和面向失败的审查相分离，同时持续维护任务状态和环境证据。这种设计支持动作修正、先决条件恢复、分支切换和长程线索关联。我们在10个近期发布的HTB目标上评估MazeRunner，将每个系统-目标运行限制为2000万个LLM令牌，并防止针对特定目标的解决方案泄漏。使用Claude Sonnet 4.5，MazeRunner完成了47.7%的标注子任务，而PentestGPT-V2为36.2%，Claude Code为34.2%。它在六个目标上实现了用户级或更高权限的访问，其中两个目标获得root权限；每个同模型基线仅在两个目标上达到用户级访问，且从未获得root权限。执行痕迹分析进一步表明，MazeRunner探索了更多攻击分支，并更高效地获取了shell。
cs.CR / 25 / 2608.14370
A Hybrid LLM-Based Framework for Automated Security Annotation Generation in Business Process Models
一种基于混合大语言模型的业务流程模型安全标注自动生成框架
large language model
大语言模型相关
Abstract
The modelling and analysis of secure business processes require the incorporation of security annotations into process models. Although BPMN extensions, including SecBPMN2, exist for this purpose, the derivation of accurate and complete security annotations from natural-language specifications remains a manual, expert-intensive, and error-prone task. This paper presents a hybrid framework that takes a BPMN process model and a security requirements document as input and automatically generates security annotations adhering to the SecBPMN2 specification. The approach combines Large Language Model (LLM)--based semantic extraction with schema-constrained mapping, rule-based normalization, and deterministic validation. The framework is evaluated comprehensively on a curated dataset of 27 process models from various domains. The results indicate that it consistently produces structurally valid SecBPMN2 annotations with high schema completeness. Compared to human security analysts, the system achieves substantially higher precision (0.58 vs. 0.29) while maintaining comparable recall (0.52 vs. 0.50) and reduces erroneous or misplaced annotations by nearly 50%. In addition, annotation generation is significantly faster than manual annotation. These findings demonstrate that hybrid LLM- and rule-based automation can reduce modeling effort while improving consistency and reliability, thereby providing a scalable foundation for security-by-design BPM.
Chinese Translation
安全业务流程的建模与分析需要在流程模型中纳入安全标注。尽管存在包括SecBPMN2在内的BPMN扩展用于此目的，但从自然语言规范中推导出准确且完整的安全标注仍然是一项人工密集、依赖专家且容易出错的任务。本文提出了一种混合框架，该框架以BPMN流程模型和安全需求文档为输入，自动生成符合SecBPMN2规范的安全标注。该方法将基于大语言模型（LLM）的语义提取与模式约束映射、基于规则的规范化以及确定性验证相结合。该框架在一个由来自不同领域的27个流程模型组成的高质量数据集上进行了全面评估。结果表明，它能够持续生成结构有效且模式完整性高的SecBPMN2标注。与人类安全分析师相比，该系统在保持相当召回率（0.52对0.50）的同时，实现了显著更高的精确率（0.58对0.29），并将错误或错位的标注减少了近50%。此外，标注生成速度显著快于人工标注。这些发现表明，混合LLM与基于规则的自动化可以减少建模工作量，同时提高一致性和可靠性，从而为安全设计BPM提供可扩展的基础。
cs.CR / 26 / 2608.14418
STINER: Automated Extraction of Strategic Cyber Threat Intelligence from X
STINER：从X平台自动提取战略性网络威胁情报
large language model
大语言模型相关
Abstract
Strategic Cyber Threat Intelligence (CTI) focuses on high-level insights, such as identifying targeted industries, attributing attacks to specific ransomware groups, and assessing the scale of data loss. Today, X (formerly Twitter) has become the fastest source for this intelligence, often hosting real-time breach announcements days before formal vendor reports. Converting this raw chatter into actionable intelligence requires navigating a complex linguistic landscape. Conventional Named Entity Recognition (NER) models struggle to parse the informal and highly irregular dialect of social media, creating a blind spot for automated defense systems. To address this challenge, we introduce STINER, a taxonomy and expert-annotated corpus for extracting strategic intelligence from social media streams. We construct a high-quality, expert-annotated dataset of 2,100 real-world alerts and propose a granular taxonomy of eight entity types centered on strategic pivots such as Threat Actor, Sector, and Location. We benchmark nine models across 12 evaluated configurations, spanning general-purpose and domain-adapted encoders, open-schema extraction, and generative LLMs in both zero-shot and fine-tuned settings. Domain-adapted encoders such as DarkBERT reach a strict F1-score of 89.33%, outperforming both general-purpose baselines and fine-tuned Large Language Models, which additionally incur substantially higher inference latency. Leveraging STINER-DarkBERT, we conduct a European threat landscape analysis for H1 2025. Our results align with official reporting on major targets while highlighting the distinct visibility profile of attacks in Spain, and illustrate how social-media-driven extraction can surface early signals of the SafePay ransomware campaign prior to its retrospective characterization in vendor threat landscape reports.
Chinese Translation
战略性网络威胁情报（CTI）侧重于高层次洞察，例如识别受攻击的行业、将攻击归因于特定的勒索软件组织，以及评估数据丢失的规模。如今，X（原Twitter）已成为获取此类情报最快的来源，通常会在正式供应商报告发布前数天就发布实时的数据泄露公告。将这种原始讨论转化为可操作的情报，需要应对复杂的语言环境。传统命名实体识别（NER）模型难以解析社交媒体中非正式且高度不规则的表达方式，从而给自动化防御系统造成了盲区。为应对这一挑战，我们提出了STINER，一个用于从社交媒体流中提取战略情报的分类体系及专家标注语料库。我们构建了一个包含2,100条真实世界告警的高质量专家标注数据集，并提出了一个以威胁行为者、行业和地理位置等战略核心为中心的八种实体类型的细粒度分类体系。我们在12种评估配置下对9个模型进行了基准测试，涵盖通用编码器、领域自适应编码器、开放式模式提取，以及零样本和微调设置下的生成式大语言模型。领域自适应编码器（如DarkBERT）达到了89.33%的严格F1分数，优于通用基线模型和微调后的大语言模型，而后者还会产生显著更高的推理延迟。利用STINER-DarkBERT，我们对2025年上半年（H1 2025）的欧洲威胁格局进行了分析。我们的结果与官方关于主要攻击目标的报告一致，同时突出了西班牙攻击事件的独特可见性特征，并展示了社交媒体驱动的提取如何能够在SafePay勒索软件活动被供应商威胁格局报告追溯定性之前，提前发现其早期信号。
cs.CR / 27 / 2608.14533
Finding Vulnerabilities via LLM-Augmented Semantics-Aware Type-Checking
通过LLM增强的语义感知类型检查发现漏洞
large language model
大语言模型相关
Abstract
Vulnerability detection via static analysis traditionally relies on security experts encoding insecure coding patterns into algorithmic rules. However, this approach often focuses on syntactic patterns and overlooks deeper semantic information in the code, such as the meanings of variable and function names. As software systems grow more complex, modeling vulnerabilities using only syntactic rules becomes increasingly challenging. In this paper, we propose a semantics-aware approach to detecting software vulnerabilities. We present SETYPE, a semantics-aware type system that can be derived directly from source code based solely on the meanings of symbols and expressions in natural language. In the SETYPE type system, both type inference and checking are performed by Large Language Models (LLMs), and a failed type check indicates a potential vulnerability. We prototype PYSETYPE to demonstrate the feasibility of SETYPE for detecting vulnerabilities in Python web applications. Our evaluation on real-world applications achieves 87% detection precision and 88% detection accuracy. Using PYSETYPE, we identified 15 potential zero-day vulnerabilities, nine of which were confirmed by developers.
Chinese Translation
通过静态分析进行漏洞检测传统上依赖于安全专家将不安全的编码模式编码为算法规则。然而，这种方法通常关注语法模式，而忽略了代码中更深层的语义信息，例如变量和函数名称的含义。随着软件系统变得越来越复杂，仅使用语法规则对漏洞进行建模变得越来越具有挑战性。在本文中，我们提出了一种语义感知的方法来检测软件漏洞。我们提出了SETYPE，一种语义感知的类型系统，它可以仅基于符号和表达式在自然语言中的含义直接从源代码中派生。在SETYPE类型系统中，类型推断和类型检查均由大型语言模型（LLM）执行，类型检查失败即表示存在潜在漏洞。我们实现了原型PYSETYPE，以证明SETYPE在检测Python Web应用程序漏洞方面的可行性。我们在真实世界应用程序上的评估实现了87%的检测精度和88%的检测准确率。使用PYSETYPE，我们识别了15个潜在的零日漏洞，其中9个已得到开发人员的确认。
cs.AI / 28 / 2608.14172
Concept Guidance: Precise, Training-Free Latent Control for Text-to-Image Generation
概念引导：用于文本到图像生成的精确、无需训练的潜在控制
diffusion
扩散模型相关
Abstract
Text-to-image diffusion models have two major drawbacks that severely limit their practical utility: (1) standard models lack an intrinsic mechanism for continuous, concept-specific guidance (e.g., for precisely controlling how aesthetically pleasing an image looks), and (2) they lack reliability for tasks requiring high local coherence (e.g., generating text or human hands). To tackle these issues, we introduce a novel notion of concept-wise mutual information and find large, concept-dependent differences between individual layers, demonstrating that the generation of specific structures is localized in distinct parts of the network. We exploit this insight by reinforcing the impact of concept-relevant layers in Concept Guidance (CoG), a precise, target-specific guidance method that works for models out-of-the-box without additional training, external models, gradients, or prompt engineering. CoG first quantifies each layer's concept-specific impact and then guides denoising using a weighted combination of predictions generated with concept-relevant layers skipped. We demonstrate performance increases across various targets and popular models like PixArt-alpha, SD3, SD3.5, and FLUX.1-dev. Code is available at https://github.com/CompVis/concept_guidance
Chinese Translation
文本到图像扩散模型有两个主要缺点，严重限制了它们的实际用途：（1）标准模型缺乏内在的机制来提供连续的、针对特定概念的引导（例如，精确控制图像看起来有多美观），并且（2）它们对于需要高度局部一致性的任务（例如生成文本或人类手部）缺乏可靠性。为了解决这些问题，我们引入了一种新颖的概念级互信息概念，并发现不同层之间存在大型且依赖于概念的差异，证明特定结构的生成位于网络的不同部分。我们通过增强与概念相关的层在概念引导（CoG）中的影响来利用这一洞见，CoG是一种精确的、针对目标的引导方法，适用于开箱即用的模型，无需额外训练、外部模型、梯度或提示工程。CoG首先量化每层的概念特定影响，然后使用跳过与概念相关层的预测的加权组合来引导去噪。我们展示了在多种目标以及流行模型（如PixArt-alpha、SD3、SD3.5和FLUX.1-dev）上的性能提升。代码可在 https://github.com/CompVis/concept_guidance 获取。
cs.AI / 29 / 2608.14530
Marionette: Predicting World States, Rendering Geometry, Painting Appearance
Marionette：预测世界状态、渲染几何、绘制外观
diffusion
扩散模型相关
Abstract
Interactive game world models typically autoregress visual observations directly in pixel or latent space, forcing structured properties such as pose, geometry, and occlusion to be implicitly maintained by the same generative sequence. Over long horizons, errors in these latent world properties accumulate, making consistency and controllability fragile. We explicitly model the evolving world state, delegate exact geometric computation to a fixed, zero-parameter renderer, and leave the neural model to synthesize appearance. We instantiate this idea as Marionette, a world model for interactive games with articulated characters. First, a two-stage autoregressive dynamics model predicts an explicit and interpretable 276-dimensional 3D world state comprising multi-entity articulated skeletons, metric root trajectories, and rotations. Second, a zero-parameter graphics bridge converts the predicted state into pose-control videos, computing world-space geometry and occlusion in closed form. Third, a control-conditioned video-diffusion observation model synthesizes photorealistic RGB observations from the resulting structured controls. Our experiments establish two properties of Marionette. First, the predicted world state is directly controllable. Forcing a mismatched action stream changes root-aligned joint error by 31% across 48 held-out segments. Second, long-horizon behaviour is determined in the state, and can be repaired there. Left free, the two generated characters drift to 21.2 m apart (recorded sessions stay near 5 m) and a third of frames show ground penetration. Two rules imposed on the explicit state, a terrain collider and a separation cap, cut penetration by 66% and keep the pair engaged, with no change to the observation model. Routing appearance through the predicted state costs no fidelity we can detect, at an FVD of 831 against 799 for recorded pose.
Chinese Translation
交互式游戏世界模型通常直接在像素或潜在空间中自回归地生成视觉观察，迫使姿势、几何、遮挡等结构化属性由同一生成序列隐式维持。在长时间范围内，这些潜在世界属性中的误差不断累积，使得一致性和可控性变得脆弱。我们显式地对演化中的世界状态进行建模，将精确的几何计算委托给一个固定的零参数渲染器，而让神经模型来合成外观。我们将这一想法实例化为Marionette，一个面向具有关节角色的交互式游戏的世界模型。首先，一个两阶段自回归动力学模型预测一个显式且可解释的276维3D世界状态，该状态包含多实体关节骨骼、度量根轨迹和旋转。其次，一个零参数图形桥接器将预测状态转换为姿态控制视频，以闭式形式计算世界空间几何和遮挡。第三，一个控制条件视频扩散观测模型从生成的结构化控制中合成照片逼真的RGB观测。我们的实验证实了Marionette的两个特性。首先，预测的世界状态是直接可控的。在48个保留片段中，强制不匹配的动作流使根对齐的关节误差改变了31%。其次，长时行为由状态决定，并且可以在状态中进行修复。如果任其自由发展，两个生成的字符会漂移到相距21.2米（记录会话保持在5米附近），并且三分之一的帧显示地面穿透。在显式状态上施加的两条规则，即地形碰撞器和分离上限，将穿透率降低了66%，并保持这对角色互动，而无需改变观测模型。通过预测状态路由外观并不会造成我们可检测到的保真度损失，其FVD为831，而记录姿态的FVD为799。
cs.LG / 30 / 2608.13695
Language-Specific Gaps in AI Safety Training Datasets
AI安全训练数据集中的语言特定缺口
large language model
大语言模型相关
Abstract
Large language model providers routinely cite multilingual safety benchmarks spanning a dozen or more languages as evidence that their models are safe for non-English-speaking users. We show that these collection-level coverage claims frequently do not survive inspection at the level of an individual language. Auditing 21 resources across 25 language slices, of which 20 count as datasets under our counting rules, spanning three languages chosen to represent low- (Hausa), mid- (Swahili), and high-resource (French) tiers, we find that gaps in provenance, annotation reliability, access, harm-taxonomy coverage, and data reuse recur in patterns that partially, but not fully, track resource level. Using a controlled within-pipeline comparison, we show a Hausa-language slice falling below its own paper's translation-quality acceptance threshold while the same pipeline's Swahili output clears the same bar comfortably; this is evidence that these gaps are measurable and addressable, not inherent. We further show that self-harm and sexual-content categories have no native-language coverage in either African-language tier we studied, a total rather than gradated gap that a purely resource-level account does not predict. We connect these findings to a documented, persistent asymmetry in multilingual jailbreak robustness (single-turn attacks largely mitigated, multi-turn attacks still effective), arguing that this asymmetry is structurally consistent with where our audit finds training and evaluation data thinnest. We contribute a reusable slice-level audit methodology, a cross-tier empirical comparison, and concrete recommendations for dataset creators, model providers, and venues aiming to make ``multilingual coverage'' claims verifiable rather than merely stated. Dataset: https://huggingface.co/datasets/ChialukaOnuoha/safety-slice-audit
Chinese Translation
大型语言模型提供商通常引用涵盖十几种或更多语言的多语言安全基准，作为其模型对非英语用户安全的证据。我们表明，这些集合层面的覆盖声明在单个语言层面进行审查时往往无法成立。我们审计了跨越25个语言切片的21项资源，其中20项根据我们的计数规则算作数据集，覆盖了代表低资源（豪萨语）、中资源（斯瓦希里语）和高资源（法语）三个层级的语言，发现来源、标注可靠性、访问权限、危害分类法覆盖和数据重用方面的缺口以部分但并非完全追踪资源层级的模式反复出现。通过受控的流水线内比较，我们显示豪萨语切片低于其自身论文的翻译质量接受阈值，而同一流水线的斯瓦希里语输出轻松超过同一标准；这证明这些缺口是可衡量和可解决的，而非固有的。我们进一步表明，自残和性内容类别在我们研究的两个非洲语言层级中均没有母语覆盖，这是一个总体性的而非渐进的缺口，纯粹的资源层级解释无法预测这一点。我们将这些发现与多语言越狱鲁棒性中一个已记录且持续存在的不对称性联系起来（单轮攻击基本被缓解，多轮攻击仍然有效），论证这种不对称性在结构上与我们的审计发现训练和评估数据最薄弱之处一致。我们贡献了一种可复用的切片级审计方法、一个跨层级的实证比较，以及针对数据集创建者、模型提供商和旨在使“多语言覆盖”声明可验证而非仅仅陈述的会议场所的具体建议。数据集：https://huggingface.co/datasets/ChialukaOnuoha/safety-slice-audit
cs.AI / 31 / 2608.14399
Whose doctor does the AI recommend? An algorithm audit of reputation and demographic signals in large language model-assisted physician choice
AI推荐哪位医生？大型语言模型辅助医生选择中声誉与人口统计信号的算法审计
large language model
大语言模型相关
Abstract
Patients increasingly ask large language model (LLM) assistants which doctor to see, making these systems AI infomediaries: algorithms that intermediate one person's choice among other people and thereby decide, silently and at scale, which physicians become visible. We report a prespecified randomized algorithm audit of what causally moves those recommendations. Seven models (six open-weight; gpt-4o-mini) each chose among five synthetic family-medicine physician cards whose attributes were independently randomized across 3,024 choice sets, three patient personas, nine prompt paraphrases and nine experimental arms, yielding 40,068 scored responses; gender and ethnicity were signaled through names following correspondence-audit methodology. Reputation signals dominate: raising a rating from 3.9 to 4.7 increases choice probability by 31.4 percentage points (pp), and raising the fee from $90 to $190 lowers it by 20.0 pp. Demographic parity is rejected, but not in the direction human audit studies predict: female-signaled names gain 2.5 pp, and Hispanic-, South-Asian- and Black-signaled names gain 1.3-2.9 pp over White-signaled names, tilts worth $7-$14 per visit in fee-equivalent terms, and a content-free first-listed position is worth $11. Yet models mentioned gender or ethnicity in at most 0.03% of their stated reasons and abstained in 0.39% of trials, so these effects are invisible in the models' own explanations, and transparency obligations relying on model self-report would not detect them. One reasoning model failed the prespecified auditability gate outright. The frozen design makes the audit repeatable: any new model can be assessed against identical stimuli, making recurring behavioural audit, rather than self-reported explanation, the monitoring technology fit for purpose.
Chinese Translation
患者越来越多地询问大型语言模型（LLM）助手应该看哪位医生，这使得这些系统成为AI信息中介：即在人与人之间调解个人选择的算法，从而悄无声息地、大规模地决定哪些医生会进入视野。我们报告了一项预先设定的随机算法审计，以探究是什么在因果层面推动了这些推荐。七个模型（六个开放权重模型；gpt-4o-mini）各自从五张合成家庭医学医生卡片中进行选择，这些卡片的属性在3,024个选择集、三个患者角色、九个提示改写和九个实验臂中被独立随机化，产生了40,068个评分响应；性别和族裔通过姓名来标识，遵循通信审计方法。声誉信号占主导地位：将评分从3.9提高到4.7会使选择概率增加31.4个百分点（pp），而将费用从$90提高到$190则使其降低20.0个百分点。人口统计均等性被拒绝，但方向不同于人类审计研究所预测的：女性信号姓名获得2.5个百分点的优势，西班牙裔、南亚裔和黑人信号姓名相对于白人信号姓名获得1.3至2.9个百分点的优势，这些倾斜按费用等价计算相当于每次就诊价值7至14美元，而一个无内容的首次列出的位置价值11美元。然而，模型在其陈述的理由中最多有0.03%提到性别或族裔，并且在0.39%的试验中放弃回答，因此这些效应在模型自身的解释中是不可见的，依赖模型自我报告的透明度义务也无法发现它们。一个推理模型直接未能通过预先设定的可审计性门槛。冻结的设计使审计具有可重复性：任何新模型都可以根据相同的刺激进行评估，从而使得定期行为审计——而非自我报告的解释——成为适合目的的监测技术。
cs.AI / 32 / 2608.13900
Agentic Transaction: Towards ACID-Compliant Agent Systems
智能体事务：迈向符合ACID的智能体系统
large language model
大语言模型相关
Abstract
Large language model (LLM) agents are evolving from conversational assistants into autonomous systems that execute long-horizon tasks through reasoning, tool use, code generation, and workspace manipulation. As agents increasingly operate over persistent environments and multi-step workflows, they face challenges analogous to those addressed by transactional database systems: reliable execution, consistent outcomes, safe concurrency, and durable state management. We introduce the concept of an agentic transaction and propose an ACID-compliant agent system framework that reinterprets the classical ACID properties for agent execution through four semantic guarantees: Semantic Atomicity, Semantic Consistency, Semantic Isolation, and Semantic Durability. Together, these properties provide a principled foundation for building reliable agent systems despite model uncertainty and dynamic execution environments. To instantiate this framework, we develop an ACID-compliant data agent that realizes these guarantees through transactional exploration-execution-validation cycles, transactional skill hubs, confidence divergence-based validation, semantic dependency-aware isolation, and transaction-aware semantic state management. Experimental results on widely used benchmarks show that our system achieves a 10.6% improvement over state-of-the-art agents, including Claude Code. This work opens a broader research agenda on extending transactional principles and system architectures toward building trustworthy, scalable, and self-evolving AI agent systems.
Chinese Translation
大型语言模型（LLM）智能体正从对话式助手演变为自主系统，通过推理、工具使用、代码生成和工作空间操作来执行长期任务。随着智能体越来越多地在持久化环境和多步骤工作流中运行，它们面临着与事务数据库系统所解决的挑战类似的问题：可靠执行、一致结果、安全并发和持久状态管理。我们引入了智能体事务的概念，并提出了一个符合ACID的智能体系统框架，该框架通过四个语义保证重新诠释了经典ACID属性在智能体执行中的含义：语义原子性、语义一致性、语义隔离性和语义持久性。这些属性共同为构建可靠的智能体系统提供了原则性的基础，尽管存在模型不确定性和动态执行环境。为了实例化该框架，我们开发了一个符合ACID的数据智能体，通过事务性探索-执行-验证循环、事务性技能中心、基于置信度差异的验证、语义依赖感知的隔离以及事务感知的语义状态管理来实现这些保证。在广泛使用的基准测试上的实验结果表明，我们的系统比最先进的智能体（包括Claude Code）取得了10.6%的提升。这项工作开辟了一个更广泛的研究议程，即将事务性原则和系统架构扩展到构建可信、可扩展和自进化的AI智能体系统。
cs.SE / 33 / 2608.14527
Validating LLM-Modernized Scientific Software Through Differential Fault Injection
通过差分故障注入验证经 LLM 现代化的科学软件
large language model
大语言模型相关
Abstract
Large language model (LLM) agents are increasingly used to modernize the legacy Fortran underlying production scientific software, but validation of these transformations emphasizes nominal executions and may not test whether a modernization preserves the original code's response to faults, perturbations, and reduced precision. We present a differential fault-injection validation method: a harness instruments the shared self-consistent-field driver of GAMESS at twelve sites and applies identical, deterministic faults to the original and LLM-modernized implementations, isolating the converted integral kernels. Across more than 2,200 runs, transient-fault absorption costs match a contraction-based model (predicted slopes 0.74 and 1.49 iterations per bit; measured 0.82 and 1.50), persistent perturbations halve final-energy error per additional bit, and the campaigns expose phase-dependent parallel deadlocks and false convergence under reduced precision. The original and modernized kernels agree in all 200 paired injections, and a measurement-guided synchronization change composes with the modernization, matching in all 40 pairs.
Chinese Translation
大型语言模型（LLM）智能体越来越多地被用于对生产科学软件底层的遗留 Fortran 代码进行现代化改造，但对此类转换的验证侧重于名义执行，可能无法测试现代化是否保留了原始代码对故障、扰动和降低精度的响应。我们提出了一种差分故障注入验证方法：一个测试工具在 GAMESS 的共享自洽场驱动程序的十二个位置进行插桩，并对原始实现和 LLM 现代化实现施加相同且确定性的故障，从而隔离转换后的积分核。在超过 2,200 次运行中，瞬态故障吸收成本与基于收缩的模型相匹配（预测斜率分别为每比特 0.74 和 1.49 次迭代；实测为 0.82 和 1.50），持续扰动使每个额外比特的最终能量误差减半，并且这些测试活动暴露了相依赖的并行死锁以及降低精度下的虚假收敛。原始内核与现代化内核在所有 200 次配对注入中结果一致，并且一项由测量引导的同步修改与现代化改造相结合，在所有 40 个配对中均匹配。
cs.AI / 34 / 2608.14132
Act2Intention: A Benchmark For Developing Active Mobile Agents Through Inferring User Intention from GUI Actions
Act2Intention: 一个通过从GUI操作推断用户意图来开发主动式移动智能体的基准
large language model
大语言模型相关
Abstract
Mobile GUI Agents powered by multimodal large language models (MLLMs) show promise in human-computer intelligence. However, current research primarily focuses on reactive task execution while lacking a comprehensive understanding-prediction-execution process for user intentions, which are the core requirements of active agents. In this paper, we propose the Act2Intention framework that builds an active mobile agent by integrating understanding, predicting user intentions, and executing decisions. First, we construct the Act2Intention Bench through data collection and validated generation, comprising 72,511 intentions and over 700,000 actions across 52 apps, thereby establishing the first benchmark for evaluating proactive agents via continuous intention-action trajectories. We further develop the Act2Intention Agent, achieving proactive services through Proactive-oriented Intention Understanding, Personalized Proactive Intention Prediction, and Experience-guided Intention Execution. Experimental results show that supervised fine-tuning on Act2Intention Bench yields absolute improvements of +32.0 Acc-S, +10.25 Acc-S, and +6.9 SSR points over non-fine-tuned counterparts under the same agent framework for intention understanding, prediction, and execution, respectively. This success underscores the necessity and value of the Act2Intention Bench, which establishes a standardized platform for developing and evaluating proactive agents and consequently paves the way for research on intention-driven human-computer interaction.
Chinese Translation
由多模态大语言模型（MLLMs）驱动的移动GUI智能体在人机智能方面展现出潜力。然而，当前研究主要关注反应式任务执行，缺乏对用户意图的理解-预测-执行完整过程，而用户意图是主动智能体的核心要求。在本文中，我们提出Act2Intention框架，通过整合理解、预测用户意图和执行决策来构建主动式移动智能体。首先，我们通过数据收集和验证生成构建了Act2Intention Bench，包含52个应用中的72,511个意图和超过700,000个操作，从而建立了通过连续意图-操作轨迹评估主动式智能体的首个基准。我们进一步开发了Act2Intention Agent，通过面向主动的意图理解、个性化主动意图预测和经验引导的意图执行来实现主动服务。实验结果表明，在Act2Intention Bench上进行监督微调，在相同智能体框架下，与未微调的对应模型相比，在意图理解、预测和执行方面分别取得了+32.0 Acc-S、+10.25 Acc-S和+6.9 SSR分的绝对提升。这一成功凸显了Act2Intention Bench的必要性和价值，它为开发和评估主动式智能体建立了标准化平台，从而为意图驱动的人机交互研究铺平了道路。
cs.AI / 35 / 2608.13786
Do AI chatbots find what experts would? Effects of model, user role, and sample size on study retrieval for medical questions
AI 聊天机器人能找到专家会找到的吗？模型、用户角色和样本量对医学问题研究检索的影响
large language model
大语言模型相关
Abstract
Large language model (LLM) chatbots are increasingly used to answer clinical questions with citations to relevant clinical studies. Prior research has largely focused on citation fabrication, leaving a gap in evaluating the quality of retrieved studies and the factors driving their selection. In this study, we evaluated three general-purpose LLM chatbots: Claude Sonnet 5, Gemini 3.1 Pro, and ChatGPT GPT-5.5. We prompted the models with clinical questions adapted from 20 review questions in Issues 6 and 7 of the 2026 Cochrane Database of Systematic Reviews, simulating patient, clinician, and evidence-synthesis researcher roles. Each chatbot was queried under each user role with four independent repetitions, yielding 720 responses. Each chatbot was asked to support its answers with primary clinical citations, which we benchmarked against the included and excluded study sets of the Cochrane reviews. On average, a chatbot response retrieved 39.2% $\pm$ 29.8% of Cochrane included studies, while citing 5.0% $\pm$ 9.4% of excluded studies. Recall of Cochrane included studies varied significantly by model and user role. ChatGPT achieved higher recall than Claude or Gemini (63.1% $\pm$ 29.5% vs. 37.0% $\pm$ 23.8% vs. 17.3% $\pm$ 13.1%; $p=2.0\times10^{-5}$). The researcher role yielded higher recall than the clinician or patient roles (42.8% $\pm$ 30.8% vs. 38.6% $\pm$ 28.9% vs. 36.1% $\pm$ 29.3%; $p=2.0\times10^{-5}$). Controlling for publication year, citations per year, and open-access status, sample size was the only independently significant predictor of retrieval (odds ratio 1.80 per 1-unit increase in log sample size, 95% CI 1.37-2.36, $p=2.34\times10^{-5}$). These findings suggest that while LLM chatbots can retrieve some studies identified by expert reviewers, their performance varies by model and user role, and they exhibit a bias toward clinical trials with larger sample sizes.
Chinese Translation
大语言模型（LLM）聊天机器人越来越多地被用于回答临床问题，并引用相关的临床研究。以往的研究主要关注引用伪造，在评估检索到的研究质量及其选择驱动因素方面存在空白。在本研究中，我们评估了三个通用大语言模型聊天机器人：Claude Sonnet 5、Gemini 3.1 Pro 和 ChatGPT GPT-5.5。我们使用改编自 2026 年 Cochrane 系统评价数据库第 6 期和第 7 期中 20 个评价问题的临床问题对模型进行提示，模拟患者、临床医生和证据综合研究者等角色。每个聊天机器人在每种用户角色下被查询四次独立重复，共产生 720 个响应。每个聊天机器人被要求用主要的临床引用支持其答案，我们将这些引用与 Cochrane 评价的纳入和排除研究集进行了基准比较。平均而言，一个聊天机器人响应检索到了 Cochrane 纳入研究的 39.2% $\pm$ 29.8%，同时引用了排除研究的 5.0% $\pm$ 9.4%。Cochrane 纳入研究的召回率因模型和用户角色而显著不同。ChatGPT 的召回率高于 Claude 或 Gemini（63.1% $\pm$ 29.5% 对比 37.0% $\pm$ 23.8% 对比 17.3% $\pm$ 13.1%；$p=2.0\times10^{-5}$）。研究者角色的召回率高于临床医生或患者角色（42.8% $\pm$ 30.8% 对比 38.6% $\pm$ 28.9% 对比 36.1% $\pm$ 29.3%；$p=2.0\times10^{-5}$）。在控制出版年份、年均引用次数和开放获取状态后，样本量是唯一独立显著的检索预测因子（对数样本量每增加 1 个单位的比值比为 1.80，95% CI 1.37-2.36，$p=2.34\times10^{-5}$）。这些发现表明，虽然大语言模型聊天机器人能够检索到一些由专家评审员确定的研究，但其表现因模型和用户角色而异，并且它们表现出对样本量较大的临床试验的偏向。
cs.AI / 36 / 2608.14068
MACS: A Hybrid Multi-Agent Framework for Reliable Conversational E-Commerce Recommendation
MACS：用于可靠对话式电子商务推荐的混合多智能体框架
large language model
大语言模型相关
Abstract
Conversational recommendation for e-commerce is increasingly mediated by large language models (LLMs), yet many real-world deployments operate under a stricter requirement: recommendations must be drawn only from a merchant's fixed catalog, without web search or unsupported product claims. In this setting, the main challenge is reliability under hard constraints: the system must satisfy user requirements, remain grounded in available inventory, and preserve preferences across multiple conversational turns. We present MACS (Multi-Agent Commerce System), a hybrid multi-agent framework for reliable conversational recommendation in fixed-catalog settings. MACS uses LLMs for language-facing tasks such as interpreting user requests, eliciting preferences, and generating responses, while correctness-critical operations, including product retrieval, hard-constraint filtering, brand exclusion, and progressive relaxation, are executed deterministically by the merchant agent. A session-persistent preference layer tracks constraints across turns, enabling consistent handling of budget overwrites and exclusion reversals. On a 140-query single-turn benchmark, MACS achieves the highest pass rate (87.1%) and perfect brand compliance (1.000). On a 10-scenario multi-turn benchmark, MACS achieves the strongest macro Pass@5 (72% vs. 56% GPT+Catalog / 52% Gemini+Catalog) with zero constraint drift. The advantage is sharpest on exclusion reversal (100% vs. 20% / 0%) and constraint accumulation (100% vs. 60% / 40%). Mean judged response quality is similar across systems (0.751 vs. 0.736). These results suggest that hybrid architectures combining deterministic constraint enforcement with session-persistent preference tracking provide stronger reliability-oriented performance than catalog-bound prompt-only baselines in the fixed-catalog merchant setting.
Chinese Translation
电子商务中的对话式推荐越来越多地由大语言模型（LLMs）中介实现，然而许多实际部署面临更严格的要求：推荐只能来自商家的固定目录，不能进行网络搜索或提出无依据的产品声明。在这种场景下，主要挑战是硬约束下的可靠性：系统必须满足用户需求，保持以可用库存为依据，并在多个对话轮次中保持偏好。我们提出了MACS（多智能体商务系统），一种用于固定目录场景下可靠对话式推荐的混合多智能体框架。MACS使用LLMs处理面向语言的任务，如解释用户请求、引导偏好和生成响应，而正确性关键操作，包括产品检索、硬约束过滤、品牌排除和渐进式放松，则由商家智能体以确定性方式执行。一个会话持久的偏好层跨轮次追踪约束，从而能够一致地处理预算覆盖和排除逆转。在140个查询的单轮基准测试中，MACS达到了最高的通过率（87.1%）和完美的品牌符合度（1.000）。在10个场景的多轮基准测试中，MACS取得了最强的宏观Pass@5（72%对比GPT+Catalog的56%和Gemini+Catalog的52%），且零约束漂移。在排除逆转（100%对比20%/0%）和约束累积（100%对比60%/40%）上优势最为显著。各系统的人工评判响应质量均值相近（0.751对比0.736）。这些结果表明，在固定目录商户场景中，将确定性约束强制执行与会话持久偏好追踪相结合的混合架构，比仅使用提示语且受限于目录的基线方法提供了更强的以可靠性为导向的性能。
cs.LG / 37 / 2608.13844
Federated Prompt Learning: A Unified Framework, Empirical Analysis, and Future Directions
联邦提示学习：统一框架、实证分析与未来方向
large language model
大语言模型相关
Abstract
Large language models (LLMs) have become core components of cloud-based intelligent services in academia and industry, yet their training and deployment are hindered by high computational costs, data centralization, and privacy concerns. Federated learning (FL) offers a decentralized training paradigm that enables clients to collaboratively train a learning model without sharing raw data, making it a promising solution for privacy-preserving LLM training and reasoning. This paper presents a comprehensive survey of federated prompt learning (FPL) to review recent advances in integrating the federated learning paradigm and large language models, answering the following research questions: RQ1: The fundamental motivations, characteristics, and enabling technologies of FPL, and how it differs from conventional FL and full-model federated fine-tuning; RQ2: The trade-offs FPL approaches exhibit in performance, communication efficiency, computational overhead, scalability, personalization, and heterogeneity handling; RQ3: The remaining security, privacy, robustness, and system challenges, along with key future research directions. To this end, we systematically examine existing FPL methods across the full model lifecycle: pre-training, fine-tuning, and practical applications, while discussing security, privacy, and robustness issues and summarizing existing defense mechanisms. Finally, we highlight open challenges and future directions, aiming to help readers understand how the insights drive research in FPL.
Chinese Translation
大型语言模型（LLMs）已成为学术界和工业界基于云智能服务的核心组件，但其训练和部署受到高计算成本、数据集中化和隐私问题的阻碍。联邦学习（FL）提供了一种去中心化的训练范式，使客户端能够在不共享原始数据的情况下协作训练学习模型，使其成为隐私保护的大型语言模型训练和推理的有前景的解决方案。本文对联邦提示学习（FPL）进行了全面综述，以回顾将联邦学习范式与大型语言模型集成的最新进展，并回答以下研究问题：RQ1：FPL 的基本动机、特征和支持技术，以及它与传统 FL 和全模型联邦微调有何不同；RQ2：FPL 方法在性能、通信效率、计算开销、可扩展性、个性化和异构性处理方面表现出的权衡；RQ3：剩余的安全、隐私、鲁棒性和系统挑战，以及关键的未来研究方向。为此，我们系统地审视了现有 FPL 方法在整个模型生命周期中的表现：预训练、微调和实际应用，同时讨论安全、隐私和鲁棒性问题，并总结现有防御机制。最后，我们强调开放的挑战和未来方向，旨在帮助读者理解这些见解如何驱动 FPL 研究。
cs.LG / 38 / 2608.13866
Geometric Filtering of LLM-Generated Samples for Few-Shot Text Classification
面向少样本文本分类的LLM生成样本的几何过滤
large language model
大语言模型相关
Abstract
Large language models (LLMs) can generate synthetic training data for text classification, but the quality of generated samples is heterogeneous: some fall in correct class regions of the embedding space while others land in peripheral or cross-class zones. We propose a geometric filtering framework that evaluates each LLM-generated sample by its Euclidean distance to real class examples in a sentence embedding space, selecting only geometrically consistent candidates. A soft weighting mechanism transforms filter scores into sample weights for classifier training. Evaluated across 13 datasets, 5 classifiers, 10 augmentation methods, and over 6,700 configurations, our method achieves +2.61 percentage points (pp) over SMOTE ($p<0.0001$, Cohen's $d=0.95$, 88.9% win rate). The approach generalizes to named entity recognition (+9.26pp, 100% win rate) without filter modification, and is robust across 5 LLMs from 4 providers. A key finding is that the simplest distance-based filter consistently outperforms complex multi-criteria alternatives.
Chinese Translation
大型语言模型（LLMs）可以为文本分类生成合成训练数据，但生成样本的质量是异质性的：一些样本落在嵌入空间中正确的类别区域内，而另一些则落在边缘区域或跨类别区域。我们提出了一种几何过滤框架，该框架通过计算每个LLM生成样本到句子嵌入空间中真实类别示例的欧氏距离来评估样本，仅选择几何上一致的候选样本。一种软加权机制将过滤分数转换为用于分类器训练的样本权重。在13个数据集、5个分类器、10种增强方法以及超过6,700种配置上进行评估，我们的方法相比SMOTE实现了+2.61个百分点（pp）的提升（$p<0.0001$，Cohen's $d=0.95$，88.9%胜率）。该方法无需修改过滤即可泛化到命名实体识别（+9.26pp，100%胜率），并且对来自4个提供商的5种LLM具有鲁棒性。一个关键发现是，最简单的基于距离的过滤器始终优于复杂的多标准替代方案。
cs.LG / 39 / 2608.13925
CForce: Boosting Parallel Decoding for dLLMs via Consistency Forcing
CForce：通过一致性强制提升dLLMs的并行解码
diffusionlarge language model
扩散模型相关
大语言模型相关
Abstract
Diffusion large language models (dLLMs) accelerate language generation by predicting multiple masks in a single forward pass. However, existing dLLMs can suffer from unreliable predictions in early denoising stages under aggressive parallelism strategies, leading to errors that can propagate to later stages. To tackle this issue, we present Consistency Forcing (CForce) for dLLMs, a distillation method to force the mask predictions of early stages to align with those of later stages. CForce trains the model on pre-collected self-rollout trajectories, thereby improving training-inference alignment. We introduce Confidence Adaptive KL Divergence as a distillation objective to conjoin the merits of forward and reverse KL. We further provide a theoretical analysis for the consistency objective to explain why CForce can approximately minimize the prediction error of early stages. Critically, the same formulation applies to both mask-to-token decoding and edit-capable decoding; in the edit-capable case, later token-to-token refinements provide additional supervision for earlier masked-state predictions. Experiments on non-edit and edit-capable LLaDA models show improved speed-quality trade-offs, especially under high-parallelism decoding budgets. Code is available at: https://github.com/inclusionAI/dFactory.
Chinese Translation
扩散大语言模型（dLLMs）通过在单次前向传播中预测多个掩码来加速语言生成。然而，现有的dLLMs在激进的并行策略下，可能会在早期去噪阶段出现不可靠的预测，从而导致错误传播到后续阶段。为了解决这一问题，我们提出了用于dLLMs的一致性强制（CForce）方法，这是一种蒸馏方法，迫使早期阶段的掩码预测与后期阶段的预测保持一致。CForce在预先收集的自推演轨迹上训练模型，从而改善训练与推理的对齐。我们引入置信度自适应KL散度作为蒸馏目标，以结合前向KL和反向KL的优点。我们进一步对一致性目标进行了理论分析，以解释为什么CForce能够近似最小化早期阶段的预测误差。关键的是，相同的公式既适用于掩码到token解码，也适用于可编辑解码；在可编辑情况下，后期token到token的细化过程为早期掩码状态预测提供了额外的监督。在非编辑和可编辑LLaDA模型上的实验显示，速度与质量的权衡得到改善，尤其是在高并行解码预算下。代码可在以下网址获取：https://github.com/inclusionAI/dFactory。
cs.LG / 40 / 2608.13932
Post-training Quantization for Hybrid Iterative Generative Models
混合迭代生成模型的训练后量化
diffusion
扩散模型相关
Abstract
Iterative Generative Models (IGMs) span autoregressive and diffusion paradigms, and hybrid variants that couple them can achieve remarkable image-generation fidelity. However, their iterative inference incurs substantial computational overhead, making Post-training Quantization (PTQ) appealing for acceleration, while directly applying vanilla PTQ to hybrid IGMs can trigger model collapse. By analyzing these failures, we identify two critical challenges: Excessive Outliers (EOs) in the activations create an irreconcilable trade-off between preserving normal precision and covering EOs, resulting in severe degradation in generation quality; Amplified Anomalies (AAs) arising unpredictably from minor quantization errors, create a mismatch between calibration and inference, thus iteratively triggering model collapse. To address these challenges, we introduce HyGenQ, a PTQ framework for hybrid IGMs. HyGenQ comprises Hierarchical Cluster Decoupling (HCD) and Scaling Recalibration (SR). HCD identifies and decouples outlier channels via a multi-stage clustering process, effectively isolating EOs while maintaining normal value precision, thereby alleviating performance degradation. SR scales AAs beyond Gaussian Bound, thereby avoiding model collapse caused by aggressive truncation. Extensive experiments demonstrate that HyGenQ successfully quantizes representative hybrid IGMs to 8-bit precision (W8A8), significantly outperforming existing baselines and validating its robustness across different model families.
Chinese Translation
迭代生成模型（IGM）涵盖自回归和扩散范式，而将二者耦合的混合变体能够实现显著的图像生成保真度。然而，其迭代推理会带来大量的计算开销，这使得训练后量化（PTQ）成为一种有吸引力的加速手段，而将朴素PTQ直接应用于混合IGM可能引发模型崩溃。通过分析这些失败案例，我们识别出两个关键挑战：激活中的过度离群值（EOs）在保持正常精度与覆盖EOs之间造成了不可调和的权衡，导致生成质量严重下降；由微小量化误差不可预测地引发的放大异常（AAs），造成了校准与推理之间的不匹配，从而迭代性地触发模型崩溃。为应对这些挑战，我们提出了HyGenQ，一个面向混合IGM的PTQ框架。HyGenQ包含层次聚类解耦（HCD）和缩放重新校准（SR）。HCD通过多阶段聚类过程识别并解耦离群通道，在保持正常值精度的同时有效隔离EOs，从而缓解性能下降。SR将AAs缩放到高斯界限之外，从而避免由激进截断引起的模型崩溃。大量实验表明，HyGenQ成功地将具有代表性的混合IGM量化到8位精度（W8A8），显著优于现有基线，并验证了其在不同模型家族中的鲁棒性。
cs.LG / 41 / 2608.13966
QUASAR: Lowering the Loss Floor of Quantization-Aware Training with Loss-Aware Reconstruction
QUASAR：通过损失感知重建降低量化感知训练的损失下限
large language model
大语言模型相关
Abstract
As large language model inference shifts toward lower precision, post-training quantization (PTQ) becomes increasingly brittle, making quantization-aware training (QAT) essential for preserving model quality. However, QAT computes the loss and surrogate gradients using a lossy reconstruction of latent full-precision weights, while applying updates to the latent weights themselves. This mismatch can lead to suboptimal training trajectories and a higher loss floor. Second-order PTQ methods mitigate a similar gap by minimizing loss-aware reconstruction error, but doing it once for a frozen model can take hours; repeating this process throughout QAT as the weights evolve is impractical. We introduce QUASAR, a QAT method that continuously performs lightweight, loss-aware reconstruction in the training loop to lower the loss floor and improve the resulting low-bit model. At each training step, QUASAR uses the exponential moving average of squared gradients as online saliency estimates, searches over a small set of clipping ranges, and fits affine dequantizers via saliency-weighted least squares. Our analysis shows that the loss-aware reconstruction error is the only reconstruction-dependent term in the QAT convergence bound and controls the loss of the final quantized model, establishing QUASAR's objective as a principled optimization target. QUASAR modifies only the training procedure and supports standard deployment formats, including integer quantization and NVFP4, with no inference-time changes or overhead. Across Qwen3 and Llama-3.1, QUASAR achieves the lowest held-out KL divergence among competitive QAT methods at 2, 3, and 4 bits, reducing KL by at least 10% at 3 and 4 bits and by 29% at 2 bits. At 2 bits, it improves average accuracy across eight tasks by 3.5-4.3 percentage points over strong QAT and PTQ baselines.
Chinese Translation
随着大语言模型推理向更低精度转变，训练后量化（PTQ）变得越来越脆弱，使得量化感知训练（QAT）对于保持模型质量至关重要。然而，QAT使用潜在全精度权重的有损重建来计算损失和替代梯度，同时对潜在权重本身应用更新。这种不匹配可能导致次优的训练轨迹和更高的损失下限。二阶PTQ方法通过最小化损失感知重建误差来缓解类似的差距，但对冻结模型进行一次这样的操作可能需要数小时；在QAT过程中随着权重演化而重复这一过程是不切实际的。我们引入了QUASAR，这是一种QAT方法，在训练循环中持续执行轻量级的损失感知重建，以降低损失下限并改进最终的低比特模型。在每个训练步骤中，QUASAR使用平方梯度的指数移动平均作为在线显著性估计，在一小组裁剪范围内进行搜索，并通过显著性加权最小二乘法拟合仿射反量化器。我们的分析表明，损失感知重建误差是QAT收敛界中唯一与重建相关的项，并且控制最终量化模型的损失，从而将QUASAR的目标确立为一个有原则的优化目标。QUASAR仅修改训练过程，并支持标准部署格式，包括整数量化和NVFP4，在推理时不需更改或产生额外开销。在Qwen3和Llama-3.1上，QUASAR在2、3和4比特下实现了竞争性QAT方法中最低的留出KL散度，在3和4比特下KL至少降低10%，在2比特下降低29%。在2比特下，相较于强QAT和PTQ基线，它在八个任务上的平均准确率提高了3.5至4.3个百分点。
cs.LG / 42 / 2608.14038
Adversarial Learning of Classifier-Free Guidance Schedules
无分类器指导调度表的对抗学习
diffusion
扩散模型相关
Abstract
Modern text-to-image diffusion models rely on classifier-free guidance (CFG) to achieve high image fidelity and text alignment. However, CFG typically applies a static, global scale across all timesteps, samples, and conditions -- a choice that is generally suboptimal and can introduce artifacts, as different states may benefit from different levels of guidance. While time-varying schedules are known to improve quality, designing them by hand is non-trivial and application-dependent. In this paper, we learn the guidance schedule as a function of diffusion time, conditioning and the current noisy sample, in order to better align sampled images with the text prompt. We frame this as a density ratio estimation problem: a discriminator is trained to estimate the time-dependent log-density ratio between the true and guided marginal distributions, while a lightweight generator network predicts the optimal, state-dependent guidance scale. Empirically, our approach outperforms both heuristic CFG schedules and prior methods for learning dynamic guidance on text-to-image generation benchmarks.
Chinese Translation
现代文本到图像扩散模型依赖无分类器指导（CFG）来实现高图像保真度和文本对齐。然而，CFG通常在所有时间步、样本和条件上应用一个静态的全局尺度——这个选择通常不是最优的，并且可能引入伪影，因为不同状态可能受益于不同级别的指导。虽然已知随时间变化的调度表可以改善质量，但手工设计它们并非易事，且依赖于具体应用。在本文中，我们将指导调度表学习为扩散时间、条件和当前噪声样本的函数，以便更好地将采样图像与文本提示对齐。我们将此问题构建为密度比估计问题：训练一个判别器来估计真实分布与指导边缘分布之间的时间相关对数密度比，同时一个轻量级生成器网络预测最优的、状态相关的指导尺度。在经验上，我们的方法在文本到图像生成基准上优于启发式CFG调度表和先前学习动态指导的方法。
cs.LG / 43 / 2608.14067
When Denoising Hurts: Rethinking the Terminal Step of Diffusion Time Series Forecasters -- Extended Version
当去噪有害时：重新思考扩散时间序列预测器的终端步骤——扩展版
diffusion
扩散模型相关
Abstract
Diffusion models offer a natural way to model uncertainty in time series forecasting, yet their iterative sampling process is often treated as a uniformly beneficial refinement procedure. Our study challenges this view by examining how forecast quality evolves throughout reverse diffusion. We find that general temporal structure is often recovered at relatively high noise levels, whereas continued low-noise refinement can introduce statistical drift and degrade the final forecast. Our analysis further suggests that this behavior explains why prior methods often favor relatively narrow diffusion architecture and schedule design. Building on this observation, we propose a label-free global stopping criterion that detects the optimal termination point, eventually speeding up inference and improving predictive accuracy. Additionally, since early stopping terminates inference in high-noise regions, we propose a Bernoulli timestep sampler that concentrates training on this region while preserving coverage of the full diffusion process. Extensive experiments conducted across eight real-world datasets demonstrate the superior performance of our method compared to existing approaches.
Chinese Translation
扩散模型为时间序列预测中的不确定性建模提供了一种自然方式，然而其迭代采样过程常被视为一种均匀有益的细化程序。我们的研究通过考察预测质量在整个反向扩散过程中的演变来挑战这一观点。我们发现，一般的时间结构通常在相对较高的噪声水平下被恢复，而持续的低噪声细化可能引入统计漂移并降低最终预测质量。我们的分析进一步表明，这一行为解释了为什么先前的方法常常偏好相对狭窄的扩散架构和调度设计。基于这一观察，我们提出了一种无标签的全局停止准则，用于检测最优终止点，从而加速推理并提高预测精度。此外，由于早停会在高噪声区域终止推理，我们提出了一种伯努利时间步采样器，将训练集中于此区域，同时保持对整个扩散过程的覆盖。在八个真实世界数据集上进行的大量实验表明，与现有方法相比，我们的方法具有更优越的性能。
cs.LG / 44 / 2608.14303
Detecting Contaminated Code-Generation Prompt Batches via Influence Functions
通过影响函数检测受污染的代码生成提示批次
large language model
大语言模型相关
Abstract
Large language models (LLMs) are increasingly used for code generation, yet they remain vulnerable to prompts that elicit insecure implementations. Existing defenses typically rely on predefined threat models or known vulnerability patterns, limiting their effectiveness against novel attacks. We propose CodeSIFT, a threat-model-agnostic detection method that leverages influence functions to identify batches of prompts that induce anomalous model behavior. Rather than detecting specific vulnerabilities, CodeSIFT measures the parameter-space influence of generated code and uses a statistical test to determine whether a candidate prompt set deviates from a benign reference distribution. To evaluate our approach, we introduce two benchmark datasets covering a variety of vulnerabilities. We evaluate CodeSIFT on three open-weight code LLMs ranging from 3B to 7B parameters, achieving AUROC scores of up to 0.98 at moderate-to-high injection rates, while maintaining well-calibrated false positive rates and substantially outperforming static analysis baselines. These results suggest that influence-function-based detection is a promising direction for identifying malicious code-generation prompts without requiring prior knowledge of the underlying attack class.
Chinese Translation
大型语言模型（LLMs）越来越多地用于代码生成，但它们仍然容易受到引发不安全实现的提示的影响。现有的防御方法通常依赖于预定义的威胁模型或已知的漏洞模式，这限制了它们应对新型攻击的有效性。我们提出CodeSIFT，一种与威胁模型无关的检测方法，利用影响函数来识别引发模型异常行为的提示批次。CodeSIFT不是检测特定漏洞，而是测量生成代码在参数空间中的影响，并使用统计检验来确定候选提示集是否偏离良性参考分布。为了评估我们的方法，我们引入了两个涵盖多种漏洞的基准数据集。我们在三个参数规模从3B到7B的开放权重代码大语言模型上评估了CodeSIFT，在中高注入率下实现了高达0.98的AUROC分数，同时保持了良好校准的假阳性率，并显著优于静态分析基线。这些结果表明，基于影响函数的检测是识别恶意代码生成提示的一个有前景的方向，无需预先了解底层攻击类别。
cs.LG / 45 / 2608.14430
Designing Reinforcement Learning for Diffusion Models: A Unified Path-Space View
为扩散模型设计强化学习：一种统一的路径空间视角
diffusion
扩散模型相关
Abstract
Reinforcement learning (RL) post-training provides a direct way to align diffusion models with human preferences and task-specific rewards. However, current RL algorithms for diffusion models remain fragmented: reverse-trajectory methods rely on discretized likelihood ratios, whereas forward-matching methods train on reward-labeled noising versions of the rollout samples. This paper shows that these seemingly different losses arise from a single path-space principle. Starting from the regularized diffusion-RL objective, we use importance sampling between sampling SDEs to obtain an explicit policy-gradient estimator on trajectory space. The estimator contains the stochastic Itô integral underlying Flow-GRPO-type updates; we derive an equivalent variance-reduced value-gradient form that recovers the forward-matching structure of AWM and DiffusionNFT. This identifies the empirical gap between these method families as a variance-reduction effect rather than a difference in RL principle. The derivation yields a unified design space organized by value-gradient estimation, weight functions, and sampling choices. Within this space, we propose a multi-sample KDE value-gradient estimator that reuses rollout groups, together with scale-bounded weight families that retain stable existing recipes while excluding singular ones. Experiments on SD3.5-M and Qwen-Image models validate the variance-reduction explanation and show that the resulting recipe improves over prior diffusion-RL baselines.
Chinese Translation
强化学习（RL）后训练提供了一种直接的方式，使扩散模型与人类偏好和任务特定奖励对齐。然而，当前用于扩散模型的RL算法仍然支离破碎：反向轨迹方法依赖于离散化的似然比，而前向匹配方法则在带有奖励标签的加噪版本的rollout样本上进行训练。本文表明，这些看似不同的损失源于一个单一的路径空间原理。从正则化的扩散-RL目标出发，我们在采样SDE之间使用重要性采样，以获得轨迹空间上的显式策略梯度估计器。该估计器包含Flow-GRPO类型更新所依据的随机Itô积分；我们推导出一个等价的方差缩减价值梯度形式，该形式恢复了AWM和DiffusionNFT的前向匹配结构。这将这两类方法之间的经验差距确定为方差缩减效应，而非RL原理上的差异。该推导产生了一个统一的设计空间，由价值梯度估计、权重函数和采样选择组织。在这个空间内，我们提出了一种多样本KDE价值梯度估计器，它重用rollout组，并配以尺度有界权重族，这些权重族保留了稳定的现有配方，同时排除了奇异的配方。在SD3.5-M和Qwen-Image模型上的实验验证了方差缩减的解释，并表明所得配方优于先前的扩散-RL基线。
cs.LG / 46 / 2608.14496
Generating Benchmark Health Data Using a Tabular Diffusion Transformer
使用表格扩散变压器生成基准健康数据
diffusion
扩散模型相关
Abstract
Cross-Tabular Data Generation (CTDG) seeks to learn a generative model from multiple heterogeneous tables and produce new synthetic tabular datasets. However, existing synthetic tabular data generation methods are largely restricted to single-input-table scenarios and struggle to effectively handle multiple heterogeneous tables with diverse feature sets. To address this limitation, we propose a two-stage framework for cross-tabular data generation. In the first stage, each heterogeneous raw table is transformed into a standardized statistical table with the same set of columns across all tables. Each statistical table captures the marginal distributions of the original columns and the pairwise correlations among them. In the second stage, a diffusion transformer model is trained to capture structural patterns across these homogeneous statistical tables and to generate synthetic statistical tables. Synthetic raw tables are subsequently reconstructed from the generated statistical tables via multivariate Gaussian sampling followed by an inverse probability integral transform. This two-stage CTDG framework enables the learning of a unified generative model from multiple heterogeneous tables and supports the generation of an unlimited number of realistic synthetic heterogeneous tables. Experimental results demonstrate high fidelity in the learned statistical representations and a favorable fidelity-diversity trade-off in the generated synthetic data, validating the effectiveness of the proposed approach.
Chinese Translation
跨表格数据生成（CTDG）旨在从多个异质表格中学习生成模型，并产生新的合成表格数据集。然而，现有的合成表格数据生成方法主要局限于单输入表格场景，难以有效处理具有多样化特征集的多个异质表格。为解决这一局限性，我们提出了一种用于跨表格数据生成的两阶段框架。在第一阶段，每个异质原始表格被转换为一个标准化的统计表格，所有表格具有相同的列集合。每个统计表格捕获原始列的边际分布以及它们之间的两两相关性。在第二阶段，训练一个扩散变压器模型，以捕捉这些同质统计表格中的结构模式，并生成合成统计表格。随后，通过多元高斯采样和逆概率积分变换，从生成的统计表格中重建合成原始表格。这一两阶段CTDG框架使得能够从多个异质表格中学习统一的生成模型，并支持生成无限数量的逼真合成异质表格。实验结果表明，所学习的统计表示具有较高的保真度，并且生成的合成数据在保真度-多样性权衡方面表现良好，验证了所提出方法的有效性。
cs.OS / 47 / 2608.14376
CoRun: Padding is Simple and Efficient for Deterministic LLM Inference
CoRun：填充对于确定性LLM推理既简单又高效
large language model
大语言模型相关
Abstract
Despite fixed sampling parameters and random seeds, Large Language Model (LLM) inference exhibits output inconsistency, which undermines downstream tasks such as model evaluation and reinforcement learning. A major source of this nondeterminism is batch-dependent GPU execution: dynamic input shapes change kernel tiling and floating-point reduction orders. Existing systems address this problem with batch-invariant kernels, but these kernels restrict optimized tiling and split reductions, increasing more than 2$\times$ latency and reducing serving throughput by up to 74 %. This paper observes that although most kernels are not batch-invariant, they are position-invariant. Leveraging this property, we present CoRun, a scheduling-based system that achieves deterministic inference without requiring batch invariance. CoRun employs isolated prefill and fixed-shape batched decode to handle the two stages of LLM inference, respectively, leveraging CUDA graphs for efficient execution and simplified implementation. Experiments on LLMs with diverse architectures, including Qwen and DeepSeek, show that CoRun ensures determinism while improving throughput by 15-324 % over batch-invariant approaches, reducing time-to-first-token by 51.8 % and time-per-output-token by 48.6 % on average.
Chinese Translation
尽管采样参数和随机种子是固定的，大语言模型（LLM）推理仍会表现出输出不一致性，这破坏了模型评估和强化学习等下游任务。这种非确定性的一个主要来源是依赖于批处理的GPU执行：动态输入形状会改变内核分块和浮点归约顺序。现有系统通过批处理不变的内核来解决这个问题，但这些内核限制了优化的分块和拆分归约，导致延迟增加超过2倍，服务吞吐量最多降低74%。本文观察到，尽管大多数内核并非批处理不变的，但它们具有位置不变性。利用这一特性，我们提出了CoRun，一个基于调度的系统，无需批处理不变性即可实现确定性推理。CoRun分别采用隔离的预填充和固定形状的批处理解码来处理LLM推理的两个阶段，并利用CUDA图实现高效执行和简化的实现。在包括Qwen和DeepSeek在内的多种架构LLM上的实验表明，与批处理不变的方法相比，CoRun在保证确定性的同时将吞吐量提高了15-324%，平均将首令牌时间降低了51.8%，每个输出令牌时间降低了48.6%。
cs.LG / 48 / 2608.13678
hint$^2$: Hierarchical World Models for Inference-Time Temporal Logic Guidance
hint$^2$：用于推理时时序逻辑引导的分层世界模型
diffusion
扩散模型相关
Abstract
A central goal of robot learning is to enable robots to execute rich instructions specified at runtime. Large-scale language-conditioned policies have made substantial progress toward this goal, yet still struggle with temporal structure and safety constraints. Linear Temporal Logic (LTL) provides a powerful language to express complex, non-Markovian instructions. However, guiding learned manipulation policies toward LTL satisfaction remains challenging because modern policies generate short-horizon action chunks and replan in closed loop, while almost all LTL specifications are evaluated over long-horizon trajectories. In this paper, we introduce hint$^2$, a method for guiding short-horizon policies toward satisfying complex LTL specifications at inference time using hierarchical world models. Our key idea is to derive two separate guidance objectives using each world model's abstraction level. A high-level model predicts future action-induced transitions in task-relevant atomic propositions to guide progress through the LTL automaton, while a low-level dynamics model predicts immediate state evolution for accurate local safety guidance. Our results show that hint$^2$ overcomes the limitations of current LTL-guided diffusion methods, outperforms existing inference-time steering methods in CALVIN, and successfully completes instructions with complex liveness and safety constraints more elegantly than language-conditioned alternatives. Finally, we demonstrate that hint$^2$ can handle complex instructions on a real UR5e manipulator.
Chinese Translation
机器人学习的核心目标之一是使机器人能够执行在运行时指定的丰富指令。大规模语言条件策略已经朝着这一目标取得了实质性进展，但仍然难以处理时序结构和安全约束。线性时序逻辑（LTL）提供了一种强大的语言来表达复杂的、非马尔可夫的指令。然而，引导已学习的操作策略满足 LTL 仍然具有挑战性，因为现代策略生成短视界动作块并在闭环中重新规划，而几乎所有的 LTL 规格都是基于长视界轨迹进行评估的。在本文中，我们提出了 hint$^2$，一种使用分层世界模型在推理时引导短视界策略满足复杂 LTL 规格的方法。我们的关键思想是利用每个世界模型的抽象级别推导出两个独立的引导目标。高层模型预测任务相关原子命题中由动作引起的未来转换，以引导通过 LTL 自动机的进度，而低层动力学模型预测即时状态演化，以提供准确的局部安全引导。我们的结果表明，hint$^2$ 克服了当前 LTL 引导扩散方法的局限性，在 CALVIN 中优于现有的推理时引导方法，并且比语言条件替代方法更优雅地完成了具有复杂活性和安全约束的指令。最后，我们证明了 hint$^2$ 能够在真实的 UR5e 机械臂上处理复杂指令。
cs.SE / 49 / 2608.13681
Fine-Tuning Qwen3-27B for C-to-Rust Code Translation: A Three-Stage Curriculum of Pretraining, Debugging-Aware SFT, and Task-Specific SFT
微调 Qwen3-27B 用于 C 到 Rust 代码翻译：预训练、调试感知 SFT 与任务特定 SFT 的三阶段课程
large language model
大语言模型相关
Abstract
Translating C code into safe, idiomatic Rust is a longstanding software-engineering goal because it can eliminate entire classes of memory-safety vulnerabilities while preserving the functional behavior of legacy systems. Large language models (LLMs) have shown promise for this task but typically underperform when applied off-the-shelf, since general-purpose pretraining rarely emphasizes idiomatic Rust generation, cross-language semantic equivalence, or the ability to reason about and repair compiler/runtime feedback. In this report we describe a three-stage fine-tuning curriculum applied to Qwen3-27B that is designed to progressively specialize the model for the C-to-Rust (C2Rust) translation task: (1) continued pretraining on Rust-centric corpora to strengthen the model's prior over idiomatic Rust syntax and standard-library usage; (2) supervised fine-tuning (SFT) on the microsoft/Verus_Training_Data dataset to instill debugging and self-repair behavior over Rust code; and (3) task-specific SFT on paired C/Rust solutions derived from LeetCode problems to teach direct semantic translation. We evaluate the resulting model using the agentic, static-analysis-guided verification framework of SACTOR, which performs structure-aware, two-phase (unidiomatic to idiomatic) translation with foreign-function-interface (FFI)-based end-to-end (E2E) testing. We report success rate, idiomaticity (Clippy lint counts, unsafe-code fraction), and failure-mode analyses, and compare our fine-tuned model against baseline Qwen3-27B and other LLMs evaluated under the same framework.
Chinese Translation
将 C 代码翻译成安全、惯用的 Rust 是一个长期存在的软件工程目标，因为它可以消除整类内存安全漏洞，同时保留遗留系统的功能行为。大型语言模型（LLM）已在此任务上展现出潜力，但直接使用现成模型时通常表现不佳，因为通用预训练很少强调惯用 Rust 代码生成、跨语言语义等价性，或对编译器/运行时反馈进行推理与修复的能力。在本报告中，我们描述了一种应用于 Qwen3-27B 的三阶段微调课程，旨在逐步使模型专门化于 C 到 Rust（C2Rust）翻译任务：（1）在以 Rust 为中心的语料库上继续预训练，以增强模型对惯用 Rust 语法和标准库使用的先验知识；（2）在 microsoft/Verus_Training_Data 数据集上进行监督微调（SFT），以灌输对 Rust 代码的调试和自我修复行为；（3）在源自 LeetCode 问题的配对 C/Rust 解决方案上进行任务特定 SFT，以教授直接语义翻译。我们使用 SACTOR 的智能体式、静态分析引导的验证框架评估所得模型，该框架执行结构感知、两阶段（非惯用到惯用）翻译，并采用基于外部函数接口（FFI）的端到端（E2E）测试。我们报告了成功率、惯用性（Clippy lint 计数、不安全代码占比）和失败模式分析，并将我们的微调模型与基线 Qwen3-27B 以及在相同框架下评估的其他 LLM 进行了比较。
cs.SE / 50 / 2608.14065
Rethinking Automated Program Repair: The Impact of Bug Complexity, Fault Localization, and LLM Cost-efficiency
重新思考自动化程序修复：缺陷复杂性、故障定位与LLM成本效率的影响
large language model
大语言模型相关
Abstract
Background: Software bugs remain a critical challenge in development, necessitating effective Automated Program Repair (APR) techniques. While Large Language Model (LLM)-based APR systems have shown promise, prior studies primarily focus on overall repair effectiveness. The effects of bug complexity, fault localization, reasoning settings, and repair cost-effectiveness remain insufficiently explored. Aims: This study presents a comprehensive empirical analysis of LLM-based APR, focusing on how repair performance is shaped by bug complexity, fault localization, reasoning settings, and costs. Method: We evaluate two APR techniques (ChatRepair and CodeCorrector) using three LLMs (DeepSeek, GPT, and Llama), and examine their performance across diverse levels of bug complexity and localization strategies through a multi-dimensional empirical framework and statistical analysis. Results: Although structurally complex bugs and imprecise fault localization make repair more challenging, LLM-based APR techniques still achieve competitive repair effectiveness. Imprecise fault localization can substantially enlarge the performance gap between APR techniques. Furthermore, higher-cost LLMs and stronger reasoning settings do not consistently yield better cost-efficiency, revealing a nontrivial trade-off between repair effectiveness and computational cost. Conclusions: Over 50% of moderately complex bugs can be repaired by low-cost LLM-based APR techniques. The repair effectiveness gap between APR techniques becomes larger as fault localization becomes less precise. GPT-5 repairs 7 and 39 more complex bugs than DeepSeek-V4-pro and DeepSeek-V3.2, respectively; whereas the total repair cost of DeepSeek-V3.2 shows the best cost-efficiency performance.
Chinese Translation
背景：软件缺陷仍然是开发中的一个关键挑战，需要有效的自动化程序修复（APR）技术。虽然基于大语言模型（LLM）的APR系统已显示出潜力，但先前的研究主要关注整体修复效果。缺陷复杂性、故障定位、推理设置和修复成本效益的影响仍未得到充分探索。目的：本研究对基于LLM的APR进行了全面的实证分析，重点关注缺陷复杂性、故障定位、推理设置和成本如何影响修复性能。方法：我们使用三种LLM（DeepSeek、GPT和Llama）评估了两种APR技术（ChatRepair和CodeCorrector），并通过多维实证框架和统计分析，考察了它们在不同缺陷复杂程度和定位策略下的性能。结果：尽管结构复杂的缺陷和不精确的故障定位使修复更具挑战性，但基于LLM的APR技术仍然取得了有竞争力的修复效果。不精确的故障定位可能显著扩大APR技术之间的性能差距。此外，成本更高的LLM和更强的推理设置并不总能带来更好的成本效率，这揭示了修复效果与计算成本之间非平凡的权衡。结论：超过50%的中等复杂度缺陷可以通过低成本的基于LLM的APR技术修复。随着故障定位精确度的降低，APR技术之间的修复效果差距变大。GPT-5比DeepSeek-V4-pro和DeepSeek-V3.2分别多修复了7个和39个更复杂的缺陷；而DeepSeek-V3.2的总修复成本表现出最佳的成本效率性能。
cs.SE / 51 / 2608.14128
DepWareTrans: Dependency-Aware Incremental Repository Migration across Co-executable Languages
DepWareTrans：跨可协同执行语言的依赖感知增量式仓库迁移
large language model
大语言模型相关
Abstract
Repository-level code translation is critical for modernizing legacy systems, yet existing approaches based on large language models (LLMs) operate at the file level and fail to scale to codebases with complex inter-file dependencies. This limitation is evident in our industrial setting, where we aim to migrate a production repository (STAR) from Java to Kotlin, but file-level approaches produce fragmented results and fail to achieve end-to-end correctness. In this paper, we show that the primary cause of failure at the repository level is dependency inconsistency. Through an empirical study on open-source and industrial systems, we find that most errors arise from unresolved cross-file dependencies that cannot be effectively addressed by iterative feedback alone. We propose a dependency-aware incremental migration framework that elevates the unit of translation from individual files to dependency-consistent batches. Our approach constructs a dependency graph, groups interdependent files, and performs batched translation with iterative compile- and test-driven validation. We evaluate our method on a 51K line of code (LOC) industrial system and multiple repositories across interoperable language pairs (Java-Kotlin, Java-Scala, and C#-F#). On the STAR repository, file-level approaches achieve 38.16% compilation and 9.39% test success, whereas our approach achieves 100% compilation and test success across the evaluated settings, converging within a small number of iterations. These results show that dependency-aware batching improves scalability and reliability in repository-level code translation.
Chinese Translation
仓库级代码翻译对于现代化遗留系统至关重要，然而现有基于大型语言模型（LLM）的方法在文件级别运行，难以扩展到具有复杂跨文件依赖关系的代码库。这一局限性在我们的工业场景中十分明显，在该场景中，我们旨在将一个生产仓库（STAR）从Java迁移到Kotlin，但文件级方法会产生零散的结果，且无法实现端到端的正确性。在本文中，我们表明仓库级别失败的主要原因是依赖不一致性。通过对开源系统和工业系统的实证研究，我们发现大多数错误源于未解决的跨文件依赖，而仅靠迭代反馈无法有效解决这些问题。我们提出了一种依赖感知的增量式迁移框架，将翻译单元从单个文件提升为依赖一致的批次。我们的方法构建依赖图，对相互依赖的文件进行分组，并执行带有迭代式编译和测试驱动验证的批量翻译。我们在一个51K行代码（LOC）的工业系统以及多个跨可互操作语言对（Java-Kotlin、Java-Scala和C#-F#）的仓库上评估了我们的方法。在STAR仓库上，文件级方法实现了38.16%的编译成功率和9.39%的测试成功率，而我们的方法在评估的所有设置中均实现了100%的编译和测试成功率，并在少量迭代内收敛。这些结果表明，依赖感知的批处理提高了仓库级代码翻译的可扩展性和可靠性。
cs.SE / 52 / 2608.14352
ATLAS: Discovering Agent Strategies through LLM-Guided Abstraction and Automata Learning
ATLAS：通过LLM引导的抽象与自动机学习发现智能体策略
large language model
大语言模型相关
Abstract
Large Language Model (LLM)-based agents are increasingly used for complex tasks such as software testing and cybersecurity assessment. While these agents demonstrate impressive capabilities, their behavior is difficult to understand, explain, and analyze. Existing evaluations focus mainly on task success and execution traces, offering limited insight into the strategies employed by the agent. We present ATLAS (Automata Learning for Agent Trajectory Analysis and Strategy Discovery), an approach for recovering interpretable behavioral models from agent trajectories. ATLAS combines trace abstraction with automata learning to infer finite-state models that capture observed agent-environment interaction strategies. These models provide human-interpretable insights and support automated analyses of recurring behaviors, decision points, successful task-completion paths, and failure loops. As a proof of concept, we apply ATLAS to trajectories generated by an LLM-based penetration-testing agent. The resulting models expose high-level behavioral strategies for exploiting vulnerable machines that are difficult to identify from raw execution traces alone. We discuss how learned behavioral models can support explainability, model-guided exploration, auditing, and analysis of agentic systems. We further demonstrate symbolic model-based knowledge transfer from powerful frontier models to compact language models. In addition, we show how model transformations can derive concise explanations of agent behavior in a penetration-testing case study comprising 12 vulnerable machines. ATLAS highlights a new opportunity for model-driven engineering: transforming agent trajectories into explicit behavioral models that enable systematic understanding and analysis of otherwise opaque AI agents.
Chinese Translation
基于大语言模型（LLM）的智能体越来越多地被用于复杂任务，例如软件测试和网络安全评估。尽管这些智能体展现出令人印象深刻的能力，但其行为难以理解、解释和分析。现有评估主要关注任务成功率和执行轨迹，对智能体所采用策略提供的洞察有限。我们提出ATLAS（面向智能体轨迹分析与策略发现的自动机学习），一种从智能体轨迹中恢复可解释行为模型的方法。ATLAS将轨迹抽象与自动机学习相结合，推断出能够捕获所观察到的智能体-环境交互策略的有限状态模型。这些模型提供人类可理解的洞察，并支持对重复行为、决策点、成功完成任务路径以及失败循环的自动化分析。作为概念验证，我们将ATLAS应用于由基于LLM的渗透测试智能体生成的轨迹。所得模型揭示了利用脆弱机器的高层行为策略，而这些策略仅从原始执行轨迹中难以识别。我们讨论了学习到的行为模型如何支持智能体系统的可解释性、模型引导的探索、审计与分析。我们进一步展示了从强大的前沿模型到紧凑语言模型的基于符号模型的知识迁移。此外，我们展示了在一个包含12台脆弱机器的渗透测试案例研究中，模型变换如何推导出对智能体行为的简明解释。ATLAS突显了模型驱动工程的一个新机会：将智能体轨迹转化为显式行为模型，从而能够系统性地理解和分析原本不透明的人工智能智能体。
cs.LG / 53 / 2608.14401
Offline Deep Q* Estimation with Diffusion Models
基于扩散模型的离线深度Q*估计
diffusion
扩散模型相关
Abstract
In offline RL, estimating the optimal action-value function $Q^*$ can be formulated as solving the optimal Bellman equation based solely on offline observations. A fundamental challenge is that the reward function and transition kernel are unknown, so the optimal Bellman operator is not directly observable from data. To address this issue, we propose a novel framework that decouples operator estimation from value function learning. In this approach, we first formulate conditional diffusion models to estimate the reward law and transition kernel, which induces a data-driven approximation of the optimal Bellman operator. We then plug these estimators into the Bellman equation and obtain a deep estimator of $Q^*$ by minimizing the empirical Bellman residual over a neural network function class. Theoretically, we first establish sharp nonasymptotic convergence rates for learning the optimal Bellman operator through an end-to-end analysis of conditional diffusion estimation in total variation distance. We then establish the oracle value-stage rate $\widetilde{\mathcal O}\bigl(n^{-\frac{2β}{d_x+d_a+2β}}\bigr)$ for the excess Bellman residual risk. Finally, under a concentrability condition, we translate this residual bound into an $L^2$ convergence rate of $\widetilde{\mathcal O}\bigl(n^{-\fracβ{d_x+d_a+2β}}\bigr)$ for the resulting deep estimator of $Q^*$, where $d_x$ and $d_a$ denote the dimensions of the state and action spaces, respectively, and $β$ denotes the Hölder smoothness index of $Q^*$. Importantly, our theoretical analysis does not rely on completeness assumptions commonly used in deep RL theory. Extensive numerical experiments demonstrate the effectiveness of the proposed method and its strong empirical performance.
Chinese Translation
在离线强化学习中，估计最优动作价值函数 $Q^*$ 可以表述为仅基于离线观测求解最优贝尔曼方程。一个基本挑战是奖励函数和转移核未知，因此最优贝尔曼算子无法直接从数据中观测到。为解决这一问题，我们提出了一种将算子估计与价值函数学习解耦的新框架。在该方法中，我们首先构造条件扩散模型来估计奖励规律和转移核，从而诱导出最优贝尔曼算子的数据驱动近似。然后我们将这些估计器代入贝尔曼方程，并通过在神经网络函数类上最小化经验贝尔曼残差来获得 $Q^*$ 的深度估计器。在理论上，我们首先通过条件扩散估计在总变差距离上的端到端分析，建立了学习最优贝尔曼算子的尖锐非渐近收敛速率。然后我们建立了超额贝尔曼残差风险的预言值阶段速率 $\widetilde{\mathcal O}\bigl(n^{-\frac{2β}{d_x+d_a+2β}}\bigr)$。最后，在可集中性条件下，我们将该残差界转化为所得 $Q^*$ 深度估计器的 $L^2$ 收敛速率 $\widetilde{\mathcal O}\bigl(n^{-\fracβ{d_x+d_a+2β}}\bigr)$，其中 $d_x$ 和 $d_a$ 分别表示状态和动作空间的维度，$β$ 表示 $Q^*$ 的 Hölder 光滑性指数。重要的是，我们的理论分析不依赖深度强化学习理论中常用的完备性假设。大量数值实验证明了所提方法的有效性及其强大的经验性能。
人工智能 (cs.AI)
86
cs.AI / 1 / 2608.14502
Universal Thermodynamic Interatomic Potentials for Crystalline Materials
Abstract
Free energies govern solid-state phase stability, yet computational materials discovery still relies largely on ground-state energies because free energy calculations require ensemble averages. We introduce the thermodynamic interatomic potential (TIP), which extends an interatomic potential from its static energy to a thermodynamically consistent Gibbs free energy model, with thermodynamic responses following from temperature and pressure by automatic differentiation. We implement TIP[UMA] using the universal potential UMA, train it on free energies from quasi-harmonic to molecular dynamics fidelity, and calibrate it to higher-resolution calculations or experiment. From a single evaluation, it returns the equation of state of a crystal and locates phase transitions among competing branches, including dynamically stabilized phases. Fine-tuning extends the model to alloy solubility limits and miscibility gaps. TIP makes the free energy as accessible as the potential energy, opening finite-temperature phase stability to high-throughput discovery.
cs.AI / 2 / 2608.13662
Ontology-Grounded Project Memory for Coding Agents
Abstract
Coding agents have become the primary means of generating new code in many software projects, and the resulting velocity of changes makes keeping track of the reasons behind those changes challenging. This paper introduces MOOSEDev, a system designed to give coding agents structured, ontology-grounded project memory. The system captures architectural decisions, lessons, constraints, and rationales in a knowledge graph exposed to agents via a Model Context Protocol (MCP) interface. Records carry lifecycle status, provenance, and supersession links, queryable via MOOSE, a proprietary neurosymbolic engine that treats the symbolic layer as the primary reasoning substrate. We compared MOOSEDev against a production vector-memory tool on a neutral public corpus of 835 typed records. MOOSEDev returned the expected answer set essentially in full (0.98-1.00) on supersession, set-completeness, and negation questions, whereas the baseline's top-k retrieval surfaced between 6% and 27%. Conversely, relevance recall and token cost were largely equivalent between the two systems. We also describe a temporal commit-history bootstrap of our own codebase, a pre-registered live trial, and lessons learned.
cs.AI / 3 / 2608.13667
Second Thought: Reasoning in Parallel as LLM Agents Act and Observe
Abstract
LLM agents in the ReAct paradigm alternate between reasoning, acting, and observing, but deliberate reasoning is confined to the Thought phase: while the agent serializes an action and waits for the environment, its reasoning is frozen. We identify this recurring interval for Action and Observation as a reasoning idle window and ask whether it can host additional reasoning in parallel that serves future turns. Therefore, we propose Second Thought, a training-free inference framework that forks four auxiliary branches the instant each Thought phase concludes, decodes them concurrently with the main loop, and merges the generated thoughts back when the environment observation arrives. In this way, Second Thought relocates the added reasoning off the main thread's sequential decoding path. Across three agentic benchmarks and three reasoning LLMs, Second Thought lowers the average turn count in all nine (model,benchmark) pairs and reduces main thread decoding in six of them by up to 43% (roughly 20% on average among those settings), while leaving it essentially unchanged in a seventh; Pass@1 shows no significant change in seven of nine pairs and the two significant differences are +12.4 and +10.2 points. Against a compute-matched control that forces an equivalent budget onto the main thread's own reasoning, it attains strictly higher Pass@1 with 1.3 to 3.2 less sequential decoding in all four settings where the control applies.
cs.AI / 4 / 2608.13684
Learning to Assemble Novel Structures with Unfamiliar Parts under Semantic Constraints
Abstract
This paper describes a neurosymbolic architecture for learning to assemble novel structures using evidence from embodied conversations and task demonstrations. We focus on scenarios where an agent encounters, after deployment, semantic constraints on structures--in other words, constraints as to which part types and features make valid structures--that were not available during training, and where it is initially unaware of the relevant structure and component part concepts. The agent must acquire and exploit such knowledge through user interactions while attempting assembly. We study this setting in a simulated toy truck assembly domain, learning from symbolic evidence encoded in natural language and from dense visual observations. Our experiments show that communicating semantic constraints through natural language (e.g., "dump trucks have a dumper") yields more data-efficient online adaptation than relying only on task demonstrations and/or only naming the parts through natural language.
cs.AI / 5 / 2608.13719
Coverage Aware Active Evaluation for Failure Discovery with Paired Systems
Abstract
Autonomous systems can fail in rare and heterogeneous ways, making real-world failure discovery difficult under limited testing budgets. Although cheaper proxies such as simulators, lower-fidelity systems, or related policies can be sampled extensively to find failures, proxy failures often do not transfer to the real world due to sim-to-real and system-to-system gaps. The key challenge is therefore to effectively leverage proxy system information for accurate prediction of severe target system failures. We propose an adaptive failure discovery method that combines proxy evaluations with limited target system results to guide scenario selection for target system testing. Our method learns a local predictor of target risk by correcting proxy failure signals using control-variate-inspired residual modeling. To find failures that are both likely and diverse, we combine this predictor with a support-aware mutual-information objective that favors realistic, well-supported regions while expanding coverage across failure modes. Across autonomous driving, manipulation, and quadruped velocity-tracking tasks, our method discovers up to 2$\times$ as many failures as random sampling and active-learning baselines, including severe and diverse failures missed by competing methods.
cs.AI / 6 / 2608.13754
Explanation Multiplicity: Circuit-Level Interpretability Evidence Does Not Survive Defensible Analytic Variation
Abstract
The EU AI Act requires providers of high-risk systems to file technical documentation describing how the system reaches its decisions. Mechanistic interpretability is the obvious source of such evidence, and circuit discovery is its most developed instrument. We ask whether that evidence survives the condition under which it would be relied upon: two competent analysts, the same system, the same tool, different defensible settings. We pre-registered a crossed grid of seven analytic axes, every level taken from a published implementation, and mapped each discovered circuit through a deterministic claim map to a structured Annex IV statement. Across 15,840 pre-registered specifications on GPT-2 small and the indirect object identification task, of which 7,561 produced a claim, the derived statement flips across 73.2% of specification pairs (95% CI 0.725 to 0.738) and the modal claim commands 41.1% of the space. The evidence fails a filability criterion at every tolerance a conformity assessment body would plausibly accept. Standardising the single most influential choice, the evaluation metric, leaves the flip rate at 59.4%. Removing circuit size from the claim entirely and holding it fixed leaves 27.1% (95% CI 0.255 to 0.286), still above the pre-registered threshold. The circuits underlying these claims are structurally near-disjoint, median pairwise Jaccard overlap 4%, and functionally uncorrelated at Cohen's kappa 0.015, so the instability is not one mechanism described in different words. We give the filability criterion as a standalone protocol, and we report that one of the seven documented discovery objectives does not execute at all on the library's own canonical task. The study covers one model and one task, and whether the conclusion holds at scale is untested.
cs.AI / 7 / 2608.13774
FLARE MCMC: Fidelity-based Layer-Adaptive REcursive proposals for MCMC
Abstract
Markov chain Monte Carlo (MCMC) requires only the ability to evaluate the likelihood, making it a common technique for inference in complex models. However, it can have a slow mixing rate, requiring the generation of many samples to obtain good estimates and an overall high computational cost. FLARE MCMC is a multi-fidelity layered MCMC method that exploits lower-fidelity approximations of the true likelihood calculation to improve mixing and leads to overall faster performance. Such lower-fidelity likelihoods are commonly available in scientific and engineering applications where the model involves a simulation whose resolution or accuracy can be tuned. Our technique uses recursive, layered chains with simple layer tuning; it does not require the likelihood to take any form or have any particular internal mathematical structure. We demonstrate experimentally that FLARE MCMC achieves larger effective sample sizes for the same computational time across different scientific domains including hydrology and cosmology.
cs.AI / 8 / 2608.13787
From Passive Delegates to Strategic Negotiators: Reinforcing Social Reasoning in Small Language Models with SocialRL
Abstract
AI agents increasingly act on their users' behalf, handling tasks such as scheduling meetings, comparing offers, and haggling over prices. These principal-driven tasks routinely place the agent across from a counterpart (another user's agent, a seller, a recruiter) whose goals may conflict with its principal's. Yet the dispositions that make an assistant pleasant can make it a poor delegate: a friendly, helpful frontier model may disclose its principal's private information unprompted and concede at the first sign of resistance. We present SocialRL, a general recipe that trains social reasoning directly, and apply it to a 4B model across six domains: Deal-or-No-Deal, CaSiNo, Craigslist, Job Interview, Calendar, and Marketplace. Every domain is trained in-domain under the same recipe, and every policy is evaluated on all six. We find that (1) in-domain training reaches the frontier: on held-out scenarios the 4B matches or exceeds the GPT-5 family per domain, closing 73-122% of the baseline-to-frontier gap on the negotiation games, with 78% of buyer openings anchoring below target versus 3% untrained; (2) cross-domain transfer follows game structure: structurally paired games lift each other, a broad multi-issue donor lifts nearly all domains, and structurally isolated games transfer nothing; (3) guided by this transfer structure, two strategies, cascade RL and multi-teacher on-policy distillation (OPD), consolidate the per-domain specialists into a single unified 4B that reaches 0.627 average utility across all six environments, matching or exceeding GPT-4.1 (0.625), GPT-5.1 (0.619), and GPT-5.2 (0.613); (4) an explicit theory-of-mind scaffold helps only through training: distilling the ToM trace, rather than actions alone, lifts utility on every environment and generalizes better across them, and of the two ToM skills, only next-action prediction predicts negotiation outcomes.
cs.AI / 9 / 2608.13863
Joint Optimization of Memory and Computing Frequency for Energy-Efficient DNN Inference
Abstract
Deep neural network (DNN) inference on mobile devices often incurs high latency and energy consumption due to limited computing and memory resources. To enable energy-efficient DNN inference, most existing studies focus on dynamic voltage and frequency scaling (DVFS) for adjusting the computing frequency, while the impact of memory frequency on the inference performance has been greatly overlooked. In this paper, we consider the impact of memory frequency and computing frequency on DNN inference time, and jointly optimize these two frequencies together with communication resources for energy-efficient DNN inference. Based on a realistic inference time model, we formulate an optimization problem to minimize the energy consumption of all mobile devices under the deadline constraint. For local inference, we derive a near-optimal closed-form solution via convex optimization, while an optimal closed-form solution for transmission power is obtained for edge inference with the given bandwidth. Furthermore, we propose a low-complexity heuristic algorithm to effectively solve the overall problem with polynomial time complexity. Simulation results based on measured data show that the proposed near-optimal solution for local inference can achieve optimal performance under strict deadline constraints, with a performance gap of up to 2.5% compared with the optimal solution. Meanwhile, our proposed algorithm significantly reduces the energy consumption of devices by up to 10.4% compared to other methods.
cs.AI / 10 / 2608.13883
MemoryLake on MemoryArena: A Matched Study of Agent Memory Backends
Abstract
Most agent-memory benchmarks test post-hoc recall, whereas MemoryArena evaluates whether memory supports interdependent, multi-session task completion. We compare MemoryLake, a structured multi-track memory backend, with Mem0, text-embedding-3-small vector RAG, and a long-context control across all five MemoryArena domains. The systems share the same agent framework, requested gpt-5-mini model alias, task samples, and scoring code; the memory integration is the intentionally changed component. Because each backend bundles write, retrieval, consolidation, budgeting, and prompt-assembly choices, the study is a matched system-level comparison, not a representation-only ablation or a cost-matched experiment. On the shared evaluation sets, MemoryLake has the highest observed success rate (SR) in mathematics (9/40), physics (12/20), and progressive retrieval (4/20). Every system has zero SR in travel planning, and web shopping yields a single bundle-level success (long context, 1/150); MemoryLake ranks third on both the travel soft process score and shopping step match. Following MemoryArena's suite-level convention, a post-hoc equal-weight average over the five SRs is 20.5% for MemoryLake versus 13.6% for the best comparator. These are point estimates: sample sizes are modest, confidence intervals overlap, and we do not report paired significance tests. A separate MemoryLake-only run over all 221 progressive queries yields a failure-counted SR of 26.7% (59/221) and is not a baseline comparison. The results support a workload-dependent view of memory backends and an observed lead among the four evaluated systems on the shared sets; they do not establish benchmark-wide state of the art or a causal advantage of representation structure.
cs.AI / 11 / 2608.13921
When Personal Memory Has No Single Answer: Evaluating LLM Agents under Irreducible Conflict
Abstract
LLM agents increasingly maintain personal memory across sessions, but it can conflict. Preferences depend on context, behavior evolves, and sources can conflict. When a query lacks context, time, or source authority to interpret conflict, treating one memory as definitive converts unresolved conflict into an unjustified, overconfident action. Existing benchmarks recover one answer from conflicting evidence, overlooking whether agents recognize underdetermination, preserve alternatives, seek missing information, and choose appropriate actions. We introduce \underline{T}esting \underline{A}gents' \underline{N}avigation of \underline{G}enuine, \underline{L}atent, and \underline{E}ntangled Memory Conflicts (\textsc{TANGLE}), a benchmark for genuinely unresolvable memory conflicts. It comprises 541 instances across 40 personas and three types: Context-Partitioned Conflict (CPC), Behavior-Oscillation Conflict (BOC), and Source-Contradiction Conflict (SCC). We evaluate two tracks---an oracle track with curated memory and a pipeline track that extracts memory from multi-session dialogues---on five dimensions: conflict perception, causal reasoning, confidence calibration, clarification seeking, and memory faithfulness. Experiments reveal pipeline challenges. With curated memory, models recognize conflicts more reliably than they calibrate actions or seek targeted clarification. With end-to-end pipeline memory, extraction fails to preserve conflict-bearing relations needed for downstream reasoning. Policy comparisons show fixed rules are insufficient when actions must reflect conflict. These findings motivate Conflict-Aware Action Policy (CAAP), which adapts actions to each conflict using available evidence. \textsc{TANGLE} frames conflict handling as recognizing underdetermination, retaining conflicting evidence, and acting without forcing a definitive answer.
cs.AI / 12 / 2608.13940
AI Research Preference Models
Abstract
AI research agents (AIRA) can now propose, implement, and evaluate their own machine learning experiments, but progress on frontier tasks is throttled by cost: a candidate solution can be written in minutes, whereas evaluating it can take hours to days of GPU time. An agent can therefore propose far more candidates than it can afford to run, and its progress depends on its research preference: how it allocates a fixed execution budget across many candidates. We introduce AI Research Preference Models (RPMs) that predict which of multiple candidate solutions are most worth executing, without paying the cost of executing them all. We build RPMs from frozen pretrained language models (with no task-specific training), in two forms: an inference-only model that reasons over candidate plans, code, and prior executed solutions, and an agentic model that additionally runs small-scale pilot experiments before deciding. We integrate both into the AIRA-dojo search agent and evaluate on AIRS-Bench, a recent benchmark of machine learning research tasks for AI research agents. The two variants raise the average normalized score from 0.684 to 0.711 and 0.729 respectively, and reach the unguided agent's 24-hour performance in roughly 15 hours, using less than two-thirds of its execution budget. Our best RPMs also yield new state-of-the-art results on two AIRS-Bench tasks.
cs.AI / 13 / 2608.13951
HELIX: Model-Harness Co-evolution for Recursive Self-Improvement
Abstract
Scaling agent capability has largely focused on improving the model, yet an interactive agent acts through a runtime harness that mediates context, tools, control flow, and stopping. The harness shapes both what a model can accomplish and the trajectories from which it learns. This coupling motivates model-harness co-evolution for recursive self-improvement: build harnesses for a fixed model, update the model from verified sibling trajectories, and rebuild the harnesses as model capabilities change. Realizing this loop requires a controlled way to evolve harnesses while preserving intervention identity and effect. We present HELIX, a source-traceable substrate for harness evolution. HELIX decomposes agent systems into typed ports, reusable atoms, recipes, product shells, and runtime policies. It makes interventions explicit and auditable while retaining trajectories, test outcomes, and provenance. Harness evolution thus serves two linked roles: improving fixed-model execution and producing matched successes, regressions, near misses, and alternative solutions as data for subsequent model improvement. We evaluate HELIX in one evolution round on code repair. A 65-candidate portfolio discovers a fixed harness that improves task coverage by 4.0% over Pi, while the full portfolio exposes up to 58.0% more verified coverage through complementary sibling behavior. Selected candidates are assessed with repeated runs and the SWE-bench evaluator. A 200-slot sibling slice yields 438 verified SFT, critic, filter, and preference records. These results show how harness, model, and data form a feedback system: harness evolution expands current capability and creates learning signal for the next model; model updates motivate the next round of harness evolution. HELIX provides an auditable interface for studying this recursive process. Code is available at https://github.com/HKUDS/HELIX.
cs.AI / 14 / 2608.13958
Implementing Computational Law in Wolfram Language for the Governance of Artificial Intelligence
Abstract
How do we govern AI systems whose reasoning we cannot fully inspect? Governance does not require understanding a system's reasoning. It requires stating what the system is obliged, permitted, and forbidden to do, and checking whether it complied. I present an implementation of Reified Input/Output Logic, the formalism behind the DAPRECO knowledge base, in Wolfram Language: the core I/O axioms, obligations, permissions, constitutive norms, reified eventualities, and temporal operators. I then test whether GPT-4 can translate English legal statements into the formalism, and report the failures: hallucinated functions, omitted temporal scope, deviation from the formalism, and (in the worst cases) code that runs, reads plausibly, but silently encodes the wrong norm. A case study, an AI guard dog operating under a computational contract, shows how formalized rules can extend from a contract directly into the operational code of an embodied agent, producing symbolic, auditable justifications for its behaviour. I argue that computational law can be used as a governance tool and that a desirable goal would be to formalize the law that can and ought to be programmatically executable.
cs.AI / 15 / 2608.13987
Nanbeige4.2-3B on Apple Silicon: Fixing Deployment Bugs and Decreasing Looped Transformer Memory Overhead
Abstract
Nanbeige4.2-3B is a 3B-parameter agentic model built around a Looped Transformer (LT) that reuses one stack of layers for a second forward pass, adding effective depth without additional parameters. Evaluated on Apple Silicon (MPS), we identify five independent bugs which prevent the released checkpoint from running via Hugging Face transformers out of the box (including a silently-zeroed RoPE buffer and calls to removed transformers cache APIs). Furthermore, we show that fixing these bugs is still not sufficient for agentic tasks, due to the LT's layer-reuse strategy (which effectively doubles peak attention memory) used to achieve parameter efficiency. We thus introduce a chunked-prefill strategy which alleviates the incurred memory-capacity penalty, extending allowable context width by $2.7 \times$ on 32~GiB shared memory. However, even with the reduced memory overhead, we show that patches are required to render Nanbeige4.2-3B usable; resolving both system prompt and MPS-native memory bugs finally allows reliable evaluation on standard MCP and tool-calling benchmarks. On a subset of MCPMark, the debugged model completes up to 30\% of real agentic tasks (up from the original's 0\%), while, on BFCL, it is near-perfect at single tool calls (yet fails the majority of multi-tool tests). We release the patched checkpoint, system prompt optimizer, and evaluation harnesses at https://github.com/johnhalloran321/Nanbeige4.2-3B-mps-fix.
cs.AI / 16 / 2608.13990
Content Depth Matters in Short-Video Recommendation: Rethinking the Attention Economy
Abstract
Driven by the attention economy, short-video Recommender Systems (RSs) are primarily optimized to maximize user engagement by promoting videos that capture attention within seconds. These systems inherently favor shallow-content videos that are effective at attracting immediate attention. However, growing evidence suggests that prolonged exposure to such content may negatively affect users' cognitive engagement and mental well-being, raising concerns about the long-term societal impact of the short-video platform. To tackle this challenge, this paper introduces a new metric, the \textbf{Content Depth Score (CDS)}, to quantify the content depth of short videos. CDS measures the extent to which a video is expected to stimulate higher-order cognitive processes, using a seven-level scale grounded in established theories of cognitive psychology and learning. As an initial step toward this vision, we present \textbf{SCOPE-Bench}, the first benchmark for content-depth evaluation in short-video recommendation. Built upon a large-scale open-source short-video dataset, SCOPE-Bench provides CDS annotations for 150K videos, enabling systematic evaluation of RSs from a cognitive-content perspective. Leveraging SCOPE-Bench, we evaluate 13 representative RSs and reveal a consistent preference for shallow-content videos. Moreover, we find that these algorithms recommending cognitively deep content are only marginally better than random selection, highlighting a previously overlooked limitation of existing recommendation objectives. Our code and datasets are available at https://liweidengdavid.github.io/SCOPE-Bench/.
cs.AI / 17 / 2608.13993
Simulation-Driven Vehicular Traffic Data Augmentation: Extending Sensor Coverage Through Virtual Sensing
Abstract
Urban traffic management relies on sensor networks whose spatial coverage is limited by deployment costs and privacy regulations. Machine learning models trained on such sparse data cannot generalize to unmonitored locations and must be retrained whenever the sensor infrastructure changes. We propose a simulation-based methodology that addresses this problem by generating augmented traffic count datasets in which each physical sensor is replaced by a virtual sensor placed at a surrogate location in the road network. Virtual sensors are selected by a graph-search heuristic that jointly maximises vehicle-flow continuity and traffic-metric similarity between the original and surrogate locations, while enforcing a minimum spatial displacement to ensure diversity of observed traffic conditions. We validate the method on two Belgian cities: Brussels, using a calibrated model, and Namur, using synthetic models. The augmented datasets preserve the bimodal daily demand profile and the dynamics of traffic at the observed locations.
cs.AI / 18 / 2608.14021
Residual Dominance as a Structural Account of Last-Item Reliance in Causal Self-Attention Recommenders
Abstract
Transformer-based sequential recommenders with causal self-attention often rely heavily on the most recent interaction at inference time, but how this behavior is structurally expressed in the representation used for prediction remains unclear. We combine prediction-time diagnostics with norm-based analysis of the full attention block. First, we show that SASRec-style models exhibit highly localized last-item reliance. We then find that, although self-attention aggregates contextual information, residual addition sharply shifts the full-block representation toward same-position contributions, which we term residual dominance. To probe this interpretation, we use inference-time residual scaling as a controlled diagnostic intervention. Changing the residual strength induces a monotonic trade-off between structural mixing and last-item reliance, while reducing residual strength recovers a subset of final-position misses for which representations at non-final positions already rank the ground-truth item correctly. Our results provide a structural account linking extreme last-item reliance to residual dominance at inference time. The code is publicly available.
cs.AI / 19 / 2608.14036
Demystifying Agent Skills: Why They Work-Until They Don't
Abstract
Skills have emerged as a practical and effective approach for enhancing LLM agents at inference time through structured packages of knowledge. However, existing evaluations largely measure whether skills improve aggregated task success, leaving a more fundamental question underexplored: \emph{\textbf{When do skills help, why do they work, and where do they fail?}} Through controlled experiments across various benchmarks, agent harnesses and LLMs, we isolate the effects of representation, outcome annotation, retrieval difficulty, and cross-framework robustness of skills. To further answer this question, we design a contrastive study that combines controlled quantitative experiments with paired trajectory analysis. We normalize 8,135 trial records from controlled experiments and retain 238 valid unique labels from 240 open-coded records. We consolidate these observations into a taxonomy of three high-level categories and twelve skill-use modes: skills work when noisy trajectories become procedural anchors that stabilize execution. Skills improve over Workflow Memory by 6.06 points in matched comparisons. Procedural anchoring accounts for 65.7\% of skill cases, versus 4.5\% for explicit knowledge injection, showing that skills stabilize action rather than inject missing facts. Retrieval is a separate bottleneck: as pools grow from 5 to 100, actual-use precision falls from 29.6\% to 3.3\%. Confusable distractors impair offline identification, yet downstream success remains stable; exact ground-truth invocation is neither sufficient nor necessary. Skills fail under brittle assumptions, incompatible contexts, or insufficient adaptation. These findings move evaluation beyond aggregate success rates and guide reliable self-evolving agents.
cs.AI / 20 / 2608.14063
Benchmarking data-driven material models on the classic Treloar dataset
Abstract
Machine learning is rapidly reshaping constitutive modeling, offers new ways to learn material behavior directly from experimental data, and challenges long-established modeling paradigms. But with a growing number of machine-learning-based approaches available, how do they compare in practice? In this paper, we use the classic experimental data of Treloar to benchmark popular frameworks for hyperelasticity: (Generalized-Invariant) Constitutive Artificial Neural Networks, Physics-Augmented Neural Networks, (Adaptive) Material Fingerprinting, and Efficient Unsupervised Constitutive Law Identification & Discovery. We compare their fitting performance, computational cost, hyperparameter sensitivity, and ease of implementation. Furthermore, we discuss the trade-offs between predictive accuracy and model complexity. The latter is assessed by quantifying both the number of material parameters in the discovered models and the computational time required to evaluate the constitutive model and its derivatives. The results show that all methods can reproduce the benchmark data remarkably well. Rather than identifying a single winner, we highlight the strengths and limitations of each approach and provide practical guidance for their use. The source code for all six methods, including the training and comparison scripts, as well as all results and data used in this study, is publicly available via https://doi.org/10.5281/zenodo.21915635.
cs.AI / 21 / 2608.14074
Mandato: Protocol-Level Enforcement of Digitally Signed Mandates on AI Agent Actions with Cryptographically Chained Audit Trails
Abstract
AI agents increasingly act on external systems through standardized tool-calling protocols such as the Model Context Protocol (MCP), yet no infrastructure layer constrains their actions to what a principal has verifiably authorized: authorization logic lives in application code, is neither signed nor independently auditable, and the resulting logs lack evidentiary value. We present Mandato, a governance proxy that enforces digitally signed mandates on agent actions at the protocol level. A mandate is a machine-readable, cryptographically signed authorization artifact specifying which tools an agent may invoke, under which parameter constraints and contextual conditions, for how long, and on whose behalf; the proxy evaluates every tool call against the applicable mandate chain, blocks non-conforming calls in line, and records every decision -- permit, deny, and the evidence for each -- in an append-only, hash-chained audit log designed for evidentiary use and periodically anchored via qualified timestamps. The mandate is deliberately modeled on the civil-law institution of delegation of authority, making the artifact legible to lawyers and auditors, not only to engineers. We give the mandate model and its decision semantics, the reference architecture as an MCP-transparent proxy with separated decision and enforcement points, and a mapping of the mechanism onto EU AI Act Articles 12 and 14, GDPR accountability, NIS2, and eIDAS 2, including a roadmap to qualified attestation through Qualified Trust Service Providers (QTSPs). We describe the implementation status of the reference system and a quantitative evaluation plan covering enforcement overhead, audit completeness, and tamper-evidence verification cost.
cs.AI / 22 / 2608.14075
A Pathway to General-Purpose Scientific AI: Multimodal Comprehension of Scientific Images
Abstract
Scientific figures and tables encode essential experimental evidence, yet remain difficult for digital libraries and multimodal AI systems to retrieve and interpret. The ALD/E-ImageMiner benchmark and ICDAR 2026 Competition on Information Extraction from Atomic Layer Deposition/Etching Scientific Figures provide 1,951 figures from 205 publications, expert-annotated for classification, data table extraction, summarization, and visual question answering. In these companion proceedings, we present a forward-looking perspective on how the benchmark can guide future scientific-image challenges. We examine how its tasks probe capabilities from visual and quantitative reading to domain-grounded reasoning and evidential justification, and how Bloom-informed question design can support deeper scientific understanding. We propose "scientific conceptual understanding from images" as a long-term benchmark objective, with future directions including broader domains and figure types, contextual and cross-document synthesis, hypothesis evaluation, provenance, uncertainty, counterfactual grounding, and open-ended multimodal research. This perspective connects the ICDAR 2026 challenge to a broader agenda for machine-actionable scientific visual knowledge and verifiable multimodal scientific AI.
cs.AI / 23 / 2608.14107
Retrieval Grounding Latent Reasoning for Dense Retrieval
Abstract
Reasoning-intensive retrieval requires text representations to capture not only semantic similarity, but also the reasoning needed to determine relevance under a given retrieval instruction. Existing reasoning-enhanced embedding models improve retrieval by incorporating reasoning information into dense representations, yet their supervision is typically dominated by the final retrieval objective. As a result, latent reasoning trajectories may learn shortcut reasoning patterns that preserve retrieval performance without producing meaningful incremental retrieval gains. We propose Retrieval Grounding Latent Reasoning (RGLT), a latent reasoning framework for dense retrieval that explicitly connects intermediate latent transitions with retrieval improvements. RGLT performs non-autoregressive reasoning in hidden space through an instruction-conditioned latent reasoning trajectory constructed from silent tokens. It combines process-supervised explicit-to-implicit distillation with retrieval-grounded supervision, using stage-wise CoT reconstruction to shape intermediate latent states and retrieval-effect credit to optimize incremental retrieval gains across the latent reasoning trajectories. Experiments on reasoning-intensive retrieval benchmarks show that RGLT consistently outperforms strong baselines while preserving efficient embedding inference.
cs.AI / 24 / 2608.14109
A Graph-Based Reinforcement Learning Framework for Structured Drift Diagnosis and Recovery in Autonomous LLM Agents
Abstract
Autonomous LLM agents are increasingly deployed in complex real-world workflows, yet they remain vulnerable to runtime behavioral drift, a silent deviation from the original task that can lead to irreversible side effects on external systems. Existing approaches address drift at the prompt level but lack structured mechanisms for step-level detection, risk assessment, and recovery decision. Because the main task-executing agent is often a large and expensive model that cannot be re-trained on every deployment, this work targets a plug-and-play recovery module instead. It introduces a graph-based framework in which a single small language model is trained via reinforcement learning to specialize at each node of a recovery graph, external to the main agent. Each node has a precise role\,: drift classification, operation detection, risk evaluation, or final decision and the model learns to produce structured XML-formatted reasoning adapted to that role. Training combines rule-based structural rewards with an LLM-as-judge semantic-quality signal, so that the model is graded both on how it answers (schema and length) and on what it says. Experiments on the public AppWorld benchmark show that the method generally exploits information about the suspected drift onset to issue correct recovery decisions using a small language model. In addition, the trained small language model reliably respects the prescribed output schema and produces semantically appropriate content in each field according to its assigned node role.
cs.AI / 25 / 2608.14122
Reinforcement Learning-Based Production Scheduling in an Industry-Based Coating Scenario Using the Digital Model Playground
Abstract
Production scheduling in complex manufacturing environments is challenging when sequence-dependent setup times, stochastic disturbances, and due-date constraints must be addressed simultaneously. While reinforcement learning (RL) methods have shown promising results in research, most studies rely on simplified benchmark processes, limiting their industrial relevance. This paper demonstrates the applicability of RL-based scheduling in an industry-inspired coating process that reflects practical complexities such as sequence-dependent setup times, machine breakdowns, and variable utilization. The open-source Digital Model Playground (DMPG), a discrete event simulation framework, is used to model the scenario and to train RL agents. Two standard algorithms, Deep Q-Networks and Proximal Policy Optimization, are benchmarked against conventional dispatching rules to illustrate feasibility and to provide a transparent testbed for further research. Results indicate that RL-based scheduling achieves balanced improvements across key performance indicators, with PPO delivering the most robust performance. The main contribution of this work is to bridge the gap between academic research and industrial practice by validating RL-based scheduling in a realistic, shareable scenario and by providing a reusable open-source framework for future studies.
cs.AI / 26 / 2608.14125
Traj-LeWM: Path-Aware World-Model Planning via Latent Trajectory Cost
Abstract
LeWM is a lightweight visual world model that learns latent dynamics end-to-end from pixels and ranks candidate action sequences by the distance between their predicted endpoints and the goal. However, LeWM has two limitations. First, during training, it learns local next-step transitions without evaluating complete trajectories relative to the task goal. Second, during planning, it ranks candidates solely by predicted endpoint distance. Because model predictions may differ from actual execution outcomes, the candidate whose predicted endpoint is closest to the goal may not perform best when executed in the environment. The evolution of the complete predicted trajectory can therefore provide complementary information beyond endpoint distance. To address these limitations, we propose Traj-LeWM, which retains LeWM's local-dynamics objective and endpoint score while introducing a goal-conditioned latent trajectory cost (LTC) that aggregates trajectory-level information as a complementary signal. During training, LTC-based trajectory-preference supervision complements next-step prediction in shaping the shared representation. During planning, LTC is combined with endpoint distance to incorporate intermediate-path information into candidate ranking. With joint endpoint-plus-LTC scoring, Traj-LeWM outperforms LeWM on Push-T, OGBench-Cube, Reacher, and Two-Room by $3$, $14$, $7$, and $7$ percentage points, respectively. Controlled experiments and ablations further verify the complementary roles of trajectory-level representation shaping and path-aware candidate ranking.
cs.AI / 27 / 2608.14149
QuaSAR: Quantization Compensation via Stable Activation-Aware Rank Truncation
Abstract
Recent training-free post-training quantization methods restore model accuracy through closed-form residual compensation. To constrain additional model storage overhead, several existing methods gate layer selection by goodness-of-fit, retaining only those layers whose compensation yields a positive residual fit score and discarding the rest. In this paper, we show that, under the low-bit W4A4 setting, this gating mechanism fails to distinguish poorly predictable quantization error from numerical solver failure. Rank-deficient input activations yield severely ill-conditioned or numerically singular Gram matrices, causing the closed-form solver to become unstable and produce spuriously negative fit scores. Consequently, existing goodness-of-fit gates misclassify affected layers as uncompensable and discard them. Many of these discarded layers can nevertheless provide substantial error recovery when their compensation is computed using a numerically stable solver. To address this problem, we propose a parameter-free truncated pseudoinverse solver which removes collapsed directions prior to inversion. On ViT-B with the W4A4 setting, our training-free method achieves 81.42\% top-1 accuracy, outperforming prior post-training methods and fine-tuning-based baselines. Combined with joint low-rank and quantization compression, the proposed method reaches a deployable operating point of 80.26\% accuracy at 54.7 MB, providing a well-balanced trade-off between model size and accuracy.
cs.AI / 28 / 2608.14157
Removing Temporal Note Redundancy Improves Multimodal Reinforcement Learning for Medicine
Abstract
Mechanical ventilation is a critical life-support intervention, requiring dynamic adjustments to ventilator settings as a patient's condition evolves. While reinforcement learning (RL) offers a promising framework for optimizing these sequential decisions, standard approaches rely primarily on structured electronic health record (EHR) data, missing crucial clinical context recorded in free-text notes. Integrating longitudinal clinical notes into RL state spaces is challenging because notes are heavily inflated by temporal redundancy, such as copy-forward text, templating, and repetitive documentation, which dilutes time-local updates and degrades state representation quality. To address this, we propose a redundancy-aware multimodal state representation framework that explicitly removes duplicated note text over time before policy learning. We evaluate two computationally efficient temporal decomposition strategies for removing duplicated note text: (1) an embedding-space decomposition using singular value decomposition on local history subspaces, and (2) an interpretable sentence-level diff operation that filters out previously documented sentences before text encoding. Using real-world ICU data, we demonstrate that state representations constructed by stripping temporal note redundancy significantly outperform both structured-only and raw-note baselines across multiple off-policy evaluation methods (Model-Based Rollouts, Fitted Q-Evaluation, Weighted Importance Sampling, and Weighted Doubly Robust Evaluation). Our findings show that explicitly isolating new clinical information from repeated note text yields higher-quality state representations and directly improves RL performance for clinical decision support.
cs.AI / 29 / 2608.14161
BiasTrace: Linking Reasoning Behaviours to Biased Outputs in LLMs
Abstract
LLMs exhibit social biases that can produce inaccurate and discriminatory inferences, posing risks in high-stakes applications. While prior work has made progress in measuring and mitigating bias, it largely focuses on final outputs of models, with limited understanding of the mechanisms that produce biased outcomes. Recent advances in LLM reasoning offers a new lens for investigating bias, yet the link between reasoning and bias remains poorly understood. Existing approaches focus primarily on final answer correctness or explicitly biased language, overlooking different behaviours in reasoning that can drive biased outcomes. We introduce BiasTrace, an annotation scheme for labelling reasoning behaviours in model-generated traces and linking them to biased outcomes. BiasTrace captures bias-specific behaviours (e.g., unsupported demographic assumptions) as well as general reasoning patterns that may implicitly contribute to bias (e.g. overthinking). We apply BiasTrace to reasoning traces in bias-sensitive contexts, scaled using validated LLM-as-a-judge methods, producing a large annotated dataset. Our analysis shows that biased outputs often stem from subtle reasoning behaviours rather than explicitly biased language, and that reasoning-level annotations improve bias detection. We further show that BiasTrace behaviours can be exploited for inference-time mitigation. These findings underscore the importance of examining a broader range of reasoning patterns to better understand bias in LLMs.
cs.AI / 30 / 2608.14205
FreeBalance: Pre-Routing Online Moe Load Balancing via Residual Workload Prediction
Abstract
Load imbalance poses a major bottleneck to the efficiency of expert parallelism in distributed inference of Mixture-of-Experts (MoE) models. The most heavily loaded rank stalls global execution due to skewed routing distributions, directly increasing latency. While offline expert placement can alleviate persistent imbalance, practical multi-task serving workloads exhibit layer- and batch-dependent routing dynamics, making online load balancing indispensable. Existing approaches rely on routing statistics collected after each MoE router, requiring expert weight load or migration to begin only after routing decisions are available, consequently placing migration overhead on the inference critical path. In this work, we observe that online balancing can instead be largely overlapped with computation before target routing (e.g., attention), if routing distributions can be predicted accurately in advance. Therefore, we propose FreeBalance, a lossless online load-balancing framework that overlaps expert migration with preceding computation stages via residual workload prediction. FreeBalance leverages cross-layer similarities in hidden representations within the residual network to build a lightweight workload predictor. This enables proactive expert migration planning before routing decisions are available, creating substantial overlap between weight transfer and computation-heavy pre-routing stages. Furthermore, a cost model constrains the number of swaps to fully hide the synchronization overhead within the available window. Experiments across models and datasets show that FreeBalance reduces the max-to-mean rank load ratio by 32.8% and end-to-end prefill latency by 13.1%. Specifically, our method hides balancing overhead of an average of 5.1 experts per layer, which would otherwise account for about 8.5% of the critical-path latency.
cs.AI / 31 / 2608.14220
A Generalized Parallelogram Rule for Proportional Analogies on Riemannian Manifolds
Abstract
Analogies are quaternary relations of the form "a is to b as c is to d", usually denoted a : b :: c : d. This notion is formalized in particular with the notion of proportional analogy, which imposes some constraints on the valid analogies. Whereas proportional analogies have been studied mostly in symbolic domains and in vector spaces, their use is limited in non-Euclidean spaces. In this paper, we introduce a proportional analogy relation in Riemannian domains, extending the parallelogram rule used for arithmetic analogies in Euclidean spaces. We illustrate the introduced analogy on various manifolds, such as the sphere, shape spaces and manifolds of probability distributions.
cs.AI / 32 / 2608.14221
MathForm: Scaling Mathematical Autoformalization with Knowledge Retrieval and Verification-Guided Refinement
Abstract
Autoformalization is commonly framed as translating natural-language mathematical statements into machine-verifiable formal languages such as Lean 4. However, faithful formalization requires more than translation. Models must map mathematical concepts to the complex hierarchy of types and definitions in formal libraries such as Mathlib, while ensuring that generated statements preserve the meaning of the source propositions. Existing approaches struggle because they rely heavily on the model's parametric memory for library-specific knowledge, while common data construction pipelines often resort to filtering single-pass outputs and lack mechanisms for feedback-driven revision. To address these challenges, we introduce MathForm, an autoformalization framework for constructing verified training data through Mathlib knowledge retrieval and verification-guided iterative refinement. Before generation, a retrieval planner gathers relevant definitions and existing formalizations from Mathlib to guide the formalization generator. Generated statements are then revised using compiler diagnostics and semantic-consistency feedback. Using this framework, we construct FormalVerse, a Lean 4 dataset containing approximately 367K verified examples across diverse mathematical domains and sources. We then train MathForm-8B through supervised fine-tuning followed by reinforcement learning. Across six benchmarks, MathForm-8B achieves average Pass@8 rates of 88.06% under Syntax Check (SC) and 72.37% under Consistency Check (CC), outperforming multiple specialized 32B autoformalizers. On the challenging FATE-H and FATE-X subsets, it attains CC pass rates of 63% and 37%, exceeding the strongest specialized baselines in both cases.
cs.AI / 33 / 2608.14227
Attributing Preprocessing Invariance in Spectral Foundation Models
Abstract
Preprocessing invariance is an appealing goal for spectral foundation models: a frozen model should remain useful when laboratories preprocess spectra differently. It is usually measured by training a classifier under one preprocessing pipeline and testing it under another, with preserved accuracy read as evidence of learning. We revisit that reading, using a Raman foundation model as a case study. Such models normalize their inputs before any learned parameter is applied. If that normalization maps two differently preprocessed spectra to the same vector, the encoder receives identical inputs, so the invariance cannot be attributed to learning. For a normalization that uses each spectrum's own statistics, this happens exactly when one spectrum is a positive multiple of the other plus a constant. Several standard preprocessing operations take that form. The encoder should therefore be measured against the normalization alone, which has no learned parameters. On six Raman evaluation datasets, the model does not measurably outperform its own normalization. It improves on raw spectra, but so does the normalization alone. Training does improve the encoder over random initialization, and a controlled experiment shows that it learns to ignore a transformation only when that transformation reaches it. A numerical test settles which transformations a given normalization removes. Across released systems in five modalities, most normalizations already remove transformations of that form, and several of those systems claim that invariance as learned. Replicating the comparison on two of them shows no gain either.
cs.AI / 34 / 2608.14246
Polaris : Multi Agentic System for Conversational Enterprise Analytics
Abstract
In today's fast-paced environment, the ability to swiftly access, understand, and act on data is no longer optional; it is essential. Yet most organizations remain data-rich but insight-poor, constrained by the complexity of querying, interpreting, and explaining enterprise-scale information. We present Polaris, a supervisor-led multi-agent framework for conversational enterprise analytics that bridges this gap. Polaris introduces Dynamic Task Coordination (DTC), a decision-theoretic orchestration layer that models agent-task assignment as adaptive bipartite matching, enabling real-time coordination, recovery, and optimization across specialized agents for querying, visualization, and reasoning. By coupling DTC with reason-first, ReAct-style agents, Polaris transforms natural-language queries into coherent analytical workflows that not only retrieve and visualize data but also explain the underlying "why." Evaluation on structured enterprise datasets demonstrates high semantic fidelity and answer relevancy, underscoring the potential of multi-agent orchestration to deliver trustworthy, end-to-end business intelligence at scale.
cs.AI / 35 / 2608.14290
Intern-S2-Mobius: Foundation Model with Decoupled Knowledge and Reasoning
Abstract
We introduce Mobius-v0, an architecture that comprises a globally shared Memory (FFN) that stores knowledge vectors and multiple Reasoners (Self-Attn) that iteratively achieve compositional reasoning. Using hidden states as cache and carrier, reasoners repeatedly query memory for required knowledge-vectors, while the knowledge is transmitted back to reasoning operators. Through this knowledge-reasoning-separation architecture, Mobius achieves better knowledge compression and reasoning efficiency. Built upon Mobius-v0 architecture: 1) Our 7B model trained-from-scratch achieves similar downstream score as a 7B Transformer baseline with 62.6% of baseline's training data. 2) Our Intern-S2-Mobius, continually-pretrained from Qwen3.5-35B, achieves similar downstream score while delivering nearly 4x end-to-end inference speedup.
cs.AI / 36 / 2608.14306
Sensor-Driven Mission Synthesis for UAV/UGV Swarms: A TB-CSPN Coordination Architecture with Hardware-Enforced Safety
Abstract
This paper presents a coordination architecture for heterogeneous UAV/UGV swarms that synthesises mission actions from uncertain, multi-modal sensor evidence while preserving hardware-enforced safety at the actuation boundary. The approach combines radar, RF, acoustic, and visual observations with Topic-Based Communication Space Petri Net (TB-CSPN) orchestration to support incremental mission formation under partial and evolving information. Consultant agents transform sensor outputs into temporally bounded semantic tokens, while supervisor agents provide authorisation and policy-governed release of mission transitions. This separation between interpretation, coordination, and execution yields auditable decision paths, constrains non-determinism within the coordination layer through guards and synchronisation, and enables bounded-time integration of heterogeneous evidence. To improve resilience in contested environments, including cyber compromise, spoofing, jamming, and communication loss, the digital coordination layer is complemented by independent analogue safety envelopes that clamp or veto unsafe actuator commands issued to individual vehicles. A coastal-surveillance case study illustrates how the proposed architecture enables dependable, governed, and physically safe swarm coordination under operational uncertainty.
cs.AI / 37 / 2608.14339
Clearing the Fog: Towards Installing and Refining Proactive Exploration Capabilities in LLM Agents
Abstract
We study proactive exploration in LLM agents, i.e., the ability to explore an environment to acquire information that improves future decision-making. In this regard, we first identify two fundamental bottlenecks that hinder this capability and then propose \ours, a novel method designed to instill and refine proactive exploration. Specifically, \ours\ consists of two components: (1) Exploratory Data Construction, which synthesizes exploration-rich trajectories to mitigate the hindsight bias of standard demonstrations; and (2) RL Optimization with Contrastive Signal Guidance, which leverages contrastive trajectory pairs to distinguish productive exploration from redundant wandering. Extensive experiments demonstrate the effectiveness of \ours\ and provide insights into the characteristics of proactive exploration. Our code is available at: https://github.com/GuanZhizhao/SAFARI.
cs.AI / 38 / 2608.14354
ScienceFlow: A long-horizon agent for ML research, scientific discovery and beyond
Abstract
Enabling LLM agents to sustain productive, stable, and goal-aligned research over extended horizons is a central challenge for autonomous machine learning and scientific discovery, as progress hinges on continuously managing evolving state, exploration decisions, and computational resources. Pioneering autoresearch agents, despite great success, still lack mechanisms for continuity, recovery from dead ends, and value-driven compute allocation, which inherently undermines overall search efficiency, wastes computational resources, and lowers the chance of ultimate success. To bridge this gap, we introduce ScienceFlow, an end-to-end autoresearch agent framework that organizes long-horizon research work into research segments grounded in executable workspaces. It represents research progress as recoverable executable states, enabling efficient exploration, revision, and execution. Transitions between research segments are governed by Executable-State Transition through Re-Anchoring (ESTRA), which selects either the live state or an archived state as the next anchor and determines whether to continue or redirect the research trajectory. An evidence-aware execution controller allocates resources to physical jobs based on resource availability, remaining budget, and validated progress. We evaluate ScienceFlow on tasks spanning machine learning, scientific modeling, and mathematical optimization. Results on diverse long-horizon benchmarks demonstrate its ability to sustain effective research processes, highlighted by a SOTA 70.22 percent Any-Medal score on the full MLE-bench within a 24-hour budget, outperforming prior reported results by 4.92 percentage points. The efficacy of ScienceFlow further demonstrates that efficient state management, adaptive exploration, and objective-aligned execution are critical for scaling autonomous research beyond short-horizon interactions.
cs.AI / 39 / 2608.14355
Disentangled Shared Representations Improve Morpho-Transcriptomic Integration
Abstract
Spatial transcriptomics (ST) enables the simultaneous profiling of gene expression and tissue morphology, creating an opportunity to learn multimodal representations capturing shared morpho-transcriptomic structure. However, standard multimodal models often compress modalities into a common latent space without explicitly separating shared and modality-specific sources of variation, which may limit downstream utility. We investigate whether explicit disentanglement of shared and private latent components improves multimodal representation learning for paired Hematoxylin \& Eosin (H\&E) and ST data. We compare VAE-based and contrastive approaches, each in standard and disentangled variants, across two cancer cohorts under matched experimental conditions. Representations are evaluated using cross-modal reconstruction, downstream probing and cross-modal probe transfer. The experiments suggest two main trends. First, contrastive objectives yield higher downstream probing performance than VAE-based models. Second, disentangled variants improve the selected reconstruction and probing metrics, although the gains depend on the model family, task, direction, and disentanglement strength. Overall, our results suggest that explicitly factorizing shared and modality-specific information can improve multimodal representation learning for spatial transcriptomics and provides a useful evaluation framework for future foundation models.
cs.AI / 40 / 2608.14359
Designing Sustainable Federated Learning as a Service using Neural Architecture Search
Abstract
The sustainability constraints of FLaaS consumers pose significant challenges to maintaining carbon-feasible federated training in FLaaS environments. These constraints often lead to infeasible consumer participation and unstable federated training under hard carbon constraints. We propose a Sustainable Federated Learning as a Service (SFLaaS), a carbon- constrained Neural Architecture Search (NAS) framework for heteroge- neous sustainable constraints. We introduce a requirement-driven search space that transforms consumer sustainability profiles into a feasible architecture region before federated execution. We develop a consumer-level carbon feasibility estimation mechanism to evaluate candidate architectures under dynamic carbon conditions. We propose a sustainable con- sumer scheduling strategy that adaptively selects feasible consumers and allocates local workloads to preserve consumer participation and statistical data coverage. An evolutionary search strategy jointly optimised for predictive performance, consumer feasibility, and participation coverage under hard carbon constraints. Experiments on real-world datasets and a simulated environment demonstrate the effectiveness of the proposed approach.
cs.AI / 41 / 2608.14375
Wrong but Useful: Trajectory Value Beyond Answer Correctness in Multi-Agent Messages
Abstract
Multi-agent reasoning systems often use agreement, confidence, or automated scores to decide which messages should shape a final answer. Such filtering assumes that a message likely to be correct is also worth keeping. Yet a wrong answer can contain a useful decomposition, constraint, or scientific principle. We test this distinction with Diverse Hypothesis Deliberation (DHD), a controlled measurement protocol that caches five independently generated messages and replays the same downstream solver, called the integrator, with each message available or hidden. The replay comparison measures a message's trajectory value: whether making the message available helps or harms subsequent reasoning. Across five mathematics and science benchmarks and two openly available model families, gpt-oss-120b and gemma-4-31B-it, wrong-helpful messages appear in every benchmark-model combination. Among wrong-answer messages that change final correctness, more than four in ten changes are helpful in each model. Controlled repeats show that the number of repeatable message effects is unlikely to arise from replay variation alone (p=0.0002). A focused intervention on repeatable wrong-helpful messages finds that the complete message works best, while retaining its reasoning preserves more success than retaining only its answer; the source of the complete-message advantage remains open. Within the same problem, repeated trajectory-value evidence also identifies a better keep-or-remove choice than answer correctness alone. Answer correctness is therefore informative but does not determine trajectory value. DHD measures this missing property and produces reusable labels for learning when agents should listen.
cs.AI / 42 / 2608.14380
AgentRewind: Recoverable Execution for Long-Horizon LLM Agents
Abstract
Many real-world tasks require LLM agents to interact with their environments over long execution horizons. Errors that occur early in execution may propagate through both the agent context and environment state, and their effects may be difficult to reverse through subsequent actions. Existing methods mainly seek to reduce such errors through plan refinement and safety checks but provide little support after errors occur. To enable recovery during long-horizon execution, we present AgentRewind, a runtime recovery framework that records aligned checkpoints of the agent context and controlled environment, allowing agents to return to an earlier state and resume execution with information from previous attempts. We also construct MettleBench, a benchmark for evaluating task completion and partial progress on long-horizon engineering assignments containing a series of related requirements. Experiments across tasks, multiple models, execution strategies, and agent harnesses show that AgentRewind improves task success rate and average checklist progress over the compared baselines.
cs.AI / 43 / 2608.14407
The Past and Future of AI Scientists
Abstract
We present a survey of the past and future of AI Scientists: machines capable of automating science. AI Scientists can originate hypotheses, deduce their consequences, design and execute experiments, interpret their results, and revise their beliefs. Such systems are integrated scientific agents, connected to the literature, formal knowledge, mathematical models, simulations, data-analysis systems and physical laboratories. Adam was the first machine to make novel scientific discoveries through cycles of hypothesis formation and physical experimentation. Eve established the architecture of the modern self-driving laboratory. Foundation models, autonomous agents and laboratory robotics now make it possible to build systems far more general than either Adam or Eve. The central problem is no longer whether individual components of science can be automated. They can. The problem is integration. AI Scientists must combine neural learning with logic, probability, mathematics, causal reasoning, simulation, experimental design, robotics and formal scientific records. AI Scientists have the potential to transform science: to make science faster, cheaper, more systematic and more reproducible. AI Scientists could investigate systems too complicated for unaided human science, and enable thousands of AI scientists to work together on single problems. The Nobel Turing Challenge sets the goal of developing by 2050 AI systems capable of automating Nobel-quality discoveries. Progress is ahead of schedule. When we succeed it will create a new form of science and transform the world.
cs.AI / 44 / 2608.14425
Knowing When to Stop: Bayesian Optimal Stopping for LLM Evaluations
Abstract
LLM evaluations often use fixed sampling budgets, testing every item the same number of times even after estimates are precise. We introduce optstop, a precision-based adaptive stopping framework that treats evaluation as a sequential measurement problem: keep sampling where uncertainty remains high, and stop where estimates are precise or stable enough. The framework builds on hierarchical Bayesian inference, supports binary, ordinal, and continuous outcomes, and keeps every benchmark item eligible for sampling, without requiring a calibrated item bank. It runs live or retrospectively, and includes a safeguard that samples more cautiously as measured performance approaches zero, where rare successes matter most. In an illustrative 200-item, 10-epoch evaluation, it removes 57%-97% of planned trials across nine validation settings, with overall conclusions equivalent to the full run. These results show that LLM evaluation compute can be allocated by uncertainty rather than by fixed repetition counts, with the magnitude of savings depending on evaluation design.
cs.AI / 45 / 2608.14426
The Dynamics of Intelligence Explosions
Abstract
AI is increasingly being used to help with AI R&D. Under certain conditions this feedback loop might be able to produce an intelligence explosion, with rapidly escalating AI capabilities. I explore the mathematics of the most explosive possibilities, with an eye to understanding what drives the dynamics. I show that singular growth (towards a vertical asymptote) is harder to achieve than would be expected from recent economics-inspired modelling, and that there is an important but neglected class of growth rates that are faster than exponential but don't lead to a vertical asymptote. I draw out the generation time (the time to go around the feedback loop) as a neglected parameter that plays a pivotal role in determining the behaviour of any intelligence explosion --- one cannot have singular growth unless the generation time rapidly approaches zero.
cs.AI / 46 / 2608.14441
PACE-Bench: Benchmarking Physics Adaptation via Code Evolution in Dynamic Environments
Abstract
Self-evolving agents improve future behavior from interaction experience, yet existing evaluations typically optimize under fixed execution conditions and do not test recovery after those conditions change. To address this gap, we introduce PACE-Bench (Physics Adaptation via Code Evolution), a simulator-grounded benchmark of 144 source-to-target adaptation pairs across six physics domains. Each pair links a source environment to a mutated target environment with the same goal and interface. A code-driven design that succeeds in the source fails in the target, where agents must iteratively adapt it into a working target design using diagnostic sandbox feedback within a limited attempt budget. We compare ten self-evolving methods from four paradigms. The benchmark remains far from saturated: Reflexion + Qwen3-14B succeeds on only 35.9\% of full-benchmark pairs, while GPT-5.5 solves 66.7\% of the Statics subset under the full budget. Together, these results show that simulator-grounded reflection is more reliable than unverified self-revision, while memory anchors agents to early designs and broad tree search explores without converging. Even revealing exact physical changes does not raise the performance ceiling, pointing to mechanism redesign rather than parameter inference as the central bottleneck. Data and code are available at https://github.com/thunlp/PACE-Bench.
cs.AI / 47 / 2608.14446
Wyvern: An Agentic Framework for Generating Grounded Multimodal Reports
Abstract
In the current artificial intelligence-driven innovation era, the pace of knowledge growth is accelerating, and is hard to keep up with. While generative models are increasingly used to synthesize content, they often lack in information grounding. To address these peculiarities of our time, we propose Wyvern, a multi-agent framework for the automated generation of grounded, multimodal technical reports. Wyvern allows for the generation of multimodal outputs, integrating images, tables, and text with supporting references in a unified report. Additionally, a particular focus is placed on the grounding of the content, with the implementation of a claims auto-revision stage. We conduct a human evaluation study to assess the quality of our proposed framework. The results show that the figures' informativeness is perceived as superior to that of a recent baseline in 87% of cases. Furthermore, Wyvern's reports are rated as more useful than those produced by three alternative methods in 63% to 100% of instances. We also carry out automatic evaluations showing that Wyvern gains up to 2.3$\times$ in citation recall and 1.6$\times$ in citation precision with respect to the baselines.
cs.AI / 48 / 2608.14456
Shift Aware Transfer Learning with Adaptive Dual-Encoder Fusion for PM Forecasting in Data-Limited Environments
Abstract
Short-horizon forecasting of fine particulate matter (PM2.5) remains difficult when observations from the target domain are limited and the statistical properties of the source and target domains differ. In these settings, models trained only on local data may not capture complex temporal dynamics, while direct transfer learning can result in negative transfer. This study develops a shift-aware dual-encoder transfer framework that combines source-domain knowledge with target-specific representation learning. The source encoder was pretrained using hourly observations from 10 U.S. monitoring locations. The framework was then adapted and evaluated using two years of hourly observations from 77 stations in Taiwan under a chronological train-validation-test protocol. Among the four principal baselines, the frozen-source dual-encoder model achieved the best performance, with MSE = 21.8960, MAE = 3.1597, and R^2 = 0.8725. This corresponds to an MSE reduction of approximately 7.1% relative to TL-v1 and 4.1% relative to TL-v2. The ablation analysis showed that removing the Taiwan-specific branch caused the largest decline in performance. Allowing the source encoder to adapt produced the best overall result, with MSE = 21.6575, MAE = 3.1383, and R^2 = 0.8739. SHAP analysis indicated that predictions were driven mainly by recent PM2.5 observations and meteorological variables related to pollutant transport and dispersion. These results suggest that source-domain knowledge is most effective when target-specific information is preserved and the transferred representation is allowed to adapt under target supervision.
cs.AI / 49 / 2608.14490
Twin: Playing an Unknown Game with a Test-Time Digital Twin
Abstract
We present a Test-time World-model Inference (Twin) system, in which a frontier coding agent writes an executable world model for completing continual learning tasks, such as ARC-AGI-3 games. Traditional approaches hand-engineer such models, one custom design per task. Each game hides its rules and goal, and our system constructs them from simulation and interaction alone. Its inductive prior over grid games is strong enough to recover the true transitions of the game and the goal on nearly all levels. Replay validation happens in a twin world model. The harness enforces that an action is not made until the program reproduces every previous observed game transition. Each mismatch between a world model prediction and the actual action result becomes a counterexample that is used to repair the world model. Twin clears 179 out of 183 levels (97.8%), and does so more efficiently than humans in 158 out of 179 levels (88.3%). The system infers the goal before any reward on 156 of the levels it clears (87.2%), and in the remaining levels automatically discovers the goal by search. The benchmark scores completion and action efficiency, between 0 and 100, against humans playing each game for the first time. Played directly, the base model scores only 7.8%; an off-the-shelf harness increases it to 61.1%, whereas our twin world model increases the same base model to 93.3%, clearing 23 out of 25 games. Building a usable world model is simpler than anticipated, whereas the harder problem is inferring the right goal.
cs.AI / 50 / 2608.14509
Split the Labor: Separating Evidence Interpretation from Decision Aggregation
Abstract
Systems that ask a language model to reach a conclusion from many sources usually concatenate them into one prompt. This conflates two operations with different requirements. Interpreting a source rewards capacity and context. Combining interpretations rewards fixed arithmetic, comparability across instances, and the option to return nothing. Once separated, the design problem becomes the interface between them. We propose a four-field evidence tuple (hypothesis, reliability bucket, rationale, provenance) and show that fixing it determines both halves. The separation also reveals a failure mode in how such systems combine, which we call count-scale drift. Thresholding a sum of unnormalized weights is exactly posterior thresholding, but at an operating point that slides with the number of sources consulted. The slide grows with reader reliability. When source reliabilities differ, the vote rule and the posterior order instances differently, and no threshold reconciles them. Pooling calibrated log-likelihood ratios addresses both problems. The fix is arithmetic rather than architectural, and applies to a class of rules beyond language models: score-summing triage engines, diagnostic panels scored by counting positives, and additive multi-signal detectors. We then instantiate the principle twice on one longitudinal corpus, once after outcomes resolve and once before. The same partition helps in both, at different granularities: over reading in the first, over learning capacity in the second. There, a small sequence encoder on an easy auxiliary objective plus a tree ensemble carrying the censored survival loss reaches 0.921 AUPRC against 0.805 for a hand-crafted baseline. We separate what transfers from what must be re-estimated per domain, and state five predictions that would falsify the framework, three negative results, and which comparisons remain confounded.
cs.AI / 51 / 2608.14522
Participatory Moral AI Is Not Neutral: The Invisible Hand of Developers
Abstract
As AI systems make more morally loaded decisions across society, one response has been moral preference elicitation. In this approach, researchers poll participants on hypothetical dilemmas and use the aggregated votes to train a policy that an AI model then applies at scale. Before any vote is cast, developers make three key choices in the moral AI elicitation pipeline: feature scoping, voter sampling, and question framing. In other words, they decide which features go to a vote, which voters to include, and how to present the question. These choices are often opaque, undocumented, and treated as technical details rather than normative ones. We examine each of these choices within a common empirical study and show that each can shape the preferences produced by moral AI elicitation. Across two phases (N = 809) in three deployment contexts (i.e., AI kidney allocation, AI agents simulating absent workers, and generative AI depictions of the deceased), we examine the three main stages of the moral AI elicitation pipeline. First, morally relevant features shift across contexts. This suggests that feature schemas should not be assumed to transfer across deployment domains. Second, preferences differ by political ideology for roughly one-third of features, with some differences reversing direction. The ideological composition of the voter pool can therefore affect the resulting aggregated preference profile. Third, the wording of the elicitation question can narrow or widen ideological gaps by up to a full scale point. The framing conditions also change how moral foundations are associated with participants' judgments. Taken together, these findings suggest that voting-based alignment cannot deliver fair or transparent AI by aggregation alone; at minimum, each stage of the moral AI elicitation pipeline should be audited and disclosed.
cs.AI / 52 / 2608.13690
MedPlex: Deep Vision-Language Co-Adaptation for Clinically Grounded Medical Segmentation
Abstract
Medical image segmentation is still largely treated as a vision-only problem, although clinical interpretation often relies on textual knowledge of anatomy, location, appearance, and surrounding context. Existing text-guided segmentation methods within the Vision-Language Model (VLM) paradigm often use language only as a late conditioning signal, limiting its influence on visual representation learning. We introduce MedPlex (Medical Plexus of Vision and Language), an end-to-end VLM framework that makes text guidance a continuous, clinically grounded component of segmentation learning. Through Bi-Fusion (Bidirectional Fusion), visual and textual representations evolve jointly across the encoding hierarchy. MedPlex further introduces class-level and region-level concept alignment to organize the shared representation at complementary granularities. Class-level alignment anchors each anatomical target to an aggregated clinical concept profile, while region-level alignment preserves individual concepts, such as shape, location, appearance, and texture, through class-specific visual evidence. In this way, language provides structured supervision throughout the encoder rather than serving only as a late-stage cue. MedPlex achieves state-of-the-art performance across CT and MR benchmarks for multi-organ, cardiac substructure, and tumor segmentation, including settings with real free-text clinical supervision. Code: https://github.com/rafiibnsultan/MedPlex.
cs.AI / 53 / 2608.13939
CMCNet: Aligning Ultrasound Image Embeddings with Textual TI-RADS Representations for Fine-Grained Thyroid Classification
Abstract
Ultrasound is the primary imaging modality for assessing thyroid nodules, and the ACR TI-RADS framework standardizes diagnosis through five ultrasound feature categories that are aggregated into five risk levels (TR1-TR5). Although widely adopted in clinical practice, most deep learning approaches focus on binary malignancy classification, while multi-class prediction and explicit utilization of feature-level supervision remain underexplored, largely due to limited annotated data. In this study, we introduce the STN dataset of 600 thyroid nodules with paired transverse and longitudinal ultrasound images, bounding box annotations, and complete labels for all five TI-RADS feature categories. Following the clinical decision process, we investigate how structured feature information can guide representation learning during training while requiring only images at inference. We demonstrate that text embeddings derived from standardized feature descriptions form a stable surrogate representation for TI-RADS risk levels. Based on this observation, we propose CMCNet, which aligns image embeddings to fixed textual embeddings via a Center-Margin Contrastive Loss that simultaneously promotes intra-class compactness and inter-class separation. Experimental results show that this embedding alignment strategy is more data-efficient and robust than direct multitask learning, and consistently outperforms InfoNCE, center loss, a strong multitask baseline, and a VQA-style multimodal model, particularly in imbalanced settings. The dataset is freely available at doi: 10.5281/zenodo.19125693 and the source code is available at: https://www.healthinformaticslab.org/supp/.
cs.AI / 54 / 2608.14015
MedClaw: Heuristic Agent Harness for Long-Horizon Surgical Video Reasoning
Abstract
Understanding tens-of-minutes surgical videos requires long-horizon temporal reasoning, answering what happens before, after, or across stages of a procedure by grounding the question in visual evidence spread across time. Existing approaches handle this poorly: a one-shot vision-language model (VLM) compresses the whole procedure to fit its context window and loses the detail a "before" or "after" question depends on, while video agents that train the model where to look are data-hungry and transfer poorly to out-of-domain surgery. We build an agent harness that separates reasoning from perception and improves by evolving context rather than optimizing weights. A text-only orchestrator plans which evidence to gather and issues an auditable sequence of tool calls, while frozen vision-language sub-agents execute each call over the pixels, viewing, cropping, inspecting frames, and retrieving external knowledge. We further propose a gradient-free, reward-gated Heuristic Skill Distillation loop that mines the agent's own low-scoring traces and keeps a candidate skill only when it raises a validation reward, yielding reusable retrieval skills, notably directed re-look. Growing an external skill library rather than tuning weights, the loop adapts from only about 100 labeled examples, far fewer than supervised or reinforcement fine-tuning requires. To evaluate this agent, we introduce MedClawBench, a de-leaked, doctor-grounded benchmark of 1,123 questions over self-built long neurosurgery recordings and a held-out public lecture-video test split. Across both datasets and all four evaluation dimensions, our agent consistently outperforms one-shot VLMs and general video-agent frameworks, with the largest gains on the long, out-of-domain neurosurgery videos. Project page: https://fyycs.github.io/medclaw/.
cs.AI / 55 / 2608.14016
Content Based Video Narration of Gameplay with Vision Language Models
Abstract
Live game commentary is scarce: it exists for professional esports broadcasts and almost nowhere else. We present a content-based video narration system that produces spoken, esports-style commentary for arbitrary gameplay recordings using a general-purpose vision-language model (VLM) and a text-to-speech back end, with no game-specific instrumentation, no engine telemetry, and no task-specific training. Three mechanisms carry the system. Temporal mosaic packing arranges nine uniformly sampled frames into a single 3x3 image, letting an image-native VLM reason about motion while consuming one image payload per segment instead of nine. Context-conditioned prompting replays the K most recent narrations as assistant-role history, suppressing the repetition that dominates per-segment captioning of static scenes. Duration-conditioned generation and elastic alignment constrain narration length in the prompt, then time-scale or symmetrically pad the synthesized audio so each utterance fills its segment slot exactly, giving frame-accurate muxing without a forced aligner. The implementation supports either cloud TTS or a 6-bit quantized 4B-parameter on-device TTS model on Apple silicon, making the speech stage fully local. We report a qualitative case study on real-time strategy footage, a cost model showing the mosaic reduces per-minute image payloads by 9x, and a candid account of observed failure modes - hallucinated game state, resolution loss from mosaicking, and prosody artifacts from time-scaling. We release the system as a reproducible baseline, with an evaluation protocol for the quantitative study a full version will report.
cs.AI / 56 / 2608.14022
ForgeWM: Progressive Causal Training for Few-Step Action-Conditioned Video World Models
Abstract
Action-conditioned video world models require low-latency causal generation and reliable responses to game-native controls. Although causal distillation enables one- or few-step video synthesis, extending it to interactive world models remains challenging, as discrete keyboard states and continuous mouse motion must remain aligned with temporally compressed latent chunks during causal training and autoregressive rollout. We introduce ForgeWM, a progressive framework that transforms a bidirectional action-conditioned video generator into efficient few-step world models through domain adaptation, teacher-forced causal training, causal consistency distillation, and on-policy distribution matching with a bidirectional teacher. The resulting budget-specialized students operate at steady-state denoising budgets of 1, 2, and 4 steps. ForgeWM further supports a dual-path deployment protocol combining latency-critical interaction with optional replay-time refinement, where the one-step student re-noises and refines its saved draft. On paired Minecraft trajectories, ForgeWM leads the evaluated systems in Imaging Quality, reference-aligned motion-profile agreement, action-sign accuracy, and mouse-control accuracy, while achieving the lowest reference LPIPS; the same four-stage recipe transfers to gamepad-controlled FPS gameplay. Replay-time refinement matches four-step reference quality while remaining roughly three times closer to the experienced trajectory than regeneration from noise. These results demonstrate ForgeWM's effectiveness for controllable few-step video generation.
cs.AI / 57 / 2608.14058
Voxel-based 3D Facies Segmentation from Seismic Data: A Comparative Study
Abstract
Seismic facies segmentation has emerged as a significant challenge in geophysics, requiring robust methods and systems to effectively identify geologically analogous facies with limited labeled data. Although existing studies have shown promising results in 2D facies segmentation, they often preprocess the original 3D seismic volumes into sets of 2D slices, typically the inline and crossline directions, and treat this problem as a purely 2D segmentation task. This simplification introduces discontinuities across slices and fails to preserve the spatial and structural continuity in 3D seismic data, thus limiting the model's ability to learn coherent geological patterns. In this work, we present a comparative and reproducible benchmark for voxel-based 3D seismic facies segmentation, built upon publicly available seismic volumes including the Netherlands F3 and the Parihaka datasets, with standardized data splits and evaluation metrics. By evaluating the three representative families of modern 3D segmentation architectures, we establish strong baseline results that highlight the potential and remaining challenges for future research in this domain.
cs.AI / 58 / 2608.14112
Fixed-Budget Gaussian Volume Encoding with Structure-Aware Allocation
Abstract
Scientific simulations often produce scalar volumes faster than they can be stored, transferred, and loaded, while in situ reduction must use only a limited share of simulation resources. This work encodes scalar fields as anisotropic Gaussian primitives under a fixed budget. The complete primitive set is allocated analytically from local field structure, including position, orientation, and shape, then refined directly against the scalar field without densification, pruning, or count changes. The selected budget determines encoded storage before refinement and, together with the iteration schedule, provides a controllable refinement-time budget. In a controlled benchmark, truncation-aware field evaluation reduces encoding time by up to 51x; 1.4 million Gaussians encode a billion-voxel volume in at most four minutes on one desktop GPU, with reduced-iteration refinement completing in under one minute. Across five datasets spanning 2.1 million to 1.1 billion evaluated voxels, compression-useful configurations achieve 15.0-38.7 dB PSNR at compression ratios from 2.2x to over 40,000x. Pre-encoding structure statistics characterize fields for which one-shot allocation yields limited gains from additional capacity. Because primitives retain scalar attributes rather than baked appearance, a single compact model serves every subsequent visualization state - supporting post-hoc transfer-function, colormap, lighting, and viewpoint changes without re-encoding.
cs.AI / 59 / 2608.14136
HiCo-GS: Hierarchical Context Aggregation and Geometric Consistency for Octree Gaussian Splatting
Abstract
Octree-based anchor Gaussian Splatting has emerged as a scalable representation for city-scale novel view synthesis, where multi-level anchors adaptively capture scene content from coarse building structures to fine architectural details. However, we identify a fundamental limitation in existing methods: cross-level feature isolation, where each level's anchor features are optimized independently with no inter-level communication, causing color drift on building facades and over-smoothing in textured regions. We present HiCo-GS, a high-fidelity reconstruction framework with two complementary modules. Cross-Level Context Aggregation (CLCA) enables bidirectional hierarchical prior injection by leveraging the octree's spatial containment structure to aggregate per-level context vectors into parent-self-child triplets, fused via a lightweight MLP with residual connection. Coarse-level structural priors flow down to inform fine-level anchors, while fine-level detail statistics feed back to prevent over-smoothing, at negligible computational overhead. Depth-Normal Geometric Consistency (DNGC) regularization enforces agreement between rendered normals and depth-derived normals through an alpha-weighted consistency loss, complemented by edge-aware smoothness losses with progressive warmup that exploit the strong planar priors ubiquitous in urban geometry to suppress floating artifacts. We further introduce the China-Pagoda dataset comprising 8 ancient Chinese pagodas with over 1,200 images each, featuring dense ornamental carvings, curved multi-layer eaves, and repetitive fine-grained textures. Extensive experiments on Mill19, UrbanScene3D, MatrixCity, and China-Pagoda demonstrate that HiCo-GS achieves state-of-the-art rendering quality and substantially cleaner geometry across real-world and synthetic urban benchmarks.Code: https://github.com/WZ-CS/HiCo-GS.
cs.AI / 60 / 2608.14138
SPARGen: Unifying Spatial Perception and Reasoning through Native Multimodal Generation
Abstract
Spatial perception and reasoning from visual observations require recovering geometric structure, establishing correspondences, and understanding spatial relations. Existing approaches typically address these capabilities separately using task-specific architectures or external geometric modules, limiting knowledge transfer among complementary representations of the same physical scene. We introduce SPARGen, a unified multimodal framework that casts 3D reconstruction, dense correspondence, and spatial reasoning as instruction-conditioned generation tasks. SPARGen serializes compact structured and linguistic outputs as token sequences while generating dense geometric fields in image-aligned forms, enabling spatial supervision to jointly shape shared representations within a native multimodal generative model. Experiments across benchmarks for 3D reconstruction, correspondence, and spatial reasoning show that SPARGen achieves competitive performance across heterogeneous spatial tasks within a single native multimodal generative framework.
cs.AI / 61 / 2608.14144
Self-Supervised Visual On-Policy Distillation
Abstract
Visual on-policy distillation relies heavily on an informative teacher-student asymmetry, through either a larger, stronger teacher or privileged supervision, such as reference answers or ground-truth regions of interest. This raises a fundamental question: where can informative asymmetry come from when nothing privileged is available? We answer this by inverting where the asymmetry comes from. Rather than adding privileged information to the teacher, we subtract information from the student. This asymmetry creates the same effective learning signal for free as a teacher with access to information unavailable to the student, without ground-truth annotations, rewards, or a separate stronger teacher model. Building on this principle, we introduce Self-Supervised Visual On-Policy Distillation (S$^2$VOPD), a simple yet effective method that constructs on-policy learning signals from asymmetric augmented views. S$^2$VOPD distills the teacher's distribution conditioned on the original image on-policy into the student distribution conditioned on a strongly augmented view of the same image. We systematically explore a broad design space of visual augmentations and uncover that (1) asymmetry matters: all four augmentation families improve performance, while symmetric self-distillation degrades it; (2) strength matters: performance peaks at a moderate strength; and (3) the gap must remain task-consistent: augmentations that completely remove the question-relevant evidence can induce large but uninformative discrepancies. Across six fine-grained perception benchmarks, S$^2$VOPD improves Qwen3.5-4B from 70.7% to 77.4%, above all open-source models compared, up to Qwen3-VL at 235B, and surpasses GPT-5.4. While holding training data the same, it recovers 96% of the improvement achieved by methods with privileged information. Website is at https://williamium3000.github.io/s2vopd
cs.AI / 62 / 2608.14286
Seeing Red, Thinking Bad: Color Bias in Vision Language Models
Abstract
Vision language models (VLMs) are increasingly used in industrial decision-making systems, such as recruitment support and recommendation. This motivates careful analysis of how VLMs process visual and textual information. In this work, we study how VLMs interpret text rendered as an image, and investigate the influence of visual styling biases. To this end, we introduce Stealth Visual Prompts, which subtly change visual styling of text, such as color and contrast, while preserving semantic content. Using these prompts, we systematically control the visual styling of words in text and measure their impact on the analysis performed by VLMs. We further analyze how such visual perturbations affect the latent representations of the vision encoder. From our experiments, we observed that coloring positive words in green consistently shifts sentiment predictions toward a positive direction. As a result, VLMs often fail to properly account for negative words present in the text. Our analysis suggests that this behavior is correlated with changes in the latent representations of the vision encoder induced by color variations. In addition, we show that reducing text--background contrast increases reliance on visually salient cues and leads to more incorrect Visual Question Answering (VQA) outputs. These results suggest that the visual styling of rendered text can guide VLMs' interpretation in ways that diverge from human semantic understanding. Project page: https://github.com/KohsukeIde/color-bias-vlm
cs.AI / 63 / 2608.14317
Intelligent Detection of Mechanical, Electrical, and Plumbing (MEP) Metrics Based on 2D Floor Plans
Abstract
This research developed a neural network-based model to extract various information from 2D floor plans. We detect lighting symbols, identify the appropriate type of light, and extract the associated texts with lights. The study aims to enable efficient floor designing and determining the number and type of lights needed per floor, i.e., allow efficient design and estimate the power requirement of the floor plan. The model was developed using Mask RCNN as the base. The images were annotated and converted into a Coco data format for training the model. The model achieved bbox\_mAP and segm\_mAP values of 0.7596 and 0.7111, respectively. It also performed well at different IoU thresholds, i.e., with bbox\_mAP 50 and segm\_mAP 75 values of 0.9850 and 0.9219, respectively. The developed model will help various industries, such as architecture and construction, to improve design time and create efficient workflows by automatically detecting Mechanical, Electrical, and Plumbing (MEP) objects from floor plans, and it is the first step towards building tools that will help energy-efficient building design.
cs.AI / 64 / 2608.14389
GBU-Palm: A Multimodal Video Dataset and Benchmark for Palm Presentation Attack Detection
Abstract
Existing palm presentation attack detection (PAD) datasets are often limited by static imagery, restricted acquisition conditions, or insufficient multimodal video data, hindering systematic evaluation across environments, modalities, and attack types. We present GBU-Palm, a large-scale multimodal video dataset and benchmark containing 21,326 videos from 105 subjects and 210 palms across six acquisition environments, including bona fide, Print, and Replay presentations, with 6,310 synchronized RGB-NIR samples. We construct leakage-controlled protocols that separate palm identity and attack lineage and benchmark four representative video architectures under environment-matched and held-out-environment settings. Results reveal substantial architecture-dependent degradation under environmental shift and show that RGB-NIR fusion does not consistently outperform RGB-only input. We further analyze model behavior through true accept (TA), true reject (TR), false accept (FA), and false reject (FR) decomposition, spectral masking, temporal-order intervention, and frozen-backbone NIR probing, revealing distinct failure patterns and evidence utilization across architectures. GBU-Palm provides a unified and challenging benchmark for developing and evaluating robust multimodal palm PAD methods under cross-environment conditions.
cs.AI / 65 / 2608.14391
Can We Defend Against AI-Generated Video Attacks on Real-World Crisis Events? A Systematic Evaluation of Detectors, Generators and Social Dissemination
Abstract
Recent video generators can fabricate realistic depictions of wars, disasters, public emergencies, and other real-world crises, creating substantial risks of misinformation. Existing benchmarks, however, provide limited evidence on detector and generator behavior in such settings, including how detectability varies with generation conditions, how people perceive generated videos, and whether detectors remain reliable during social dissemination. To address this gap, we introduce RA-Bench, a benchmark for AI-generated video detection that uses Real videos as Anchors. RA-Bench contains 17,886 videos, comprising 1,830 real-video anchors across 10 social-risk categories and 16,056 generated clips from four open-source and five closed-source generators. Based on RA-Bench, we organize our evaluation along three dimensions. We first assess detector generalization across seven traditional detectors, ten zero-shot multimodal models under three review settings, and two MLLMs specifically fine-tuned on AI-generated video detection. Across these methods, none of the three detector families generalizes consistently across RA-Bench instances. We then examine how detectability varies with generation quality, conditioning information, and sampling seeds. These analyses show that generation properties affect detector families differently, while source-level detection patterns remain stable across seeds. Finally, we study human authenticity judgments and detector reliability during social dissemination. We find that videos that mislead people are also difficult for current detectors, and that social dissemination makes detection harder. Together, these findings show that current methods struggle to detect realistic AI-generated videos, highlighting the need for detectors robust to evolving video generators.
cs.AI / 66 / 2608.14539
Decoding the Past: An Uncertainty-Aware Deep Learning Framework for Sex Attribution in Prehistoric Hand Stencils
Abstract
Determining the biological sex of the individuals who created Upper Paleolithic hand stencils remains a challenging problem due to the absence of ground truth, population differences between contemporary and prehistoric groups, and the uncertainty introduced by image degradation. Traditional morphometric methods suffer from high structural overlap across sexes, poor cross-population generalizability, and subjective feature engineering. This study presents an uncertainty-aware deep learning framework for sex attribution in prehistoric hand stencils that explicitly models, propagates, and aggregates uncertainty throughout the analytical pipeline. The methodology combines dual image processing, dual contour extraction, structured silhouette augmentation, model architectural diversity, and ensemble-based decision aggregation. The pipeline generates twelve plausible silhouette realizations per stencil to capture boundary uncertainties, which are processed by two ensembles of ten deep neural networks each (EfficientNet-B3 and MobileViT-S) trained on 14,036 contemporary hand samples. Furthermore, a triangulated validation scheme integrates ensemble predictions with unsupervised 2D latent-space manifold mapping (UMAP + k-NN) and explainable AI spatial attributions (LayerCAM) to ensure anatomical consistency. On contemporary data, ensemble models achieve strong classification performance, with accuracies exceeding 88% in older age groups. When applied to prehistoric stencils, the framework produces both sex predictions and confidence measures of internal agreement, enabling the distinction between morphologically stable and ambiguous cases. Convergence across ensemble predictions, latent-space structure, and interpretability analyses shows that uncertainty can become a measurable component of archaeological inference, enabling robust and reproducible decoding of ancient rock art.
cs.AI / 67 / 2608.13712
Reading Between The Lines: Modeling and Evaluating Behavioral Realism in Legal Simulation
Abstract
Deposition training requires attorneys to manage dynamic witness behavior, yet legal-AI evaluations largely focus on factual accuracy, reasoning, or response-level plausibility. We introduce WitnessSim, a deposition simulator driven by controllable legal personas. We use an evaluation framework separating behavioral realism from pedagogical usefulness. We assess realism through adversarial testing, blinded attorney comparison, and analysis of longitudinal behavioral trajectories. WitnessSim generally maintained plausible behavioral boundaries, and attorneys did not systematically prefer either original testimony or WitnessSim generated testimony. Pedagogical tests showed that witness behavior changed meaningfully in response to question form and attorney intervention without uniformly collapsing the assigned persona. Together, these results showcase a model of behavioral fidelity in legal simulations, and provide a framework for evaluating its performance.
cs.AI / 68 / 2608.14254
Meteorology-driven Causal Nowcasting of Fugitive Landfill Emissions Enables Proactive Public Health Response
Abstract
Fugitive emissions from waste sites increasingly expose communities to toxic and odorous gases, yet public-health responses remain largely retrospective, with episodes investigated only after residents have been exposed. Here we show that the meteorological drivers of elevated hydrogen sulphide (HS) at a long-monitored European landfill, and the timescales over which they act, can be identified directly from routine monitoring data. We introduce CAIRN (Causal-Anchored Inference for Receptor Nowcasting), a machine-learning framework whose internal memory is matched to these measured timescales: a fast component tracking hour-scale wind-borne transport and a slow component tracking multi-hour weather changes. Trained to predict gas measurements, CAIRN operates using only routine weather variables and the calendar, without hand-engineered features. Its behaviour is consistent with the identified transport mechanisms, and the framework transfers unchanged to a second monitoring station and to co-emitted methane. Combining four such nowcasters produces a site-level, tiered alert aligned with WHO odour guidance that closely reproduces the alert generated by a direct sensor network and tracks an independent record of community odour complaints. Weather-driven nowcasting can therefore estimate community impact as an emission episode unfolds, providing public-health authorities with a validated, graded trigger for intervention and enabling exposure to be reduced during events rather than after them.
cs.AI / 69 / 2608.13944
Musical Mirrors: The LLM as Sounding Board in Songwriting
Abstract
This paper examines a use of AI in creative practice as an interpretive sounding board for human-generated material, rather than the more familiar pattern of AI generation followed by human curation. Through the lens of resonance as theorized by Hartmut Rosa, I present a first-person case study of songwriting from July 2025 to March 2026, drawing on 16 original pieces in English, French, and other languages along with piano solos. I describe a configuration in which resonance is not located between user and model, but in the author's deepening contact with their own material, mediated through the model. This kind of resonance was supported rather than inhibited by AI when sounding-board behavior was cultivated through sustained calibration by the user. Two failure modes appeared when calibration was absent: sycophantic drift and magical overinterpretation. This account suggests both the potential and the risks of AI as an interpretive partner in creative practice.
cs.AI / 70 / 2608.14405
From Style Replication to Style Exploration: Enabling Art Style Exploration with Analyze-Experiment-Resituate Framework
Abstract
Art style is a signature of professional digital artists that develops through repeated experimentation, reflection, and adaptation. While generative AI (GenAI) can reproduce styles with high fidelity, current tools provide limited support for exploring new stylistic directions and may encourage style replication over exploration. To address this gap, we propose Analyze-Experiment-Resituate (AER), a framework for AI-assisted style exploration derived from interviews with 10 professional digital artists. Rather than prioritizing visually appealing outputs alone, AER supports three core practices of style exploration, including interpreting references, trying out stylistic possibilities, and reflecting on how emerging styles may be received. Specifically, AER enabled artists to (1) analyze artworks into interpretable stylistic elements, (2) have controllable experimentation guided by their own choices, and (3) resituate emerging styles through simulated social perspectives. We implemented AER in a prototype system and evaluated it in a controlled study with 16 artists. Compared with a direct style-transfer workflow, AER increased artists' agency and reflection as they pursued new stylistic directions. A two-week field study with four artists revealed how the AER framework influenced daily style exploration, such as reflection, experimentation, and stylistic decision-making at each stage. We discuss opportunities and challenges in designing AI-assisted style-exploration workflows, and outline implications for future artistic support tools.
cs.AI / 71 / 2608.13833
AdsWorldEngine: A Self-Evolving Conversational Advertising Agent through Orchestrator and Tool Coevolution
Abstract
Conversational advertising aims to deliver useful ads within multi-turn assistant interactions. Unlike conventional query-based advertising, where the user's intent is often expressed in a short standalone query, conversational ads must infer latent commercial intent from the current user query, the assistant response, and dialogue history while also deciding whether an ad would be helpful rather than intrusive. We propose AdsWorldEngine, an agentic framework for conversational advertising. AdsWorldEngine uses an Opportunity Gate to determine whether ads should be shown, an Orchestrator to generate commercial intents, call advertising tools, and construct a top-3 ad slate, and an Evaluator to score delivered ads for offline optimization. The central contribution is an iterative actor-tool training procedure: we first train the Orchestrator with supervised fine-tuning and agentic reinforcement learning, then use high- and low-reward rollouts to construct preference data to train tools. This creates a self-improving loop in which the system learns not only how to use advertising tools, but also how to improve them from rewarded behavior. To support subjective production decisions, we introduce label grounded judgment modeling, which trains judgment models from human labels collected under explicit guidelines. It enriches labels with thinking traces, filters inconsistent rationales through reflection, and further optimizes binary judgments with a cost sensitive GRPO variant that preserves asymmetric reward gaps. Offline, AdsWorldEngine improves diversity by 60% and relevance by 80% over the current production ad delivery system. In an online A/B test, it increases RPM by 22% and ads coverage by 74%.
cs.AI / 72 / 2608.14011
EchoRec: Multi-Item Prediction-Empowered Generative Recommendation via Cycle-Consistent Preference Alignment
Abstract
Generative recommendation autoregressively generates the semantic IDs of the target item, unifying preference modeling and index retrieval within the shared token space. Recent attempts have introduced Multi-Token Prediction (MTP) into this field, yet they primarily inherit its efficiency merit, leaving its potential as dense supervision unexplored. Unlocking this potential hinges on whether future behaviors qualify as informative supervision. Our analysis reveals that future behaviors carry a semantic echo of the current one far above that of random pairs, which nevertheless decays along horizons under intent transitions, making them informative yet order-dependent signals. Motivated by this, we propose EchoRec, which empowers MTP with cycle-consistent holistic preference alignment across multi-horizon for generative recommendation. It comprises two synergistic modules. Horizon-aware Preference Generation (HPG) sequentially chains lightweight auxiliary branches upon the base recommender, where each branch conditions on its predecessor to respect preference evolution. Verifiable Holistic-Preference Alignment (VHA) further consolidates them into the holistic preference and echoes it back through cycle-consistent projectors to suppress spurious alignment, with theoretical guarantees that exclude the rank-collapse form of spurious alignment under an invertible transport, enabling the holistic preference to be retained in the decoding representation. All auxiliary components serve as disposable scaffolding discarded at inference, introducing negligible online serving overhead. Extensive experiments on three datasets demonstrate the superiority of our EchoRec, together with its naturally acquired multi-item generation ability. Our code and datasets will be available upon acceptance.
cs.AI / 73 / 2608.14032
HAM-RAG: Hierarchy-Aware Multimodal RAG for Structure-Faithful Interleaved Generation
Abstract
Existing multimodal RAG methods often flatten structured documents into isolated text and image units, weakening the source organization and local text-image logic needed for faithful evidence selection and placement. We propose HAM-RAG, a Hierarchy-Aware Multimodal RAG framework for structure-faithful interleaved generation. HAM-RAG uses document hierarchy as a grounding signal across retrieval and generation, contextualizing textual and visual evidence and preserving source position and local text-image relations in the prompt. We further introduce HAM-Bench, covering Wukong, Wiki, arXiv, and Recipe across game walkthroughs, web pages, scientific papers, and step-wise recipe documents. Across multiple backbones, HAM-RAG improves the main multimodal average by 17.3% over the strongest non-hierarchical baseline. On Wukong, HAM-RAG improves Img-CBS by 24.2% over the strongest non-hierarchical baseline, demonstrating substantially better local text-image alignment. The main experiments and ablation study together demonstrate that document hierarchy is a key grounding signal for faithful image selection, placement, and local text-image alignment. These findings highlight the value of hierarchy-aware grounding for reliable multimodal assistants that generate answers faithful to the source organization, procedural structure, and local text-image evidence of structured documents, such as technical manuals, maintenance guides, and industrial SOPs. The code is available at https://github.com/MCCodeAI/HAM-RAG.git.
cs.AI / 74 / 2608.13809
Optimal Power Allocation and AI Receiver Design for Superimposed DMRS and Data Transmission
Abstract
In this paper, we consider transmissions with superimposed (SI) demodulation-reference-symbol (DMRS) and data in orthogonal frequency-division multiplexing (OFDM) based multiple-input multiple-output (MIMO) systems. First, we derive an analytical framework to characterize the iterative behavior between the mean-square errors (MSEs) of channel estimation (CE) and MIMO detection (MD) within an iterative CE and detection (ICED) process. This framework is subsequently utilized to optimize power allocation and pilot patterns between the DMRS and data symbols for SI-DMRS transmission. Second, we design an artificial intelligence (AI) based receiver built upon Transformer encoders for SI-DMRS transmissions, which incorporates an iterative CE and detection (ICED) structure. Simulation results demonstrate that the proposed AI-ICED receiver, combined with SI-DMRS, effectively increases spectral efficiency (SE) compared to conventional systems using non-overlapped DMRS and data symbols.
cs.AI / 75 / 2608.14511
Learning-to-Transition for Large-scale and High-Order MIMO Detection
Abstract
High-order multiple-input multiple-output (MIMO) detection requires efficient search over a large discrete symbol space while producing reliable soft information for channel decoding. This paper develops a learning-to-transition (L2T) framework that formulates MIMO detection as a stochastic sequence of complete-vector transitions. At each transition, a channel-coupled Transformer updates both the instance embedding and the sampling policy, while a blockwise autoregressive factorization captures inter-stream dependence with moderate sequential complexity. For hard-output detection, a transition network is applied recursively and trained through a residual-to-BER curriculum, which first learns the MIMO search geometry from the exact residual metric and then aligns the policy with transmitted-bit accuracy. For soft-output reception, the well-trained hard policy is cloned at the parameter level into every layer of an untied soft-input soft-output iterative detection and decoding (IDD) receiver. This tied-to-untied transfer preserves the learned zero-prior search dynamics while enabling layer- and round-specific specialization under decoder feedback. Within each IDD round, decoder priors tilt candidate generation according to Bayes' rule, and likelihood-weighted terminal hypotheses produce posterior and extrinsic log-likelihood ratios for LDPC decoding. A multi-stage training strategy further stabilizes the hard-to-soft transfer by progressively exposing the receiver to synthetic and in-loop decoder-generated priors.
cs.AI / 76 / 2608.14104
Rewrite Once, Validate Anywhere: Producing OWL-Aware SHACL Constraints (Extended Version)
Abstract
The Shapes Constraint Language (SHACL) is a W3C recommendation to express syntactic constraints, called shapes, on RDF graphs. SHACL validators are used to test whether a given graph adheres to such a shape. However, RDF graphs often come with OWL ontologies, whose implicit knowledge needs to be taken into account. This is classically handled by first applying reasoning and then performing the constraint checking on the results, often using different technologies which makes the process inefficient and vulnerable for mistakes. To overcome this, we propose to internalise the OWL axioms in the SHACL constraints; we construct a rewriting which takes as input both shapes and an OWL EL$^-$ ontology -- a fragment of OWL EL restricting the usage of existential restrictions -- and produces SHACL constraints. This output can then be evaluated by any validator supporting SHACL core regardless of its reasoning support, while yielding the same results as the traditional approach. The implementation of our translation is evaluated both against applying state-of-the-art reasoners and validators consecutively, as against validators with built-in reasoning support. For our benchmark, we show that our approach is in general more efficient in finding violations compared to the sequential approach, thus providing a powerful tool which simplifies combining reasoning with validation.
cs.AI / 77 / 2608.14130
AlignFace: Human-Aligned Face Similarity Metric with Interpretable Concept Relations
Abstract
Computer vision models for generated facial content, such as face editing and privacy protection, increasingly affect people, requiring similarity metrics that serve as faithful proxies for human perception. While perceptual evaluation has progressed from signal-based heuristics to representation-based metrics, current approaches are limited to behavioral modeling without cognitive alignment. They rely on implicit and spurious relations while assuming a universal observer, failing to account for inherent variations across diverse human populations. This leads to inaccurate evaluative models of stakeholders and misleading guidance for generative model debugging. Rather than treating perception as a black box, we leverage scientific findings from cognitive psychology of human face similarity perception: dependence on facial featural and configural attributes, nonlinear psychophysical response scaling, and own-group biases. We introduce the FACETS dataset and propose AlignFace, an interpretable, human-aligned, face similarity metric that encodes these cognitive principles through ante-hoc modeling. It employs visual-language modeling (VLM) to encode paired face images and text-based attributes, gated cross-attention (CA) to extract attribute-specific facial difference representations, concept bottleneck modeling (CBM) to constrain reasoning via interpretable face attributes, and neural generalized additive model (GAM) to model their nonlinear influence. Experiments found AlignFace significantly improves alignment with human subpopulation perceptions compared to baseline metrics, including recent domain-free learned perceptual metrics. By bridging learned representations and human cognitive processes, this work enables more transparent and aligned perceptual evaluation metrics for face images.
cs.AI / 78 / 2608.14028
AdvDex: Learning Dexterous Manipulation from Human Demonstrations via Joint-Aligned Actions and Adversarial Learning
Abstract
Dexterous manipulation is a fundamental capability for embodied intelligence, but scaling it remains difficult because robot demonstrations are expensive to collect and action spaces vary across embodiments. Policies trained on heterogeneous data can also entangle task-relevant visual cues with embodiment-specific appearance, limiting cross-embodiment generalization. We present AdvDex, a unified Vision-Language-Action framework for learning dexterous manipulation from human and robot demonstrations. First, we introduce OmniShare, a large-scale multimodal dataset of human manipulation demonstrations that provides high-quality kinematic supervision and tactile measurements while reducing reliance on robot teleoperation. Second, we propose the Joint-Aligned Action Space (JAAS), a canonical action representation comprising an $\mathrm{SE}(3)$ wrist pose and 15 finger joints, thereby functionally aligning human hands, dexterous robot hands, and parallel grippers. Finally, we use domain-adversarial learning to reduce embodiment-specific information in the learned visual representation. Experiments on hand-action prediction and real-world dexterous manipulation show consistent improvements over baselines, effective zero-shot human-to-robot skill transfer, generalization to unseen objects and environments, and data-efficient few-shot adaptation.
cs.AI / 79 / 2608.14047
Evolve Vision-Language-Action Model into an Agent with On-the-fly Tool-use
Abstract
This paper integrates end-to-end Visual-Language-Action (VLA) models with agentic tool-use to propose Agentic Robot with Tool-use (ART). ART is a tool-injection framework that tunes any VLA model to leverage off-the-shelf tool modules for low-level vision, high-level affordance, and embodiment enhancement. Compared to vanilla VLA models with a whole continuous action solution space, ART reduces the complexity of the action solution space through tool-use, which not only improves generalizability across different tasks but also reduces data dependency. To demonstrate the advantages (high generalizability and low data dependency) of this framework, we first built a dataset of 30K tool-use trajectories and action demonstrations, which is much smaller than those used by baseline methods. We then designed a training regimen for long-trajectory tool-use reasoning in challenging environments. Experiments show that ART achieves a 20% higher success rate than mainstream baselines on simulation and real-world tasks, such as pick-and-place in the dark at novel viewpoints. Empirical results highlight the benefits of an agent-based approach: modular tool utilization enables more efficient training, lightweight deployment, and scalable integration of new tools. This design fosters robustness, adaptability, and extensibility, paving the way for the practical deployment of VLA systems in complex real-world scenarios.
cs.AI / 80 / 2608.14379
Reflex: Enabling Fast and Predictive Vision-Language-Action Models for Reaction-Critical Manipulation
Abstract
Vision-Language-Action (VLA) models have recently achieved promising performance in robotic manipulation. However, existing benchmarks mainly evaluate generalization on static manipulation tasks and largely overlook dynamic interaction scenarios. To address this gap, we present ReflexBench, a benchmark for reaction-critical manipulation. ReflexBench contains six dynamic tasks and introduces an evaluation framework that decouples simulator stepping from robot control while supporting configurable latency under synchronous and asynchronous inference. Building upon ReflexBench, we propose ReflexVLA, an efficient VLA model designed for reaction-critical manipulation without large-scale robot-data pretraining. ReflexVLA enhances temporal reasoning through latent future prediction and multi-frame temporal fusion within the vision backbone, while reducing deployment latency through batched visual encoding and CUDA Graph replay. Experiments show that ReflexVLA consistently improves dynamic manipulation performance while maintaining competitive accuracy on standard static manipulation benchmarks, and real-world experiments further demonstrate its effectiveness under practical deployment conditions. Project website: https://reflexvla.github.io
cs.AI / 81 / 2608.14481
Ensuring Safe Physical AI in Urban Mobility via Hazard-Informed Synthesized Envelopes
Abstract
As heterogeneous robotic systems deploy across diverse urban zones, maintaining safety amid complex human-robot interactions remains a critical challenge. We present a unified framework that bridges systematic hazard analysis and runtime enforcement using hazard-informed safety envelopes. Rather than treating safety as a static constraint isolated within individual software modules, we introduce a cross-layer safety transformation process spanning symbolic, spatial, and dynamic world models. We show how this representation naturally interfaces with physical AI runtime harnesses to guarantee safe urban mobility.
cs.AI / 82 / 2608.14287
Acoustic UAV Detection in Battlefield Scenarios: Handling Noise, Domain Shift, and Weak Labels
Abstract
Passive acoustic sensing offers a critical, cost-efficient, and, crucially, passive alternative for detecting small unmanned aerial vehicles. However, the practical deployment of acoustic systems is discouraged by extreme environmental noise and sensor-induced domain shift caused by heterogeneous hardware. This paper addresses these challenges by introducing a robust framework optimized for real-world battlefield conditions. We propose the integration of Per-Channel Energy Normalization (PCEN) and attention-based pooling to enhance feature extraction under low signal-to-noise ratio scenarios. We further propose a domain-aware training strategy that leverages auxiliary classes and multi-microphone data to mitigate cross-domain performance degradation. Evaluated on a unique dataset of combat-zone recordings from the Ukrainian frontlines, our approach significantly outperforms existing baselines, increasing the F1 score from 55.4% to 78.6%. This paper was originally presented at the International Conference on Military Communication and Information Systems (ICMCIS), organized by the Information Systems Technology (IST) Scientific and Technical Committee, IST-224-RSY - the ICMCIS, held in Bath, United Kingdom, 12-13 May 2026.
cs.AI / 83 / 2608.14491
Optimal Scheduling of Road Maintenance Jobs Considering Impact on Traffic Flows
Abstract
Network-level maintenance planning requires repeated evaluations of equilibrium traffic flows under road capacity reductions. While equilibrium traffic assignment models are well established, their repeated solution quickly becomes computationally prohibitive and challenging to embed within maintenance scheduling problems. This paper investigates data-driven surrogate models that approximate equilibrium arc flows directly from origin-destination demand, using optimization-based equilibrium solutions as ground truth. A real-world case study based on traffic data from the Newark, New Jersey area demonstrates the effectiveness of the proposed approach as a scalable building block for future maintenance scheduling frameworks.
cs.AI / 84 / 2608.14396
AI-Assisted Discovery and Construction of a Counterexample to the Convergence of Three-Block ADMM with the Identity Matrix as its Third Constraint Block
Abstract
The alternating direction method of multipliers (ADMM), as a landmark algorithm, has attracted tremendous research attention and extensive practical applications over the past two decades. It is well known that, although the two-block ADMM enjoys well-established theoretical convergence guarantees, its direct extension to the three-block case may fail to converge, as demonstrated by existing counterexamples [5]. However, to the best of our knowledge, the case in which the third constraint block is the identity remains unresolved: the existing literature gives neither a general convergence proof nor a counterexample for this subclass. In this paper, we give a negative answer: direct three-block ADMM may fail even when the first two blocks are strongly convex quadratics. Using Codex with GPT-5.6 Sol, we construct an explicit rational counterexample candidate and verify it along a piecewise-affine reduction path; exact checks show that direct three-block ADMM on this instance produces a bounded nonconvergent orbit of period 66. Within the same Codex workflow, we further guide a study of multiplier relaxation and clarify when convergence can be restored at the fixed-instance and class levels: a problem-dependent small dual step can restore convergence, whereas no positive relative step works uniformly over the whole class. Furthermore, we also test the recent Kimi Code with Kimi K3 model without the Codex candidate or project-specific route guidance; along a different path it produces an exact locally attracting period-23 certificate, convertible to an equivalent all-identity instance. The comparison suggests that different research-harness configurations can shape the mathematical objects explored and the certificates pursued.
cs.AI / 85 / 2608.14076
Reaction-Transformation-Aware Flow Matching for Generalizable Transition State Generation
Abstract
Transition-state (TS) structures define the energetic barriers and mechanistic pathways of elementary chemical reactions, yet their identification remains computationally demanding because conventional saddle-point searches require expensive quantum-mechanical calculations. Recent machine-learning approaches have accelerated TS generation by predicting structures from reaction endpoint information, but they primarily learn geometric correspondence between endpoints and TSs, leaving the structural transformations underlying elementary reactions implicitly represented. To address this limitation, we introduce TransTS, a reaction-transformation-aware framework for generalizable TS generation from atom-mapped reactant-product pairs. TransTS explicitly learns atom-level structural transformations between reaction endpoints and integrates them with a unified atom-aligned geometric representation of reactants, TSs and products, enabling reaction-aware equivariant generation of TS geometries. TransTS is designed to provide reliable TS initial guesses for subsequent quantum-chemical refinement, where generated structures are evaluated not only by geometric similarity but also by their ability to converge to validated saddle points and recover the intended reaction pathways. Across IID and zero-shot OOD benchmarks, TransTS demonstrates improved TS initialization quality, with particularly strong generalization to unseen reaction distributions. On the challenging GDB-10-rxn and GDB-17-rxn OOD benchmarks, TransTS generates TS candidates that more frequently converge to validated saddle points and recover the intended elementary reactions after refinement than existing approaches under the same training regime. Scaling reaction coverage and model capacity further improves both geometric fidelity and refinement outcomes.
cs.AI / 86 / 2608.13749
Data-driven techniques for translational neuroscience and personalized neuro-health
Abstract
Neurodegenexrative diseases such as Alzheimer's disease and Parkinson's disease are diagnosed most reliably only after substantial, often irreversible, neuronal loss has already occurred, creating an urgent need for quantitative tools that can detect subtle, early, and individual-specific brain changes from neuroimaging data. This review surveys a broad and rapidly evolving toolkit of data-driven techniques for translational neuroscience and personalized neuro-health, organized around four complementary methodological pillars. Throughout, we emphasize how these methodologically diverse approaches converge on a common translational goal: personalized, mechanistically grounded, and clinically actionable models of individual brain health, and we close by discussing the principal open statistical, computational, and clinical challenges that remain.
机器学习 (cs.LG)
73
cs.LG / 1 / 2608.13826
SPEAR: Structure Property Explainability with Attention Regularization
Abstract
Machine learning is increasingly used to learn structure property relationships from spectroscopic and diffraction data, yet its adoption in materials discovery is often limited by poor interpretability of model predictions. Although attention mechanisms are frequently treated as inherently explainable, unregularized attention can yield unstable, fragmented, or intensity driven attribution patterns that obscure the physical origin of these relationships. Here we introduce SPEAR (Structure Property Explainability with Attention Regularization), a framework that constrains attention distributions during training to improve their stability, selectivity, and physical interpretability. SPEAR augments attention based regression with a learnable temperature that controls attention concentration and a smoothness penalty that enforces coherence across neighboring spectral positions, treating attention as a learnable explanatory object rather than a post hoc visualization. Using synthetic spectral benchmarks with known generative structure, we show that attention regularization produces smooth, contiguous attribution profiles aligned with causal features while preserving predictive accuracy. Applied to experimental X ray diffraction data from a combinatorial rare earth zirconate thin film library, the regularized model selectively emphasizes physically relevant diffraction features and decouples feature importance from raw peak intensity. The reflection it identified prompted a reassessment of our earlier structural analysis, revealing a correlation between the 220 peak position, the tetragonal distortion that accommodates cation size disorder, and the local thermal conductivity. Attention regularization therefore provides a principled training constraint for explainable structure property regression, yielding mechanistically meaningful explanations without sacrificing predictive performance.
cs.LG / 2 / 2608.13660
What to Preserve, Where to Adapt: A Depth-Wise Analysis of Forgetting in Continual Gynecological Image Segmentation
Abstract
Medical image segmentation models are typically trained under the assumption that all data are available simultaneously. However, in clinical practice, datasets often arrive sequentially, requiring models to adapt continuously to evolving data distributions. We study this problem in gynecological image segmentation, where substantial heterogeneity across imaging modalities, anatomical structures, and annotation protocols creates a particularly challenging continual learning setting. Under these large distribution shifts, existing continual learning methods struggle to preserve previously learned knowledge, leading to catastrophic forgetting. To better understand forgetting in this setting, we investigate how different encoder--decoder regions influence segmentation performance and forgetting during continual gynecological segmentation. Through block-wise ablation analysis, we observe that ablating early encoder and late decoder regions results in the largest performance degradation, indicating that segmentation performance depends unevenly across the network hierarchy. Using controlled adaptation experiments, we further show that forgetting remains limited when updates are restricted to bottleneck-adjacent regions, but increases sharply once shallower encoders and decoders become trainable, even when only a small subset of parameters is updated. These findings suggest that forgetting in the encoder-decoder architecture is strongly influenced by where updates occur across network depth during continual learning. Full code and analysis pipelines will be made publicly available upon acceptance.
cs.LG / 3 / 2608.14293
Conditional Neural Optimal Transport for Predicting Cellular Phenotypes from Molecular Structure
Abstract
High-content microscopy enables systematic profiling of cellular responses to chemical perturbations, but the scale of the chemical space makes exhaustive phenotypic characterization experimentally infeasible. This motivates computational models that can predict image-derived phenotypes without acquiring the corresponding treated cells. We formulate molecule-induced phenotype prediction as an inductive conditional transport problem in image representation space. Given a negative-control phenotype and the structure of a molecule, we aim to predict the phenotype induced by the corresponding molecule. We first evaluate classical optimal transport baselines and show that static couplings do not yield useful predictions on large-scale phenotypic image datasets. We then introduce a molecule-conditioned Neural Optimal Transport (NOT) model with a Monge-Gap regularization training objective that learns to transport negative-control unperturbed phenotypes toward perturbed phenotypes using molecular structure as conditioning information. NOT recovers molecule-specific phenotypic effects while reducing microscopy-associated technical variation, thereby facilitating comparisons across experimental batches. On unseen active molecules, the model outperforms baseline approaches, demonstrating that chemically conditioned transport can generalize beyond the molecules observed during training. We identified the molecular encoder as the main limitation to this generalization, while transport in a compressed representation space improves performance and scalability. These results establish NOT as a promising framework for predicting cellular phenotypes from molecular structure and negative-control phenotypes, while highlighting the development of more informative molecular representations as a key direction for improving out-of-distribution performance.
cs.LG / 4 / 2608.14435
Style or Signature? Artist-Disjoint Evaluation of Style Classification in Frozen Vision Embeddings
Abstract
Frozen image embeddings from models such as CLIP are increasingly used to classify paintings by art-historical style, with high reported accuracy. We ask whether this accuracy reflects an understanding of style or the recognition of individual artists. Standard evaluation uses random splits in which works by the same artist appear on both sides, so a classifier can succeed by recognising the painter rather than the movement. We re-evaluate style classification under an artist-disjoint protocol, holding out every artist in turn so that no work is ever classified using other works by its own painter. On a balanced dataset of 320 paintings across four twentieth-century movements, 5-NN style accuracy falls from 0.87 to 0.77 under this protocol, and the drop is sharply uneven. Impressionism and Cubism barely move, while Surrealism falls twenty points. The pattern holds across four image encoders, including a vision-only self-supervised model, which places the effect in visual structure rather than language. Where an encoder captures genuine shared form, individual artists are barely recognisable yet style is robust, while Surrealism shows the opposite. We argue that artist-disjoint evaluation is necessary to measure stylistic understanding in frozen embeddings.
cs.LG / 5 / 2608.13810
What preferences can - and cannot - predict in multi-agent online learning
Abstract
We examine the interplay between ordinal, preference-based solution concepts in games and the long-run behavior of game dynamics, asking in particular to what extent the combinatorial data of a game -- its preference graph -- determine the outcomes of no-regret learning dynamics -- such as follow-the-regularized-leader (FTRL). In one direction, we show that the skeleton of every dynamically stable set (i.e. the set of pure profiles it contains) must also be preferentially stable, that is, it must be closed under profitable deviations. We then ask the converse question: when do preferences determine the long-run behavior of the players' learning dynamics? We begin by showing that preferences characterize asymptotic stability in the case of subgames -- i.e. subsets of pure profiles obtained by restricting players' action sets. Beyond this case however, the equivalence between dynamic and preferential stability collapses: concretely, we construct a three-player game with a preferentially stable set whose span is dynamically unstable, showing in this way that preferences do not suffice as a criterion of dynamic stability. We then bridge this gap via the notion of resilience under aggregate deviations, an easy-to-check payoff-based condition that guarantees asymptotic stability of arbitrary spans of pure strategies.
cs.LG / 6 / 2608.13652
Contrastive Learning for Interpretable Anomaly Detection at Collider Experiments
Abstract
Generic event-level anomaly detection for collider physics has two recurring problems: anomaly scores are hard to interpret, and they correlate strongly with energy scale and object multiplicity. We present Organized Representation via Contrastive learning for Anomaly detection (ORCA), a two-stage framework that first learns an embedding space via supervised contrastive learning across a diverse set of physics processes, then runs a standard autoencoder in that space to generate event-level anomaly scores. On a simulated dataset consistent with conditions at the High-Luminosity Large Hadron Collider, ORCA delivers significant gains in both breadth and depth of sensitivity to new physics signals with respect to a baseline autoencoder architecture. Beyond improved sensitivity, the contrastive embedding makes the anomalous sample interpretable: because known processes occupy distinct regions of the space, a maximum-likelihood template fit to the embedding distributions can attribute events in an anomalous sample to template physics processes with quantified uncertainties. We demonstrate that the fit accurately recovers injected signal yields, including for signals excluded from the training of the embedding, and characterizes signals absent from the template library through the known processes they most resemble. These results establish ORCA as a route to interpretable anomaly detection-based searches at colliders, where the embedding geometry carries higher dimensional physics information compared to standard one-dimensional output fits, enhancing downstream statistical analysis.
cs.LG / 7 / 2608.13668
The Query Knows What to Forget: A Second Erase Direction for Linear Attention
Abstract
Linear attention keeps a state of fixed size. At long context, many stored items share this state, and interference between them degrades retrieval. Gated DeltaNet-2 (GDN-2), like every delta-rule model before it, derives its erase vector from the key of the current token. However, the interference in its reads is measured through the query, and the erase step cannot reach it. We introduce the Query-derived Erase Direction (QED). QED adds a second erase direction derived from the query and orthogonal to the key. In the fast-weight view, a key-directed delta edit cannot change the key-orthogonal part of a read. It uses the editable part to cancel old-state content measured along the query. It also improves retrieval at every length past the training window, and it about doubles the usable context length on S-NIAH-1.
cs.LG / 8 / 2608.13675
From BERT to Frontier Agents: Eight Years of Language-Model Progress, the Collapse of the Capability-Cost Curve, and the Rise of Task-Targeted Models
Abstract
Between October 2018 and July 2026 AI models progressed from simple systems like BERT to massive agents that solve complex math and write software. The ability to resolve real coding issues improved by nearly six times per year since late 2024. During this time costs dropped sharply with OpenAIs budget model GPT 5 point 6 Luna matching flagship capabilities for just one to six dollars per million tokens beating older versions at a fraction of the price. Top performance is now split across specialized models as Claude Opus 5 leads in frontend coding Claude Fable 5 excels at repository level coding and GPT 5 point 6 Sol dominates terminal tasks. In a grade school math test using the Qwen 2 point 5 model basic methods solved 58 of 100 problems while advanced sampling solved up to 79. A confidence ranking tool correctly identified 47 right answers in its top 50 choices proving highly useful for sorting tasks with all research materials made fully public.
cs.LG / 9 / 2608.13676
EEG-PRISM: Physiologically-Grounded Interpretability of Predictions by EEG Foundation Models
Abstract
Objective: Foundation models represent the next advancement in AI for EEG analysis; however current explainable AI techniques provide attribution scores in the time-channel input space, which is mismatched to clinical intuition about EEG. Thus, there is a critical need for a universal method that can extend the interpretability of any foundation model to alternative and physiologically relevant domains without modifying or retraining the underlying model. Methods: EEG-PRISM leverages linear transformations and established backpropagation rules to map time-channel attribution scores into alternative domains. We derive mappings to the frequency domain via an invertible DFT and to the source domain via an approximately invertible EEG generative model. We evaluate EEG-PRISM in simulated and real data, assessing recovery of ground-truth phenomena across domains with five foundation models and four AI explainers. Results: In simulation, EEG-PRISM achieves near-perfect spectral recovery and 69.2% spatial accuracy. In epilepsy, EEG-PRISM correctly determines that delta-theta activity is most salient and correctly localizes the seizure onset region with 50% accuracy. In autism, EEG-PRISM localizes the predictive delta-alpha biomarkers to frontal and temporal regions, consistent with prior work. Conclusion: EEG-PRISM is a theoretically-grounded post-hoc attribution method with accurate mapping into the spectral and spatial domains. It supports window-level analysis of transient events (e.g., seizures) and group-level identification of clinically relevant biomarkers (e.g., autism), thus advancing interpretable EEG foundation models. Significance: This work enables physiologically-grounded interpretation of EEG foundation models and supports clinically relevant insights such as event localization and biomarker identification.
cs.LG / 10 / 2608.13702
SAGE: Surrogate-gradient Adaptation via Attention-Guided Entropy for Spiking Transformers
Abstract
Spiking neural networks (SNNs) offer an energy-efficient alternative to conventional deep neural networks by exploiting sparse event-driven computation, but their training remains challenging because the non-differentiable spike function requires surrogate gradients whose fixed shape may be suboptimal across layers and training stages. In this work, we introduce SAGE, an uncertainty-modulated surrogate-gradient mechanism for Transformer-based SNNs. SAGE estimates block-level uncertainty from normalized self-attention entropy and uses this signal to adapt the surrogate-gradient slope during training while leaving the inference model unchanged. By modulating only the training-time surrogate parameter, the proposed method preserves the original architecture and deployment cost while improving optimization flexibility. Experiments on CIFAR-10/100 demonstrate that SAGE achieves improved accuracy over fixed-surrogate baselines, with results up to 1-2\% consistent gains across multiple simulation time steps. These results highlight the potential of attention-derived uncertainty as a lightweight training signal for adaptive surrogate-gradient learning in transformer-based SNNs.
cs.LG / 11 / 2608.13721
Capacity-Dependent Effects of Data Selection for Reasoning
Abstract
In reasoning supervised fine-tuning, candidate responses for the same instruction can differ substantially in how well they match the student's current distribution. Recent likelihood-based response selection methods suggest that responses closer to the student distribution provide more effective supervision, motivating the hypothesis that high-likelihood responses may generally be preferable for fine-tuning. In this paper, we revisit this intuition and show that the value of likelihood-based data selection depends critically on model capacity and training duration. Through controlled experiments on mathematical reasoning, using students ranging from 1.5B to 8B parameters and supervision generated by stronger teacher models, we observe a clear \emph{capacity-dependent} ``{\color{SMALLCOLOR}\textbf{Fast-Fit}} / {\color{LARGECOLOR}\textbf{Slow-Gain}}'' pattern. High-likelihood data provides faster and more stable early improvements, especially for smaller models, but low-likelihood data becomes increasingly beneficial for larger models when training is allowed to continue longer. To explain this phenomenon, we analyze learning dynamics, showing that small models often fail to absorb low-likelihood supervision and instead fall into shallow or repetitive behaviors, while larger models are better able to move toward the teacher distribution under such data. We further provide a capacity-constrained theoretical view of distillation that clarifies how data difficulty, data span, and student capacity jointly govern transfer. Overall, our findings show that effective data selection for reasoning should be aware of model capacity and computing budget rather than based on a single universal preference for high-likelihood supervision.
cs.LG / 12 / 2608.13756
The Integer Alibi: Localizing Cross-Kernel Divergence in INT8-Quantized LLM Inference
Abstract
Two GPU kernels implementing the same scaled INT8 GEMM interface are usually treated as interchangeable. We test that assumption: holding the checkpoint, prompts, hardware, inference engine, decoding, and quantization configuration fixed, we swap only the INT8 linear kernel (CUTLASS versus Triton) inside vLLM. At 1.7B each arm reproduces itself bit-for-bit across cold restarts, yet the arms agree on no sequence in any end-to-end comparison we ran (0/8, 0/16, and 0/64). What makes this more than a benchmark discrepancy is an integer alibi: for shared INT8 operands under a verified no-overflow bound, the INT32 dot product is exact and order-independent, so the accumulator cannot be the source of any difference. Feeding both kernels identical operands from every linear layer of Qwen3-1.7B and 8B (196 and 252 layers), we find bit-identical outputs under power-of-two scales, confirming a pinned prediction list 196/196 and 252/252 (pre-registered at 1.7B, pinned but not blind at 8B), and observed differences of at most one bfloat16 spacing under the checkpoints' real scales. This localizes the divergence to scale application and output rounding after the exact accumulator. Applied as a probe checkpoint, the same intervention restores end-to-end bitwise agreement (8/8 and 16/16 sequences). Cross-implementation FP8 GEMM shows a different signature: both the prevalence and the magnitude of differences grow with reduction depth, while the INT8 fraction stays at parts per million and within one spacing over a 64x range of K. Teacher-forced replay ties layers to tokens: flips concentrate at small logit margins, which predict flip risk with ROC-AUC 0.94 on 16,384 positions. We will release the pre-registration, per-layer predictions, manifests with kernel-selection evidence, and a conformance procedure that turns these controls into a concrete check for kernel interchangeability.
cs.LG / 13 / 2608.13773
CutClean: Neural Network Pruning for Privacy-Preserving Inference
Abstract
Neural networks are increasingly deployed in high-stakes applications with growing privacy leakage concerns. We show that this privacy leakage can occur even in the absence of representation imbalances that lead to traditional dataset biases. This poses significant privacy risks when deploying models that process sensitive attributes. In this context, we propose CutClean, a privacy-aware pruning method that allows to reduce privacy information flow through the network, while increasing its sparsity. Our approach employs auxiliary linear privacy heads placed at each network's block to quantify information leakage, and further applies increasing levels of sparsity to remove the private attribute leakage, measured in terms of the accuracy of the privacy head attached to the last block. Experiments on synthetic and real-world datasets demonstrate that our approach effectively minimizes private information flow while achieving high sparsity rates and preserving classification target accuracy.
cs.LG / 14 / 2608.13790
PPAPlace: Differentiable Cross-Stage Objectives for Chip Placement Optimization
Abstract
Macro placement significantly affects a chip's post-route performance, power, and area (PPA). Most placement methods optimize half-perimeter wirelength (HPWL) as the primary objective. However, recent benchmarking shows a near-zero correlation between HPWL and post-route timing metrics such as the worst negative slack (WNS) and total negative slack (TNS). As a result, all six evaluated artificial intelligence (AI) placers degraded PPA relative to the hierarchical baseline. Recent efforts have tried to train cross-stage predictors to close this gap. However, existing methods focus on macro-only representations and use pre-route metrics as training labels. A label fidelity study of ten circuits at four design flow stages reveals that HPWL and pre-route timing poorly reflect final post-route timing rankings. In contrast, post-global-routing achieves the best balance between final timing fidelity and label generation cost-effectiveness. Based on this finding, PPAPlace is a timing-driven differentiable surrogate predicting post-route PPA from macro and standard-cell placements. The surrogate is a dual-stream predictor that combines graph attention over the chip netlist with spatial convolution over the placement grid. It is trained on post-global-routing labels. The predicted WNS and TNS gradients flow end-to-end back to cell coordinates. PPAPlace exploits these gradients in two ways: as a co-objective injected into an analytical placer's optimization loop (PPAPlace-CoOpt), and as a post-placement refinement step that adjusts macro positions via projected gradient descent (PPAPlace-Refine). On five ChiPBench test circuits excluded from training, PPAPlace improves average WNS and TNS by 22\% and 51\% over the hierarchical baseline while preserving power and routability, using the same predictor without test-circuit retraining. Code is available at https://github.com/ValleyC/PPAPlace.
cs.LG / 15 / 2608.13797
Recent Advances in Deep Learning-Based Drug-Target Binding Affinity Prediction
Abstract
Computational approaches to drug discovery involve multiple sub-problems, and among them, drug-target binding affinity prediction plays an important role. Despite recent advances, accurately predicting binding affinity remains an open research area. The major objective of our paper is to perform a comprehensive review and comparative analysis of recent machine learning methods for drug-target binding affinity prediction, with a focus on identifying strengths, limitations, and research gaps. We review representative recent deep learning approaches that use common benchmark datasets and evaluation metrics, covering a range of neural network architectures and representation strategies. In addition, we analyze seven widely used benchmark datasets and commonly adopted evaluation metrics for drug-target binding affinity prediction. Our analysis indicates that although many methods report strong performance on standard benchmarks, their effectiveness is often influenced by dataset bias and limited evaluation settings. Furthermore, most methods exhibit reduced performance in cold-start scenarios, highlighting challenges in generalization. We identify several limitations of current approaches, including dataset imbalance, the lack of standardized evaluation, limited real-world applicability, and challenges in cold-start scenarios. We also discuss future research directions, including better dataset design, more robust evaluation methods, improved handling of cold-start problems, and the integration of multimodal representations.
cs.LG / 16 / 2608.13799
Dynamic Multi-Depot Vehicle Routing with Online Requests: Event-Driven Transformer--DRL and Rolling-Horizon Benchmarking
Abstract
This paper presents an event-driven learning and benchmarking framework for the Dynamic Multi-Depot Vehicle Routing Problem with progressively revealed requests and evolving vehicle states. Masked MLP and Transformer policies are trained through behavior cloning and proximal policy optimization. Deterministic feasibility masking prevents invalid vehicle--request assignments, while fixed-prefix/flexible-suffix route commitments protect completed, active, and near-term decisions and separately measure vehicle reassignment and resequencing. The learned policies are compared with dynamic insertion heuristics and time-limited rolling-horizon optimization. In a 20-scenario policy benchmark, all methods completed every request without invalid actions, but nearest feasible achieved the lowest mean objective and outperformed the learned policies in routing quality, waiting time, stability, makespan, and runtime. Across five independent training runs, PPO had little average effect on the MLP and improved the Transformer on average, although with greater seed variability. Under the common protocol, nearest feasible achieved the lowest combined objective and route disruption, whereas rolling horizon achieved the lowest waiting times and makespan at substantially higher computational cost. The learned policies retained millisecond-level decisions and transferred to instances with up to 80 requests without retraining, but did not outperform the strongest heuristic. No single method was best across routing efficiency, service responsiveness, stability, and online computation.
cs.LG / 17 / 2608.13800
Stochastic Control Policies for Robust Molecular Transition Path Sampling
Abstract
Transition path sampling (TPS) aims to efficiently generate rare molecular transition trajectories between metastable states and is essential for understanding biomolecular mechanisms. Beyond traditional molecular dynamics (MD)-based sampling, machine learning has become central to state-of-the-art TPS. One major class of methods learns control forces during explicit MD rollouts. By preserving the underlying molecular dynamics, these methods tend to produce more physically plausible trajectories than endpoint-conditioned generators that construct paths directly. However, rollout-based control methods have been reported to exhibit unstable and strongly seed-dependent performance. We recast rollout-based control as learning a path-space proposal distribution and investigate stochasticity placement as a design choice for improving exploration and optimization robustness. We develop two stochastic policies: FS-TPS, which directly parameterizes a state-dependent Gaussian distribution over the control policy output, and LaS-TPS, which samples a compact latent control variable and decodes it into structured, cross-atom-correlated force variation. We conduct extensive multi-seed experiments on three biomolecular systems of increasing size: alanine dipeptide, chignolin, and BBL, a fast-folding protein. Stochastic policies consistently improve transition success and path quality over deterministic-policy baselines while substantially reducing sensitivity to random initialization.
cs.LG / 18 / 2608.13827
HI-MeshGraphNets: Efficient and Accurate Mesh-based Physics Learning with Hierarchical Multi-scale Graph Neural Networks
Abstract
Machine-learned physical surrogate models have become promising alternatives to mesh-based numerical solvers. Among them, graph neural networks (GNNs) are well suited for representing simulation meshes and learning nodal state evolution through message passing. However, conventional flat message passing becomes inefficient on large, high-fidelity meshes because information propagates only one hop per layer, requiring deep processors for long-range interactions and increasing computational cost, memory usage, and the risk of over-smoothing. To address this limitation, we propose Hierarchical Interpolating MeshGraphNets (HI-MGN), a multiscale extension of MeshGraphNets for efficient long-range communication on unstructured meshes. HI-MGN replaces the flat processor with a hierarchical multiscale processor that coarsens graphs using farthest-point sampling and Voronoi partitioning while preserving the original mesh topology. Message passing on coarse graphs enables information to travel over larger geometric distances with fewer layers, and a learned graph interpolation network reconstructs fine-resolution features. Across three structural and fluid benchmarks, HI-MGN achieves improved accuracy compared with MeshGraphNets and the Bi-Stride Multi-Scale GNN while reducing training time and peak memory usage. The results show that topology-aware hierarchical message passing and learned coarse-to-fine interpolation provide an effective and practical framework for scalable mesh-based physics surrogate modeling.
cs.LG / 19 / 2608.13882
Variation Brownian Kernel Ladders
Abstract
Claims about the benefit of depth depend on the complexity assigned to a representation. We introduce the \emph{Variation Brownian Kernel Ladder} (VBKL), a path-atomic function-space framework that separates nonlinear recursive dictionary construction from linear variation superposition. Starting from linear projections, each atom recursively composes unit-ball profiles from the Brownian reproducing kernel Hilbert space; the full VBKL space is then the signed-measure variation hull of the completed dictionary. We identify each recursive dictionary as a union of Brownian pullback RKHS balls and establish variation-controlled Hölder regularity, compactness and attainment, and strict growth with depth under a local non-degeneracy condition whose trace lies in the support of the input measure. For associated finite lower-support architectures, we derive Rademacher and generalization bounds through Brownian quadratic chaos, signed threshold traces, and VC entropy. We also construct two-stage approximants by discretizing the outer measure and the selected outer Brownian profiles, obtaining an $M^{-1/2}+m^{-1/2}$ error bound, a sharp interpolation constant $\sqrt{A/2}$, and at most $2M$ active outer-profile basis contributions per evaluation. Controlled experiments illustrate the approximation mechanisms and indicate a favorable limited-data accuracy--complexity trade-off.
cs.LG / 20 / 2608.13888
Fashion Outfit Generation via Unified Sequential Composition Models
Abstract
The task of synthesizing stylistically coherent fashion outfits from massive item libraries, known as fashion outfit generation, remains a non-trivial challenge, primarily due to the non-monotonic and implicit nature of aesthetic compatibility, coupled with the exponentially large combinatorial search space. In this paper, we formalize this task as Constrained Ensemble Generation (CEG) and model it as a finite-horizon deterministic Markov Decision Process. To address CEG in fashion, we propose the Unified Sequential Composition Model (USCM), which jointly models set-level compatibility and latent composition intents. Guided by USCM's learned priors, a Latent Expansion Monte Carlo Tree Search (LE-MCTS) mechanism is proposed to handle item retrieval during composition, balancing local aesthetic synergy with global structural balance. Extensive experiments on the Polyvore Outfits dataset, along with zero-shot evaluations on the iFashion and PolyvoreU datasets, demonstrate that our framework achieves state-of-the-art performance across independent human preference evaluations, automated aesthetic proxies, and structural validity metrics for constrained fashion outfit generation.
cs.LG / 21 / 2608.13911
MedMix: Specialization-Consistent Federated Sparse MoEs under Modality Heterogeneity
Abstract
Federated multimodal medical AI faces modality heterogeneity at both the client and sample levels: clients may systematically lack access to specific modality types, while individual records within the same client may contain different partial modality subsets. Sparse Mixture-of-Experts (MoE) architectures are a promising remedy for modality-adaptive computation, but their use in federated learning is fragile under cross-client modality heterogeneity, where locally learned routing policies can diverge across clients and drive experts toward incompatible specializations. Different clients may assign the same observed modality configuration to different experts, or train similarly indexed experts on different missing-modality configurations, causing standard aggregation to misalign or overwrite the expert specialization that sparse MoEs are intended to learn. To address this challenge, we propose MedMix, a semantic-alignment framework for federated multimodal sparse MoEs that coordinates cross-client routing and expert specialization using modality context. At the client side, MedMix uses modality-context-aware routing to guide expert selection using each token's modality identity, position, and incompleteness context. Across clients, it uses consensus-guided routing alignment to construct server-side consensus anchors for shared modality patterns and align local routing distributions across clients. Complementing these routing mechanisms, client-adaptive expert aggregation leverages client-specific modality-pattern prototypes to match and aggregate functionally similar experts across clients. Experiments on real-world multimodal medical datasets show that MedMix achieves the best average F1 across diverse modality heterogeneity and modality incompleteness settings, with especially clear gains under severe heterogeneity.
cs.LG / 22 / 2608.13914
Hybrid Quantum-inspired Kolmogorov-Arnold Networks for Privacy-Aware Federated Biosignal Learning
Abstract
Electrocardiogram (ECG) recordings are sensitive biomedical data, limiting the ability of hospitals and wearable devices to share raw signals for centralized model training. Federated learning addresses this practical privacy constraint by enabling collaborative model training while keeping raw biosignal data at their respective sources. However, federated ECG classification remains challenging due to limited client-side samples, imbalanced arrhythmia labels, and non-independent and identically distributed (non-IID) data across clients. These constraints require classifiers that are both communication-efficient and robust to cross-client distribution shifts. In this work, we evaluate a hybrid quantum-inspired Kolmogorov-Arnold network (HQKAN) against a multilayer perceptron (MLP) for five-class arrhythmia classification on the MIT-BIH dataset and three-class classification on the INCART dataset under federated averaging (FedAvg). Across multiple client configurations, HQKAN improves most aggregate and minority-class metrics while using 37.35% fewer trainable parameters and reducing communication cost by 24.89% on MIT-BIH; on INCART, it achieves corresponding reductions of 44.81% and 36.41%. These results indicate that HQKAN offers a compact, communication-efficient and robust alternative to the MLP baseline for privacy-aware federated learning on biosignal data.
cs.LG / 23 / 2608.13922
High-dimensional nonparametric changepoint detection via low-rank degree-two density projection
Abstract
Detecting distributional changes in high dimension is difficult when neither the pre-change nor post-change density is parametrically specified. We introduce a representation-based approach that retains all degree-at-most-two density information while replacing density estimation by matrix mean estimation. For observations in $[-1,1]^d$, a symmetric feature matrix $H_2(X)\in\R^{(d+1)\times(d+1)}$ is constructed so that $M(f)=\E_f H_2(X)$ is an isometric encoding of the degree-two orthogonal projection of the density. We scan matrix CUSUMs after rank-$r$ truncation, exploiting the low rank of the projected jump rather than sparsity of individual coordinates. The resulting \LRD{} estimator has a tent-shaped population objective and a nonasymptotic operator-norm analysis whose leading stochastic term scales as $\sqrt{rd\log(nd)}$. For multiple changes, we give a seeded narrowest-over-threshold procedure and prove exact recovery by an induction that preserves an isolating interval for every undetected change. A cross-fitted scalar refinement learns the changing low-rank direction on one fold and localizes on the other, attaining $\widetilde O_{\Pp}(κ^{-2})$ error; a matching Le Cam lower bound shows optimality up to logarithms. A geometrically $β$-mixing extension follows from a dependent matrix Bernstein inequality. Experiments with ambient dimension up to $200$, a three-change $d=100$ sequence, and a $128$-feature human-activity benchmark show that the method remains computationally practical and accurately detects pure dependence changes that are invisible to mean CUSUMs.
cs.LG / 24 / 2608.13934
Probabilistic indirect models for undrained shear strength: addressing significant data missing and variability with advanced imputation and machine learning techniques
Abstract
Accurate prediction of undrained shear strength (su) is crucial for geotechnical design, but is often hampered by substantial uncertainty in traditional empirical methods. This study uses the CLAY/10/7490 global database to develop probabilistic indirect models to predict su based on Atterberg limits and piezocone cone penetration (CPTU) measurements. Firstly, the dataset has a high missing data rate and variability. We test three imputation methods - multivariate normal (MN), multiple imputation by chained equations (MICE), and miss forest (MF) - to fill the missing values. To validate their effectiveness, a Probabilistic Extreme Gradient Boosting (PXGB) model is developed, and the imputation methods are evaluated by comparing the PXGB's performance when trained on the imputed datasets against that on the original incomplete data. Secondly, the indirect model is built by integrating a multi-head attention (MHA) mechanism into an artificial neural network (ANN) to enhance information extraction from limited data, which leads to the MHA-based probabilistic neural networks (MHA-PNN) model. The models' performance, alongside a conventional MN-based prediction model, was evaluated using root mean square error (RMSE), coefficient of determination (R2), mean absolute percentage error (MAPE), conditional interval width (wCI), and coverage rate (CR). Results demonstrate that the proposed MN-enhanced MHA-PNN model substantially outperforms other models in both prediction accuracy and uncertainty quantification. These findings highlight the potential of this integrated strategy for building robust probabilistic indirect models in geotechnical applications, particularly when confronted with sparse and incomplete datasets.
cs.LG / 25 / 2608.13961
Polar Code Based Federated Learning: Convergence Analysis and Resource Allocation
Abstract
Federated learning (FL) enables collaborative model training across distributed devices without sharing raw data; however, it faces significant communication bottlenecks and channel impairments in practice. Conventional network layer treatments either idealize the channel as error free or apply equal error protection (EEP) to transmitted model updates, failing to account for the inherently unequal importance of quantization bits within a single local model. To address this limitation, we propose a cross layer polar code based FL scheme that leverages the unequal error protection (UEP) property of polar codes under finite block lengths. Specifically, the proposed design selectively protects more significant quantization bits, thereby mitigating the detrimental effects of channel noise. We further provide a rigorous convergence analysis of the proposed scheme, deriving an upper bound on the convergence gap, which we then jointly optimize over the number of quantization bits and the polar code block length across all training iterations. Experimental results demonstrate that both constant and variable block length configurations of our polar code based scheme consistently achieve substantial performance gains over uncoded and LDPC-based EEP benchmarks, with the advantage becoming increasingly pronounced as the channel quality deteriorating. These findings confirm the efficacy of our cross-layer design in enhancing FL robustness and efficiency under realistic channel conditions.
cs.LG / 26 / 2608.14004
Identifiability and Order-Dimension Limits of In-Context Learning on Partial Orders
Abstract
In-context learning is commonly formalized as inference from examples of a function. Partial orders instead combine transitivity, antisymmetry, and incomparability, so a finite prompt may not determine a queried comparison. We develop a theory of in-context learning on partial orders that separates logical identifiability, prompt teaching cost, structural complexity, and the exact capacity of a formal coordinate-decoder class. A version-space semantics makes background knowledge and open- versus closed-world assumptions explicit. For finite open-world prompts with positive and negative comparisons, we prove an exact completion trichotomy: after taking the reflexive transitive closure of the positive demonstrations, a query is forced true, forced false because every true completion creates a cycle or violates a negative demonstration, or remains genuinely ambiguous. For a known $n$-element universe, we characterize the open-world teaching number as the number of covers plus a blocker-set hitting number, prove that its maximum over all $n$-element posets is $n(n-1)$ and is uniquely attained by the antichain, and identify the blocker term as the exact cost of open-world rather than complete-Hasse semantics. We formalize prompt-dependent $s$-coordinate decoders and use the classical coordinate-order equivalence to obtain an exact representation boundary: dimension at most $s$ is necessary and sufficient, while width at most $s$ is a convenient sufficient condition.
cs.LG / 27 / 2608.14020
When Does More Correct Data Hurt? Insertion-Stability and the Limits of Dimension-Based Theory
Abstract
Adding data known to be correct ought to be safe. Not always. Larsen, Pabbaraju and Shetty model the failure with a monotone adversary, which reads an i.i.d. training sample and may append as many further examples as it likes, provided the target hypothesis labels them all. Mehrotra has since settled the cost, showing that for classes of VC dimension d >= 2 no learner can guarantee expected error better than Theta((d/n)log(en/d)), a logarithmic factor above the clean PAC rate. Because that rate is a worst case over all classes, it says nothing about which classes actually suffer the penalty, and the answer turns on the learner. We call a learner insertion-stable if feeding it more correctly labeled examples can only shrink the region where it errs. Such learners are immune to the adversary, since on any given sample the risk after insertions never exceeds the risk on the clean part alone, however much is added and however cleverly it is chosen. High- probability guarantees carry over unchanged, and because Closure is insertion-stable every intersection-closed class keeps its clean rate of E[Err] <= (21d+34)/n. Immunity is not something the classical dimensions can predict. Two classes can agree on VCdim = Ldim = 2 and still split, one at Theta(1/n) and the other at Theta(log(en)/n), while intervals have unbounded Littlestone dimension and are immune anyway. On Mehrotra's hard class we prove more than the failure of a single algorithm, showing that no monotone permutation-invariant compression scheme of any finite size attains the clean rate. The question is therefore not whether a class is hard, nor whether a learner is good, but whether the two suit each other. Given an insertion-stable learner that is optimal on clean data, correct additions are free, and without one the cost belongs to the class, so changing the learner will not avoid it.
cs.LG / 28 / 2608.14054
Model-agnostic Retrieval-Augmented Extended Forecasting for time series
Abstract
Time series forecasting with pretrained foundation models has demonstrated strong zero-shot capabilities. However, achieving optimal performance on time series with short or negligible historical data in domain-specific applications typically requires adaptation via either fine-tuning or RAG. While fine-tuning is effective, it incurs substantial computational costs. This work explores RAG within univariate time series (Retrieval Augmented Generation) as a more efficient alternative, in particular RAF (Retrieval Augmented Forecasting), and introduces RAEF (Retrieval-Augmented Extended Forecasting), a model-agnostic method built upon RAF. RAEF incorporates key refinements to the retrieval and aggregation mechanisms: (1) direct retrieval in input-space rather than embedding-space, reducing inference overhead, and (2) concatenation-based aggregation that preserves temporal structure instead of averaging. Empirical evaluation across multiple benchmark datasets demonstrates that RAEF outperforms RAF in both accuracy and inference overhead. Furthermore, comprehensive comparisons with zero-shot and fine-tuned foundation models show that RAEF achieves competitive or superior performance to fine-tuning while avoiding its computational burden, establishing it as a practical and scalable approach for domain adaptation in time series forecasting.
cs.LG / 29 / 2608.14096
Resource-Adaptive Primal-Dual Learning for One-Warehouse Multi-Store Systems with Censored Demand
Abstract
The one-warehouse multi-store (OWMS) system is a fundamental inventory network in which a nonreplenishable warehouse allocates shared stock across multiple stores over time. Existing OWMS learning policies are built around a fixed target calibrated to the initial average resource rate, but such a fixed-target architecture cannot re-center after realized sales change the remaining resource available per future period. We develop Resource-Adaptive Primal-Dual Learning, a new learning framework that tracks the primal-dual resolving path with censored demand as the remaining-resource state evolves. In each period, the current resource rate indexes the target store allocations and dual variable, while censored sales provide gradient estimates for updating both. The analysis combines expected-sales geometry with a moving-target argument to yield logarithmic expected regret, improving on the state-of-the-art square-root-order guarantees of existing OWMS learning policies. The underlying design and analytical ideas may inform other online learning problems with depleting shared resources. Numerical experiments further demonstrate good finite-horizon performance of a practical variant across different horizon lengths and inventory regimes.
cs.LG / 30 / 2608.14102
Sequence prediction under a lying oracle
Abstract
We consider the problem of sequential prediction of an $m$-ary sequence, where at each epoch, (i) the environment selects an outcome from an $m$-ary alphabet, (ii) the learner selects a probability distribution over the same alphabet (unaware of the outcome generated by the environment), and finally, (iii) the learner incurs a cost that depends on the probability assigned to the outcome. The cost function we consider captures the complexity of predicting the outcome generated by the environment, in a scenario where the aforementioned prediction is performed via comparative queries to a lying oracle. We consider both stochastic and adversarial environments, propose algorithms for both settings, and establish logarithmic upper bounds on their regret.
cs.LG / 31 / 2608.14106
Forecast Collapse in Time-Series Foundation Models
Abstract
When forecasting hourly returns for 1,000 US equities, we observe an unexpected phenomenon: predictions become nearly flat and show poor stock ranking, as measured by cross-sectional correlation. We call this forecast collapse. Surprisingly, the phenomenon largely disappears when forecasting trading volume under the same setting. We investigate forecast collapse across time-series foundation models (TSFMs), twelve deep-learning forecasting models, and 97 public benchmark configurations, and find that it is closely tied to target predictability. We identify two distinct reasons behind it: low predictability limits the amplitude of calibrated point forecasts, while per-series objectives leave cross-series structure unidentified. These findings reveal a calibration-ranking tradeoff: optimizing squared error leads to flat predictions, whereas directly optimizing cross-sectional correlation improves ranking but can inflate forecast amplitude by more than an order of magnitude. To address this tradeoff, we introduce CalibRank, a simple objective that balances calibration and ranking. On Finance1K, CalibRank nearly triples cross-sectional correlation while keeping amplitude close to the target, and improves correlation on all tested models. Our results reveal a blind spot in conventional time-series evaluation: per-series metrics can hide failures in cross-series structure needed by downstream decisions.
cs.LG / 32 / 2608.14114
Learning to Run Power Networks: Effective AlphaZero-inspired Topological Control
Abstract
As the integration of volatile renewable energy sources increases the strain on modern power grids, the use of Reinforcement Learning (RL) for autonomous topological reconfiguration has emerged as a promising research field to keep strained grids stable and operational. Compared to traditional redispatching measures, topological actions offer a cheaper and more cost-effective way to manage grid congestion. However, their implementation is hindered by a vast combinatorial action space and strict operational constraints. This paper investigates the effectiveness of model-based AlphaZero-inspired approaches that utilize Monte Carlo Tree Search (MCTS) for proactive grid management. We systematically evaluate how reward functions, observation density, and search guidance influence an agent's survivability. Our results demonstrate that the optimized AlphaZero approach achieves a peak survivability of 98.43%, significantly outperforming the proximal policy optimization (PPO) variant. We find that conducting the MCTS without guidance from a prior learned policy or value function can enhance training efficiency, and that a straightforward binary survival reward provides more effective search guidance than complex, multi-objective functions. Our findings demonstrate that while AlphaZero is a powerful framework for topological control, pure reinforcement learning is not sufficient; rather, an effective and reliable system requires a 'minimalist' integration of domain-specific heuristics, binary rewards, and a restricted observation space of line loads.
cs.LG / 33 / 2608.14120
From Fixed Grids to Moving Particles:A Transferable Latent Operator for Fluid Dynamics
Abstract
Lagrangian modeling is vital to fluid dynamics, as it characterizes particle transport and complements the Eulerian description.However, Lagrangian trajectories are less commonly available than Eulerian fields, while most neural operators are trained and evaluated primarily in the Eulerian representation. This mismatch motivates a new learning problem: can a model trained solely on Eulerian observations generalize zero-shot from Eulerian field prediction to Lagrangian particle rollout, without Lagrangian supervision or task-specific adaptation? To address this problem, we propose the Transferable Latent Operator (TLO), which learns a unified flow representation shared by Eulerian field prediction and Lagrangian particle rollout. TLO decouples latent flow evolution from coordinate-dependent decoding: querying the evolving latent representation at fixed spatial coordinates yields Eulerian fields, whereas querying velocities at particle positions and recursively updating these positions enables Lagrangian rollout. Across five fluid-dynamics benchmarks, TLO consistently outperforms existing neural operators in both Eulerian field prediction and zero-shot Lagrangian rollout, with further gains from limited Lagrangian fine-tuning.
cs.LG / 34 / 2608.14121
Overcoming Shortcut Learning in Graph Neural Networks through Active Explanation Guidance
Abstract
Graph Neural Networks (GNNs) can solve prediction tasks by unintentionally exploiting shortcuts---that is, edges, nodes, and features that correlate with but are not causal for the prediction---which compromise their reliability in out-of-distribution tasks. We introduce XIGL, an architecture-agnostic human-in-the-loop strategy for removing such shortcuts from GNNs. Our key insight is twofold. On the one hand, reliance on shortcuts can be detected by inspecting GNN explanations. On the other hand, once made aware of such shortcuts, sufficiently expert users can provide tailored corrective feedback, which helps deconfound the model. XIGL supports any query strategy; however, since corrective feedback can be expensive to acquire, we develop an active learning strategy for prioritizing explanations that are more likely to display shortcut behavior, lowering annotation and cognitive costs. We showcase the effectiveness of XIGL, including both existing and proposed explanation-based strategies, on several GNN architectures. Our implementation is available online.
cs.LG / 35 / 2608.14140
Smart routes: a system for development and comparison of algorithms for solving vehicle routing problems with realistic constraints
Abstract
The problem of route optimization with realistic constraints is becoming extremely relevant in the face of global urban population growth. While we are aware of approaches that theoretically provide an exact optimal solution, their application becomes challenging as the problem size increases because of exponential complexity. We investigate the Capacitated Vehicle Routing Problem with Time Windows (CVRPTW) and compare solutions obtaining by exact solver SCIP with heuristic algorithms such as LKH, 2-OPT, 3-OPT, the ORTools framework, and the deep learning model JAMPR. We demonstrate that for problem of size 50 deep learning and classical heuristic solutions became close to SCIP exact solution but requires less time. Additionally for problems with size 100, SCIP exact methods around 13 times slower that neural and classical heuristics with the same route cost and on around 50% worse for the first feasible solution on the same time. To conduct experiments, we developed the Smart Routes platform for solving route optimization problems, which includes exact, heuristic, and deep learning models, and facilitates convenient integration of custom algorithms and datasets.
cs.LG / 36 / 2608.14156
Deep Reinforcement Learning solution for pickup and delivery routing problems with time window and capacity constraints
Abstract
The task of constructing vehicles optimal routes for pickup and delivery of goods is one of most promising tasks in the context of global urban population growth. Although this kind of problems with small size can be solved by various classical approaches, a fast (or realtime) route optimizer under the constraints of the real world (such as capacity and time windows constraints) for medium-large size problems still remains a highly challenging task. In this work we, for the first time, successfully applied a deep Reinforcing Learning approach (modified JAMPR model) to solve Pickup and Delivery problem with Capacity and Time Window constraints (CPDPTW). We obtained a robust model that gives a fast optimal solution for problems of small and medium size, and gives fast suboptimal solution for problems of larger (> 200) size.
cs.LG / 37 / 2608.14177
Structure-Guided Spatiotemporal Attention Graph Neural Network for Traffic Flow Prediction
Abstract
Deep spatiotemporal models integrating graph convolutions and attention mechanisms have demonstrated excellent performance in network-level traffic flow prediction, owing to their exceptional ability to capture complex spatiotemporal dependencies. Despite their predictive success, deployment of such models in safety-critical urban systems remains constrained by their inherent lack of transparency. Existing post-hoc diagnostic methods often struggle with spurious correlations and fail to unveil the intrinsic decision-making mechanisms governing traffic dynamics, resulting in suboptimal interpretability and limited operational trustworthiness. To address these challenges, this paper proposes the Structure-Guided Spatiotemporal Attention Graph Neural Network (SGSAN). Departing from traditional architectures that rely on unconstrained adaptive graphs, SGSAN explicitly learns a static Directed Dependency Graph (DDG) to identify the invariant macroscopic propagation paths of traffic states. We further introduce an InfoNCE-based soft-coupling mechanism that anchors the model's dynamic spatiotemporal attention to this structural prior, offering a mechanistic account of the model's decision-making process while ensuring robust forecasting by aligning attention-based reasoning with identified macroscopic dependencies and preventing over-reliance on ephemeral local noise. Furthermore, a decoupled two-stage optimization framework is developed to resolve the fundamental conflict between structural discovery and predictive error minimization. Extensive experiments on multiple real-world datasets demonstrate that SGSAN achieves state-of-the-art predictive accuracy while providing built-in interpretability that organically aligns with the physical logic of traffic networks.
cs.LG / 38 / 2608.14186
Revisiting Energy-based Tabular Anomaly Detection: Energy and Reconstruction are Complementary
Abstract
Tabular anomaly detection is dominated by classical density-proxy methods (Isolation Forest, OCSVM, LOF), reconstruction-based detectors (Autoencoders, VAEs), and modern non-parametric scorers (COPOD, ECOD, Deep SVDD), all of which approximate the inlier distribution only indirectly; explicit energy-based models are largely absent. Motivated by the recent revival of EBMs in deep learning (e.g., Energy-Based Transformers, JEPA), we revisit the classical Deep Boltzmann Machine (DBM) for this task and hypothesize that its mean-field energy combines more effectively with a reconstruction-based score than same-lineage pairs do. We evaluate a two-hidden-layer DBM on two tabular benchmarks spanning distinct domains (UCI Bank Marketing and NSL-KDD) against eight classical and modern baselines across twenty random seeds. The DBM mean-field energy matches the strongest baseline (the Autoencoder) on Bank Marketing and statistically beats it on NSL-KDD, while significantly outperforming the remaining seven on both datasets. When fused with the Autoencoder via rank fusion, the DBM energy yields a statistically significant improvement on both datasets (AUROC=+0.014, p<0.01 on Bank Marketing; +0.002, p<0.001 on NSL-KDD); every non-DBM-derived base model instead fails to improve or significantly degrades the AE-paired ensemble. Our position is that classical EBMs, exemplified by the DBM, deserve a place in the tabular anomaly detection toolbox as a non-redundant complementary view to the reconstruction-based scores that dominate current practice.
cs.LG / 39 / 2608.14191
KV Cache Compression Through the Lens of Transform Coding
Abstract
The key-value (KV) cache stores information from past tokens and is a major memory bottleneck in long-context inference. Existing quantization methods address this bottleneck by representing the KV cache uniformly with lower-precision data types and designing quantization schemes to minimize reconstruction error in the cache itself, without accounting for how that error propagates through attention mechanisms. We prove that, under a white-noise quantization model, the expected attention-aware distortion decomposes into additive key and value contributions that factor across tokens and channels. Building on transform coding and reverse water-filling, which are classical tools from signal processing and rate-distortion theory, we introduce Attention-Aware Transform Coding (AATC), which allocates bits over a calibration set to minimize attention-aware distortion. On Llama-3.1-8B-Instruct and Qwen-2.5-7B-Instruct, evaluated across LongBench, RULER, GSM8K, MMLU-Pro, and MATH-500, our method achieves near-lossless accuracy at approximately $5.8\times$ compression, whereas each baseline degrades in at least some settings.
cs.LG / 40 / 2608.14198
MINT: A Universal Zero-Shot Predictor for Transaction Data
Abstract
Banks analyse sequential financial transaction data to perform many tasks, including fraud prevention, credit risk assessment and offer personalization. To improve the predictive accuracy of these tasks, Payments Foundation Models encode transaction sequence data as rich contextual embeddings, which can then be provided to task-specific models as features. However, these Foundation Models are not designed for flexible zero-shot reasoning across novel downstream prediction tasks, limiting their adaptability and utility. Existing LLM-based approaches to zero-shot prediction often fail to fully exploit the predictive signal within transaction data, while relying on costly text serialization or task-specific architectures that scale poorly. To address these limitations, we present the Multimodal Instruction Network for Transactions (MINT), a framework that connects a pretrained transaction sequence encoder to a decoder-only LLM through lightweight embedding injection, transaction-language alignment, and instruction tuning. We find that MINT achieves state-of-the-art predictive question-answering performance in both in-distribution and out-of-distribution questions, while substantially reducing input tokens, latency, and memory consumption compared to text-serialization baselines. Through comprehensive analyses of representations, alignment strategies, training data, and history length, we establish that compact transaction embeddings are a superior approach to transaction representation than text serialization for multimodal reasoning and zero-shot prediction tasks.
cs.LG / 41 / 2608.14209
Adaptive Protection for Evolutionary Feature Construction in Symbolic Regression with Application to Credit Classification
Abstract
Evolutionary feature construction has shown strong promise in symbolic regression by automatically discovering informative transformations of input features that enhance a simple base learner. However, existing approaches often lack explicit mechanisms to preserve important constructed features discovered during evolution, and valuable genetic material can be lost when genetic operators disrupt effective features. This paper introduces an adaptive protection mechanism that leverages feature importance metrics to selectively preserve constructed features during evolution. The mechanism provides stronger protection for more important constructed features while still allowing less important features to be modified and to incorporate useful building blocks from more important features. We evaluate the approach using multiple feature importance calculation methods and demonstrate its robustness across different base learners. Experimental results on 98 regression benchmark datasets show that the proposed mechanism consistently improves solution quality over baseline approaches, and experiments on two credit classification datasets demonstrate that the method also extends effectively to improve search effectiveness beyond symbolic regression.
cs.LG / 42 / 2608.14211
Training Fair Tabular Foundation Models
Abstract
Tabular Foundation Models (TFMs) have emerged as leading methods for tabular predictive tasks, leveraging in-context learning to predict on new data without task-specific training. Despite the increased use of TFMs in high-stakes decision-making, their fairness properties remain largely unexplored. In this work, we incorporate fairness constraints directly into TFM training, enabling fair predictions in a single forward pass. Our approach addresses two key challenges: limited access to sensitive attributes in training data, and the incompatibility of existing fairness techniques with the in-context learning paradigm. We propose FairTFM, a scalable training strategy based on synthetic fairness tasks and a fairness-aware architecture using a gradient reversal layer, which encourages the model to learn representations invariant to sensitive attributes. Experiments on 132 fairness tasks show consistent improvements in fairness while maintaining competitive accuracy.
cs.LG / 43 / 2608.14215
Connected Subspace Clustering: Hardness, a Scalable Heuristic, and an Application to Sea Level Geodesy
Abstract
Constrained optimization extends classical optimization by integrating side information, making it widely applicable across scientific and engineering domains. Consider a setting where we measure variables at different physical locations. When grouping these measurements, we often want clusters that are both internally similar and physically coherent. Thus, we have a constrained clustering problem where the constraint models coherence. Motivated by an application in geodesy, where contiguous regions of the sea surface must be identified for principal component analysis, we introduce the Connected Subspace Clustering problem: given high-dimensional points and a connectivity graph, partition them into $k$ connected clusters, minimizing their total squared distance to the clusters' best-fit $m'$-dimensional affine subspaces. We prove that, even for $m' = 0$ and a grid graph with holes, the problem is NP-hard to approximate within $Ω(n^{1/2-\varepsilon})$ for every $\varepsilon>0$, where $n$ is the number of measurements. We then introduce an efficient Lloyd-style heuristic that alternates subspace fitting with an iterative merging procedure to enforce connectivity. Our method returns exactly $k$ connected regions by construction, whereas unconstrained methods leave up to $1{,}966$ disconnected fragments at higher cost. In a study of 160 configurations on global sea level time series, our merging-based repair is the strongest of four strategies in $73.75\%$ of cases, and consistently outperforms competitors such as (connected) Ward's method across all tested cluster counts. The resulting regions isolate signals aligning with climate indices such as the El Nino-Southern Oscillation and Indian Ocean Dipole. Although developed for geodesy, the approach applies to other spatially embedded multivariate time series, such as climate fields, remote sensing, neuroimaging, and sensor networks.
cs.LG / 44 / 2608.14228
AutoSchema: Live Schema Grounding for Agentic Text-to-Sparql over Heterogeneous Knowledge Graphs
Abstract
Life science knowledge graphs make large collections of structured data available through SPARQL, but each resource uses its own schema, identifiers, and links. TogoMCP helps language model agents query these resources by providing curated Metadata Interoperability Exchange files. Creating and maintaining these files still requires language model assisted drafting, validation, and manual review. We study \emph{live schema grounding}, where an agent obtains the schema evidence needed for a question directly from the current endpoints. We present \textsc{autoschema}, a general framework for live schema grounding that requires no training. It inspects live schemas, maps entity names in a question to graph identifiers, explores relation paths, and finds possible connections between resources during iterative query construction. We use TogoMCP as our main comparison framework. We evaluate \textsc{autoschema} on Resource Focused Biomedical KGQA, Multi Resource Biomedical KGQA, Longitudinal Biomedical Semantic QA over BioASQ Task B, and Chemistry Knowledge Graph Transfer to a previously undocumented RDF graph. \textsc{autoschema} improves mean factoid accuracy over TogoMCP in the biomedical KGQA tasks and gives consistent gains in the longitudinal BioASQ evaluation. It also reduces iteration budget exhaustion and uses fewer tool calls on average in the core evaluation. The transfer study gives preliminary evidence that live schema grounding can support irregular and previously unseen graphs without first creating a curated schema file.
cs.LG / 45 / 2608.14264
Multi-Objective Bayesian Optimization for Model Merging
Abstract
Model merging combines trained models directly in weight space, offering a compute-efficient alternative to additional fine-tuning. Selecting merge parameters is nevertheless difficult because downstream evaluations are expensive, gradients are unavailable, and source capabilities can conflict. We formulate merge-parameter selection as a black-box multi-objective optimization problem and introduce MOBO-Merge, a merge-operator agnostic framework that uses multi-objective Bayesian optimization to approximate the Pareto front under a limited evaluation budget. We evaluate Qwen3-4B and Llama-3.1-8B in two-model instruction-math and three-model instruction-math-code settings using Linear, SLERP, TIES, and block-wise merge operators. On held-out benchmark partitions, MOBO-Merge obtains higher mean hypervolume than random search in 11 of 12 reported comparisons. The gain is small for one-dimensional Linear interpolation but substantially larger for several TIES, block-wise, and three-objective searches. No merge operator is uniformly best: TIES leads in three of four family-setting combinations, whereas Block-Linear 4x is strongest for the Llama three-model merge. These results show that multi-objective Bayesian optimization is valuable as a search layer for expressive merge parameterizations.
cs.LG / 46 / 2608.14288
Convex losses and their applications to SVM, SVR, and Shallow Neural Networks
Abstract
We propose multiple new convex losses for SVM and Neural Networks, applied to binary classification tasks. While there are practical limitations in exploiting them with the dual SVM models, we are able to use them with SVM primal formulation and Neural Networks. In detail, the primal SVM problem with the modified losses has been solved with the Particle Swarm Optimization algorithm. We prove that the proposed losses are a generalization of the standard loss, and we experiment them with several small data-sets. This preliminary study shows that using pattern correlations inside the loss function could in theory enhance the generalization performances on some data-sets. To evaluate the performance of each loss, we adopt a Nested Cross-Validation procedure. Results show that generalization measures are the same with or without the new losses.
cs.LG / 47 / 2608.14319
Quantum Multi-Armed Bandits and Linear Bandits: Lower Bounds and Algorithms
Abstract
We study quantum multi-armed bandits (QMAB) and quantum linear bandits (QLB) in the model of Wan et al. [2023], where the learner queries each arm or action through a quantum reward oracle or its inverse. Prior work gives algorithms over horizon $T$ with regret $O(K\log T)$ for QMAB with $K$ arms and $O(d^2\operatorname{polylog} T)$ for $d$-dimensional QLB. This leaves open whether the $K\log T$ scale is unavoidable and whether the $d^2$ dependence can be improved. We prove the first minimax lower bounds of $Ω(K\log(T/K))$ for QMAB and $Ω(d\log(T/d))$ for finite-action QLB, resolving the question raised by Wan et al. [2023] of whether regret independent of $T$ is achievable. At the heart of our argument is a high-confidence single-arm quantum testing lower bound for distinguishing a fixed reward mean from an interval of alternatives, proved by the polynomial method and a Remez-type inequality for trigonometric polynomials. A bandit-to-testing reduction then lifts it to the QMAB lower bound, while a linear embedding gives the finite-action QLB lower bound. Complementing the lower bounds, we give a design-based elimination algorithm for finite-action QLB. When the action set has size $\operatorname{poly}(d)$, its regret is linear in $d$, improving the prior $d^2$ dependence and matching our lower bound up to polylogarithmic factors. The algorithm couples a low-bias low-variance quantum mean estimator with a small-support $G$-optimal design through a query allocation matched to the design weights. The design-based elimination reduces the dimension dependence from $d^2$ to $d^{3/2}$ when using Quantum Monte Carlo estimates. The low-variance estimator then makes reconstruction error aggregate through variance rather than worst-case absolute error, removing the remaining $\sqrt d$ factor.
cs.LG / 48 / 2608.14349
Non-Parametric Spatiotemporal Trajectory Prediction via State-Conditioned Transition Sampling
Abstract
We present a training-free method for multi-modal trajectory prediction that achieves comparable accuracy to a 57M-parameter transformer while requiring no GPU and zero learned parameters. The method builds a transition table of historical state-to-next-position pairs and retrieves neighbors using a product kernel over spatial proximity, bearing, speed, and temporal context. Two inference modes operate over this shared representation: diversity-penalized sampling produces trajectories covering distinct plausible routes, while beam search finds the highest-likelihood path. On the TrAISformer benchmark (Danish Maritime AIS), our method achieves competitive accuracy at full data availability and dramatically outperforms the transformer in data-scarce regimes---remaining stable down to 10% of training data where TrAISformer degrades catastrophically. This enables deployment in new geographic regions from an order of magnitude less historical data, and with no GPU training.
cs.LG / 49 / 2608.14367
Mind the Long Tail: Understanding the Difficulty of Delay Detection in Business Processes
Abstract
The early detection of delayed cases in business processes is a critical capability for organizations. Predictive process monitoring (PPM) supports this task by using historical event logs to predict the remaining time of ongoing cases, enabling timely interventions to avoid missed deadlines and service level violations. Although remaining time prediction has advanced considerably through sophisticated deep learning architectures, little is known about the intrinsic difficulty of delay detection itself. Since performance is typically assessed using aggregate metrics, prior work provides limited insight into how models perform across the target distribution, especially on the operationally most critical cases with large delays. In this paper, we address this gap by analyzing the difficulty of delay detection. Across 14 event logs, we show that remaining times are typically strongly right-skewed, with only a small fraction of cases exhibiting large delays. Existing models capture the mode of this distribution well but perform poorly on high-delay cases. We further uncover pronounced heteroscedasticity, showing that predictive uncertainty increases with delay magnitude. Based on these findings, we evaluate approaches to mitigate the imbalance problem, but find only limited benefits, suggesting that the key underlying problem may not be imbalance but higher uncertainty associated with delayed cases. We show that this correlation can be exploited to substantially improve the identification of delayed cases. Overall, our work provides new insights into the sources of difficulty in delay detection and identifies uncertainty-aware modeling as a promising direction for future PPM research.
cs.LG / 50 / 2608.14372
Catching the Imposter: Self-Supervised Learning of Physical Coherence with Cross-Entity Feature Permutations
Abstract
Scientific data often describe entities whose features are jointly governed by the laws of physics, yet existing self-supervised learning (SSL) objectives largely ignore this physical coherence. We introduce imposter, a discriminative pretext task that replaces subsets of an entity's features with real observations donated by another entity and trains the encoder to identify the swapped features. Because every donated value is individually plausible, the task can only be solved by learning cross-feature physical dependencies. We evaluate the proposed objectives on global ERA5-Land reanalysis data using 21 environmental variables and assess the learned representations on seven downstream tasks spanning climate classification, carbon flux estimation, and streamflow prediction. Our study includes, to our knowledge, the first systematic comparison of self-supervised objectives for land-surface modeling under a shared architecture and pre-training budget. We find that the most effective pretext task depends on the downstream task family rather than any single objective's superiority, and that imposter provides complementary information when combined with existing SSL objectives. These results suggest that physical coherence is a valuable new source of self-supervision for scientific foundation models.
cs.LG / 51 / 2608.14373
Boosting Data Augmentation with Stochastic Weight Averaging
Abstract
The symmetries of a learning task have become an important factor in designing modern deep learning solutions. Data augmentation is a straightforward and effective way of incorporating symmetries into a generic neural network. Recent results show that infinitely large deep ensembles show perfect symmetry when trained on augmented data. However, since training ensembles requires repeating the training process many times, this method is costly. In this work, we study stochastic weight averaging (SWA) as an alternative ensembling technique that does not require repeated training runs. We analyze SWA by approximating the stochastic training trajectory at the end of training with an Ornstein--Uhlenbeck process. We show that in the infinite-width limit, SWA on augmented data provides an equiviariance boost that goes beyond what could be expected from the performance increase due to SWA alone. We verify our results with extensive numerical experiments on numerous models spanning computer vision and graph classification with both discrete and continuous symmetries.
cs.LG / 52 / 2608.14385
DeaMoE: Efficient MoE Structure for Fast Small-Batch Decoding
Abstract
Mixture-of-Experts (MoE) models have been widely adopted in real-time interactive applications such as coding assistants, real-time audio-video interaction systems. To meet the extremely low response latency requirements of these scenarios, practitioners commonly employ small-batch decoding, under which MoE inference becomes memory-bound and is severely bottlenecked by expert weight loading. However, this bottleneck has received limited attention, and existing solutions such as post-training weight compression or fine-grained expert design during pre-training either degrade model accuracy or introduce additional computation and communication overhead. To tackle this issue, we propose DeaMoE, a decoding-efficient MoE architecture, in which the experts are grouped into several departments, and the experts belonging to the same department share most parameters since they come from the same professional field, and additionally each expert contains a few private parameters to reflect its uniqueness. Moreover, we design customized two-stage routing strategy for DeaMoE to avoid redundant loading, under which DeaMoE greatly improves the efficiency during LLM decoding. Compared with vanilla MoE, DeaMoE reduces per-step loaded weights by up to 50.9% and achieves up to 1.33 end-to-end TPOT speedup for the pre-trained 7B model on A40, and up to 2.00x and 1.97x peak speedup for DeepSeek-V3 on A40 and H100 in microbenchmarks.
cs.LG / 53 / 2608.14414
CytoBERT: A Foundation Model for Cytometry Data
Abstract
Cytometry measures the complex characteristics of single cells (e.g., counts and protein expression of immune cells) and is widely used across immunological research and clinical settings. However, cytometry data is highly heterogeneous and unstandardized due to experimental protocols and the choice of measured features. While machine learning methods hold the potential to gain deeper insights into cell biology, these challenges make them difficult to apply and transfer across studies. Recent advances in foundation models can alleviate these issues, but corresponding approaches are still scarce in this field. To address this, we provide CytoBERT, a publicly available, open-source, open-weight foundation model for single-cell cytometry data with variable marker panels. CytoBERT is pretrained in a self-supervised manner on a large-scale cytometry corpus (15 human datasets with heterogeneous marker panels and more than 50 million cells) curated through marker standardization, enabling it to learn transferable inter-marker relationships within cells. Fine-tuning CytoBERT for sample-level classification demonstrates that transfer learning across heterogeneous cytometry datasets is feasible, providing a starting point for scalable, generalizable cytometry analysis. Code is available at GitHub.
cs.LG / 54 / 2608.14420
More Correct Mass, Worse Answers: Why Power Sampling Can Fail and How to Fix It
Abstract
Power Sampling sharpens a language model's distribution over complete generation trajectories, offering a verifier-free way to improve reasoning at inference time. It also has the potential to serve as a general-purpose front end for a broad range of downstream sampling methods. However, we uncover a striking paradox: Power Sampling can drive more probability mass toward correct trajectories while degrading the downstream inference it is intended to enhance. Using self-consistency as a representative case, we observe accuracy drops of up to 18.5 percentage points across models and reasoning benchmarks. We trace this paradox to two mismatches. Dose mismatch arises because a fixed exponent induces drastically different amounts of distributional change across problems. Coverage mismatch arises because global sharpening concentrates mass on a narrow set of dominant paths: high pass@k, often interpreted as evidence of preserved diversity, can therefore coexist with the loss of broad reasoning-path support required for downstream aggregation, search, and selection. Guided by this diagnosis, we replace uniform trajectory exponentiation with a deformation-controlled, support-preserving Power target that calibrates sharpening across problems while limiting the suppression of moderate-probability paths. In a same-budget instantiation with weighted self-consistency, the repaired sampler reverses the losses caused by global Power and outperforms standard multi-sample inference across reasoning benchmarks.
cs.LG / 55 / 2608.14443
Designing Compact Neural Architectures via Neuron Gating and Mixed Activation
Abstract
Neural Architecture Search (NAS) is naturally formulated as a bilevel optimization problem, where the upper-level optimizes the architecture using validation performance and the lower-level trains network parameters using training loss. However, NAS is computationally expensive due to discrete architectural decisions, exponentially growing search spaces, and the high cost of training candidate architectures. This work develops a general bilevel optimization framework for NAS across diverse architectures, including MLPs, CNNs, RNNs, and Transformers, to identify compact architectures with strong predictive performance. We propose three scalable formulations that replace discrete neuron- and activation-level decisions with continuous relaxations, enabling differentiable optimization over otherwise combinatorial architecture spaces. These formulations give rise to three NAS methods: NAS based on Neuron Gating (NAS-NG), NAS based on Mixed Activation (NAS-MA), and NAS based on Neuron Gating and Mixed Activation (NAS-NGMA). Experiments on MLPs and CNNs using MNIST and CIFAR-10 show that the proposed methods consistently identify compact architectures with competitive or improved predictive performance. On MNIST, NAS-NGMA achieves 98.68% test accuracy with 7.69M MLP parameters, while NAS-NG achieves 99.63% accuracy with only 0.26M CNN parameters. On CIFAR-10, the proposed methods consistently outperform vanilla DARTS. Further experiments demonstrate that NAS-NG can optimize substantially over-parameterized and literature-optimal architectures, improving accuracy while reducing parameters. These results establish relaxed bilevel optimization as a scalable alternative to discrete NAS and provide a general framework for efficient neuron- and activation-level architecture optimization.
cs.LG / 56 / 2608.14472
LP-NAS: Linear Programming-based Neural Architecture Search
Abstract
Neural Architecture Search (NAS) aims to automate neural network architecture design, reducing reliance on human expertise. Among the various NAS methods, differentiable NAS has gained prominence due to its efficiency and accuracy compared to conventional NAS approaches. Since differentiable NAS relaxes the architecture search space into a continuous domain, it is possible to apply principles from continuous optimization to NAS. In this paper, we propose Linear Programming-based NAS (LP-NAS), a mathematical programming-based framework for differentiable NAS that is applicable to a wide range of continuous search spaces. LP-NAS formulates a linear program (LP) using the validation-loss gradient and the training-loss Hessian to compute an architecture update direction that improves generalization while preserving the optimality of the model parameters. By following this LP-derived descent direction, LP-NAS efficiently navigates the architecture search space, leading to faster and more effective architecture optimization. We introduce two computationally efficient variants of LP-NAS, namely S-LP-NAS and R-LP-NAS. Applying LP-NAS to the Differentiable Architecture Search (DARTS) search space results in two algorithmic variants, S-LP-DARTS and R-LP-DARTS. Both variants achieve faster convergence and significantly higher validation performance during the early search iterations than the standard DARTS algorithm. Extensive experiments on CIFAR-10 and CIFAR-100 show that LP-DARTS outperforms standard DARTS in both the architecture search and evaluation phases. Additionally, we compare our approach with several DARTS variants (P-DARTS, PC-DARTS, and STO-DARTS) on the CIFAR-10 dataset and demonstrate its effectiveness. Furthermore, we validate the transferability of the discovered architectures through experiments on the ImageNet dataset.
cs.LG / 57 / 2608.14492
Approximate Muon with low-rank adapters
Abstract
The Muon optimizer shows clear benefits versus alternatives when pretraining neural networks. However, it is used less frequently for parameter-efficient fine-tuning (PEFT). One potential reason is that the most common PEFT method, LoRA, does not naturally combine with Muon since it is not mathematically possible to orthogonalize the weight update given by a low-rank parameterization. In this paper, we address this issue by approximating the solution to a relaxed Muon objective in the low-rank setting via linearization and then least-squares. We provide an efficient implementation that uses matmul operations only, as opposed to more complex linear algebra decomposition routines. Our method, sMuon (small Muon), performs favourably across SFT and a ReLoRA pretraining experiment. While results are model- and eval-dependent, we find overall that using Muon for low-rank fine-tuning provides moderate performance improvements.
cs.LG / 58 / 2608.14498
Rollplex: Cross-Phase GPU Spatial Sharing for Vision Language Model Post-Training
Abstract
Vision-language models (VLMs) enable embodied agents to reason and act from visual observations and language instructions. Reinforcement learning (RL) post-training enhances these capabilities using task feedback, but current on-policy RL runtimes execute rollout, reference scoring, and actor training in strict serial phases. While effective for text-only RL, this phase-granular execution is wasteful for VLMs, where processing dense video inputs and prompt prefixes occupies a large fraction of each phase. Because prefix processing is independent of the generated response, it can be run alongside rollout decoding, which leaves GPU compute capacity underutilized, without breaking synchronous on-policy semantics. We present Rollplex, a runtime that decomposes the reference and training phase and moves the prefix computation into the rollout decode window. Realizing this schedule requires more than concurrent kernel launches: naive colocation of Qwen2.5-VL-32\,B requires roughly 165\,GiB per GPU, while rollout and training prefer different tensor-parallel (TP) degrees and weight layouts. Rollplex addresses these constraints with two mechanisms. Phase-aware memory management controls HBM residency according to producer--consumer lifetimes. Parallelism-aware weight sharing uses the same physical storage for layout-compatible tensors across distinct TP degrees and reconstructs only incompatible tensors, avoiding a complete second actor copy. On 32 H800 GPUs, Rollplex achieves $1.23\times$--$1.30\times$ speedup over serial colocation and $1.57\times$--$2.24\times$ over disaggregation under the same GPU budget, while preserving the synchronous RL update.
cs.LG / 59 / 2608.14505
RecipeNet: A Hierarchical Transformer for Recipe Data
Abstract
Recipe data arises in domains such as materials synthesis, pharmaceutical formulation, and industrial manufacturing, where procedures are represented as ordered sequences of steps containing heterogeneous structured fields. Existing tabular learning methods typically flatten this structure into fixed-schema representations, limiting their ability to capture hierarchical field interactions and procedural dependencies. We propose RecipeNet, a hierarchical Transformer architecture that encodes field-level interactions within each step and sequential dependencies across steps through stacked Transformer encoders. Experiments on multiple recipe datasets and tasks demonstrate that RecipeNet consistently outperforms existing tabular models, highlighting the value of hierarchical and sequential modeling for recipe representation learning.
cs.LG / 60 / 2608.14135
AgilePE: Autonomous UAV Pursuit-Evasion via Self-Play Reinforcement Learning
Abstract
Autonomous pursuit-evasion is a fundamental challenge for Unmanned Aerial Vehicles (UAVs), requiring rapid decision-making under tightly coupled dynamics and continuously changing opponent behaviors. Traditional rule-based or differential-game approaches often struggle with high-dimensional aerial interactions and agile maneuvering. We present AgilePE, a complete system for autonomous UAV pursuit-evasion via self-play reinforcement learning. AgilePE integrates agile low-level control, competitive policy optimization, and sim-to-real deployment in a unified framework. The policy directly maps onboard state observations to Collective Thrust and Body Rates (CTBR) commands, enabling end-to-end agile maneuvering without intermediate trajectory planners or waypoint controllers. For training, we use competitive self-play with Prioritized Fictitious Self-Play (PFSP) and a diversified opponent pool, enabling agents to improve against historical policies while stabilizing optimization and reducing policy oscillation. This process leads to the emergence of sophisticated pursuit and evasion strategies. For real-world deployment, we develop a hardware-aligned simulation pipeline that models actuator-response dynamics, communication latency, and domain randomization. The learned policies transfer zero-shot to real quadrotors without task-specific tuning. Real-world experiments reproduce pursuit-evasion tactics observed in simulation, including rapid dodging and flanking, and demonstrate interactive two-agent zero-shot deployment.
cs.LG / 61 / 2608.14332
CORAL: Curriculum-Optimized Reward Adaptation for LiDAR-Based Goal-Directed Urban Driving
Abstract
Reinforcement learning is promising for autonomous urban driving, but long-horizon goal-directed navigation asks a policy to acquire several competing behaviors at once--reaching a distant goal, tracking a route, avoiding obstacles, obeying signals--and a fixed objective gives no order in which to learn them. This paper presents CORAL, which advances two schedules together: a five-stage curriculum that progressively lengthens routes and tightens behavioral constraints, and a stage-aware reward whose component weights shift emphasis from mission progress toward route following, safety, smoothness, and rule compliance as the task hardens. The policy is a multi-stream actor-critic network trained with Proximal Policy Optimization (PPO) in CARLA on a compact 99-dimensional state pairing a polar LiDAR histogram with vehicle telemetry, ego-frame route geometry, and traffic-rule indicators--no point-cloud encoder, no bird's-eye-view rasterization. Against two PPO baselines under an identical protocol, CORAL reaches the goal in all twenty evaluation episodes on the longest routes under the full set of behavioral constraints, where the baselines reach 5% and 10%; a factorial ablation shows that neither schedule alone matches their combination: removing either lowers both success and route completion, and disabling both drops success to 55%. Trained in one town, the policy transfers zero-shot to seven unseen towns, succeeding in 68-98% of episodes on routes of the same 100-150 m length, with mean lateral deviation below 0.35 m.
cs.LG / 62 / 2608.14466
Expected Free Energy-based Informative Path Planning for Robotic Mars Exploration
Abstract
An autonomous robot efficiently exploring an unknown environment, such as looking for water sources on Mars, faces two simultaneous demands: building an accurate information map while quickly finding the regions of greatest value, and paying for every meter of travel and the cost of every measurement it takes. Classical information-seeking and reward-seeking criteria address only one of these objectives at a time. Here, we propose Expected Free Energy (EFE), the principled action-selection objective from active inference, as a unifying criterion for budgeted robotic informative path planning. Maintaining a Gaussian-process belief over the information field, our agent plans continuous trajectories that minimize expected free energy under hard path-length constraints. The results from multiple realizations show that EFE-based planning yields accurate posterior maps and locates the highest-value regions simultaneously, outperforming information-theoretic baselines under the same settings. In robotic exploration, these unified, easy-to-tune principled information-gathering strategies facilitate autonomous deployment while enforcing efficiency and resource constraints.
cs.LG / 63 / 2608.13724
Architecture and Affordances of PLAUD: Performative Latents and Unsupervised DDSP
Abstract
PLAUD (Performative Latents and Unsupervised DDSP) is a neural synthesizer and Max for Live instrument for live electronic music, built on NoiseBandNet and trained on small personal sound corpora. We present its architecture, combining a variational DDSP synthesis model, latent smoothing, multi-scale spectral and adversarial losses, and an optional transformer prior, alongside a set of bending operations that intervene directly in the synthesis chain: component limiting, waveshaping, and prior feedback. The Max for Live interface exposes control generation, trajectory sampling, and modulation as primary modes of interaction. Throughout, we thread an affordance analysis arguing that the system's performative character follows from architectural decisions rather than being designed on top of them. The paper contributes both a technical account of the system and a situated affordance analysis of its role in live electronic music performance.
cs.LG / 64 / 2608.13817
Trajectory Dynamics in Self-Supervised Learning Latent Space for Audio Deepfake Detection
Abstract
Human speech production is constrained by physiology, giving rise to characteristic temporal structure on acoustic signals. We hypothesise that these constraints manifest as structured trajectory dynamics in the latent space of Self-Supervised Learning (SSL) models, and that synthetic speech violates them detectably. To test this hypothesis, we train a causal Long Short-Term Memory (LSTM) next-frame predictor on bonafide speech only (Stage 1), using the deepfake-specialised SSL backbone Wav2Vec2-Large-AntiDeepfake, and compare against a static global-average-pooling baseline using identical features, thus isolating the contribution of temporal modelling. A supervised Stage 2, which trains a Multi-Layer Perceptron on the frozen LSTM internal states using labelled data, is included to characterise the role of spoof supervision. Our system achieves competitive or state-of-the-art performance across six benchmarks: ASVspoof 2019/2021, Codecfake, In-the-Wild, MLAAD-EN, and Deepfake-Eval-2024, including best published EER on ASVspoof 2021 (0.75\%) and, notably, Stage 1 trained on bonafide speech only surpasses the published supervised baseline from the same backbone on DE2024 (30.35\%). On near-domain benchmarks, static and dynamic approaches perform comparably. On harder cross-corpus benchmarks with diverse synthesis methods, trajectory dynamics provide substantial gains, confirming that temporal physiological constraints carry detection signal beyond utterance-level statistics.
cs.LG / 65 / 2608.13711
TRUE-Colon: Exposing a Consistent Transfer Asymmetry in Real-Time Polyp Detection
Abstract
Computer-aided detection (CADe) systems for colonoscopy promise to reduce clinical miss rates, yet reliable real-world deployment remains elusive. This translational gap stems in part from a structural flaw in model development: the reliance on curated datasets that under-represent the long negative stretches and procedure-related artifacts characteristic of routine examinations. Training and evaluating architectures strictly on these lesion-centric benchmarks creates an illusion of success, since such benchmarks cannot capture clinically crucial metrics. To expose this gap, we establish TRUE-Colon, a standardized benchmarking protocol that measures key deployment characteristics alongside localization accuracy, and evaluate four real-time architectures (Faster R-CNN, YOLOv8, YOLOv11, RT-DETR) across curated benchmarks (SUN, PICCOLO) and 60 unedited, full-length procedures (REAL-Colon). We observe a consistent transfer asymmetry: models trained strictly on curated clips suffer a severe performance collapse when evaluated on full procedures, whereas procedure-trained models substantially improve rejection of non-polyp content on REAL-Colon, and largely retain their accuracy on curated benchmarks. Beyond transferability, we find that the Transformer detector attains the strongest sensitivity and the earliest, most persistent detections, while the convolutional detectors stay competitive at a higher throughput. Together, these results indicate that both training and benchmarking for deployable CADe should shift from curated, lesion-centric clips toward full-procedure data and deployment-relevant operating points. Source code is available at https://github.com/sdoerrich97/true-colon.
cs.LG / 66 / 2608.13651
Consistent Model Chasing Is Minimax Optimal: The Exact Value of Scalar Adversarial Adaptive Control under Large Parametric Uncertainty
Abstract
We solve exactly a fundamental problem of adaptive control against adversarial disturbances: regulate the scalar system $x_{t+1} = ax_t + u_t + w_t$, $x_0=0$, $\|w\|_\infty \le 1$, where the constant pole $a \in [-Δ, Δ]$ is unknown in sign and magnitude and $Δ$ is arbitrarily large. Elementary as the system looks, the least worst-case peak $\|x\|_\infty$ that a causal controller can guarantee against an adversarial pair $(a, w)$ (the value of this game) has, to our knowledge, never been determined for any adaptive control problem with parametric uncertainty of arbitrary size under this criterion; existing theory supplies stability certificates, gain bounds, and regret rates, not the value. That value is $γ^\star(Δ) = 1 + Δ$ for every $Δ>0$. The summand $1$ is the irreducible price of the disturbance, and $Δ$ the exact price of a single, unavoidable identification spike. The optimal policy is certainty-equivalent deadbeat control at the midpoint of the set-membership consistent interval, an instance of the robust oracle $\times$ consistent model chasing architecture. The architecture is forced, not merely sufficient: writing $θ_t := -u_t/x_t$ exhibits every causal controller as an oracle-selector composition, and optimality pins the selector to the midpoint at the critical histories. The standard tools, classical and modern, each fail quantifiably: probing is punished before it pays, commitment is fatal at sub-disturbance excitation once adaptation is necessary, optimism degenerates to tie-breaking or pays asymptotically at least twice the optimum, and regret certificates are blind to the worst-case peak in both directions. The optimal law contains no exploration mechanism, its learning purely passive. These results give the first exact optimality certificate for consistent model chasing as a design principle for adversarial adaptive control.
cs.LG / 67 / 2608.14278
Pairton: Iterative Reconstruction of Short-Lived Particles
Abstract
We present Pairton, an iterative framework for reconstructing short-lived particles in high-energy collision events. By formulating particle reconstruction as a masked prediction process over graph structures, Pairton learns conditional distributions consistent with a factorised decomposition of decay products and iteratively predicts edges in the adjacency matrix representing particle decay relationships. Leveraging a pairformer-based architecture with dynamically updated pairwise representations, our method incorporates global event consistency. We demonstrate state-of-the-art performance on fully hadronic $t\bar{t}$ decays. Pairton provides a general, flexible paradigm for particle reconstruction and can be readily extended to other topologies, bridging ideas from modern generative modelling and high-energy physics.
cs.LG / 68 / 2608.14369
Non-Shattering at and Above the Dynamical Temperature in the Spherical Pure p-Spin Model
Abstract
We consider the notion of shattering introduced by Ben Arous and Jagannath for spherical pure $p$-spin glasses with overlap $q$. For every $p\geq 3$ and $0<β\leqβ_{\mathrm{sh}}(p)$, we rule out shattering whenever $q\leq2^{-1/2}$ or $q>\sqrt{(p-2)/(p-1)}$. The proof combines a deterministic $N+1$ bound for disjoint bands in the first range with a general-$p$ sign law showing that their total marked weight has subdominant free energy in the second. A spherical-code bound and Hölder's inequality give an additional $q$-dependent obstruction; in particular, they rule out every fixed overlap for $0<β\leq\sqrt{\log2}$. For $p=3$, the first two ranges already exhaust every fixed $q\in(0,1)$, so the landscape is not shattered at any $T\geq T_{\mathrm{sh}}$. For $p\geq4$, the cases not covered by our criteria are confined to $2^{-1/2}<q\leq\sqrt{(p-2)/(p-1)}$ and $\sqrt{\log2}<β\leqβ_{\mathrm{sh}}(p)$. In particular, this paper partially resolves Conjecture 1 of the paper above and also suggests new methods to show non-shattering.
cs.LG / 69 / 2608.14245
Body size predicts how long ant workers live - but not how they age or how they die from heat
Abstract
In social insects, mortality risk comprises distinct components that may not share the same predictors: lifespan duration, senescence trajectory, and thermal vulnerability. We tested these three axes in 18 Australian ant species using paired field-laboratory survival assays (2,363 cohort-day observations; 1,148 workers). Body size predicted duration (Cox HR = 0.67, p = 0.002), while colony size (p = 0.60) and the size x temperature interaction (p = 0.72) showed no detectable moderating effect. A weak but significant size x foraging-rate interaction was detected (LRT p = 0.014), suggesting that intrinsic physiology remains the most parsimonious explanation for the main size-longevity pattern, although ecological context may contribute. Senescence trajectory was associated with circadian niche rather than size: it was steepest in matinal species (Kruskal-Wallis p = 0.009; matinal vs. crepuscular p = 0.002) and was uncorrelated with body mass (Spearman p = 0.32). Thermal hazard plateaued above 20 degrees C (Delta AIC = -38; p < 0.001), with elevated thermal sensitivity in Rhytidoponera (Ectatomminae) above the plateau (5% per degree C, p = 0.015). Circadian regime and lineage identity, not body size, therefore emerge as the most climate-relevant axes, although they are strongly collinear (Cramer's V = 0.85). These results show that body size captures only one dimension of mortality risk and that size-based vulnerability indices may misrank taxa when senescence and thermal sensitivity are decoupled from body size.
cs.LG / 70 / 2608.14169
Classical Limits of Spectral Filtering in Quantum Generative Models
Abstract
Spectral filtering has been proposed as a route to regularization in quantum generative models: the quantum Fourier transform exposes the amplitude spectrum of a quantum circuit Born machine, and a diagonal filter suppresses the high frequencies associated with finite-sample noise, an operation whose classical counterpart seemingly requires manipulating an exponentially long amplitude vector. We examine whether this coherent operation produces anything that classical post-processing of samples from the unfiltered model cannot match. Measuring the filter against convolution with a symmetric probability kernel at matched sampling cost, which accounts for the post-selection overhead of attenuation, we derive necessary and sufficient conditions for the gap between the two to vanish. Magnitude (attenuating) filters obey a dichotomy: at a fixed affordability threshold, the filtered output is either a constant-size Fourier object with an efficient classical sampler, or the passband must widen until no fixed frequency is attenuated and the filter no longer smooths. In neither case does the filter create a quantum-classical separation. Whatever separation survives is inherited from the spectral phase of the input state. Numerical experiments on trained circuit Born machines confirm the classification and show that the deciding phases are invisible to the Born-rule training loss and set by the initialization. Within the diagonal family, pure phase filters remain the only spectral operations exempt from these constraints.
cs.LG / 71 / 2608.13793
On the Brittleness of Maximum Likelihood Estimation for Gaussian Process Hyperparameter Optimization
Abstract
Machine learning (ML) has become an indispensable part of modern engineering design workflows. A crucial step in training an ML model is the selection of the loss function which can be systematically formulated via various techniques such as maximum likelihood estimation (MLE) and cross-validation . While MLE is one of the most popular, effective, and intuitive mechanisms for training ML models, it is brittle: if the assumptions underpinning it are not met, the trained ML model may generalize poorly. This brittleness affects even Gaussian processes (GPs) which are widely used in engineering design and are often (incorrectly) presumed to be very robust to overfitting. In this paper, we fundamentally evaluate the brittleness of MLE in the context of training GPs for probabilistic regression or classification tasks. We compare theoretically grounded metrics against MLE and propose practical solutions. Our extensive studies demonstrate the effectiveness of our solutions in downstream design tasks such as Bayesian optimization and provide a blueprint for practitioners to build accurate and robust GPs that can even outperform tabular foundation models in terms of prediction accuracy, uncertainty quantification, and inference cost. Our contributions are publicly available via GitHub at https://github.com/Bostanabad-Research-Group/GP-vs-TabPFN-vs-GPyTorch.
cs.LG / 72 / 2608.13937
Deep Vision in Smart Manufacturing: MODERN Framework for Intelligent Quality Monitoring and Diagnosis
Abstract
Smart manufacturing processes are often installed with a large number of sensors, imaging devices and computers, which not only enable instant communication across various modules of a production system but also aid in intelligent manufacturing management. In this paper, we introduce MODERN, a deep learning framework for quality monitoring and fault isolation, which integrates these enhanced capabilities into the practice of industrial quality control. Using the architecture of an inception residual neural network, we develop a control chart that monitors the likelihood of a product containing defects. We also propose a faulty region estimator that identifies the defective area using transfer learning. To extend our framework to cases where there are not sufficient training data, we suggest a transfer monitoring technique that requires only a small sample size and a hypothesis testing approach for quantitatively assessing the applicability of our method. Theoretically, we establish the minimax optimal convergence rate for both our defect likelihood estimation and fault diagnosis. Our results lead to a seemingly counter-intuitive managerial implication - it may not always be in a manufacturer's best interests to keep upgrading its monitoring equipment regardless of the cost. Empirically, we demonstrate the superior performance of our method in comparison with a state-of-the-art approach using both simulated experiments and real data.
cs.LG / 73 / 2608.14408
Online Inference in Distributional Temporal-Difference Learning
Abstract
We study online statistical inference for functionals of the return distribution under a fixed policy. The return distribution is estimated by nonparametric distributional temporal-difference learning from a single Markov trajectory. For the Polyak--Ruppert averaged estimator, we prove that its root-$T$ error converges weakly to a centered Gaussian random element in Cramér space. We also prove that, conditionally on the observed trajectory, the root-$T$ difference between the bootstrap and original averages converges weakly to the same Gaussian limit. These results justify bootstrap inference for smooth statistical functionals, including variance, CVaR, expected shortfall, and expectiles. For nonsmooth statistical functionals, we develop a local asymptotic theory for the estimated return CDF over $T^{-1/2}$-neighborhoods of finitely many thresholds, together with its bootstrap analogue. This theory allows us to conduct inference for nonsmooth statistical functionals characterized by CDF equations, including return quantiles.
神经与进化计算 (cs.NE)
2
cs.NE / 1 / 2608.13952
Reducing ANN-SNN Conversion Error via Residual Membrane Potential Alignment
Abstract
Spiking Neural Networks (SNNs) serve as core architectures for neuromorphic computing thanks to event-driven operation and ultra-low power consumption. Direct SNN training is hindered by non-differentiable spikes that induce vanishing gradients and unstable optimization. ANN-SNN conversion circumvents such issues by reusing well-trained ANN weights for low-latency, energy-efficient inference. Nevertheless, existing conversion schemes suffer from severe accuracy drops at small timesteps, large inference delays and cumulative quantization errors, even with marginal performance loss at large $T$. To address these limitations, we first analyze flaws of conventional conversion pipelines from residual membrane potential statistics and propose a novel conversion strategy combining dynamic initial potential tuning and feature enhancement. We then introduce a regularization loss $\mathcal{L}_{\mathrm{RMPD}}$ to adapt initial potential of IF neurons and mitigate systematic truncation bias from boundary aggregation. A dedicated SCR-Conv2d competitive refinement layer with grouped convolution is further built to sharpen feature discrimination, eliminate redundant spikes and stabilize encoding under tiny time windows. Integrated with the state-of-the-art QCFS baseline, our approach delivers consistent low-latency performance gains and generalizes to ReLU CNNs, ANN Transformers, and multi-threshold SNN variants. Evaluations on CIFAR-10, CIFAR-100 and ImageNet verify prominent accuracy improvements at $T=2,4,8$, with negligible extra computation overhead. This work offers an effective conversion paradigm to facilitate real-world SNN deployment on neuromorphic chips.
cs.NE / 2 / 2608.14019
Emergent Models: Intelligence from Tiny Substrates
Abstract
Emergent Models (EMs) are a machine learning paradigm based on simple yet open-ended substrates, such as cellular automata, in which modeling is treated not as the learning of a closed-form input-output map but as the emergence, within simple dynamical systems, of computational behaviors that solve external tasks. Such substrates typically iterate a fixed local rule over a latent space for an adaptive number of steps, with an interface linking the latent state to external input/output signals. Training proceeds by evolutionary search. We hypothesize that some instances of this framework are biased toward global generalization: capturing the rule generating the data over its full domain, and therefore extrapolating beyond the training range. Theoretically, we prove that some EMs are latent-universal: with the update rule and interface held fixed, they can realize any partial computable function by varying only the initial condition of the latent state. Empirically, we study a zoo of minimal EM instantiations across discrete and continuous substrates, showing that local-recursive computation at a tiny scale (tens to hundreds of parameters) can extrapolate exactly on simple arithmetic functions, can support control behaviour and online adaptation, while still exposing several limitations. This work is foundational: it does not propose a competitive architecture, but a framework meant to widen the design space of machine learning beyond differentiable feed-forward maps.
计算语言学 (cs.CL)
24
cs.CL / 1 / 2608.13698
GRPO Beyond English: A Large-Scale Study of GRPO in Non-English and Multilingual Settings
Abstract
Reinforcement Learning with Verifiable Rewards (RLVR), often optimized with Group Relative Policy Optimization (GRPO), has become a central recipe for improving the reasoning capabilities of pretrained language models but current studies remain heavily English-centric. We conduct a large-scale empirical study of multilingual and non-English GRPO across a wide range of base models, training languages, and different reasoning language rewards. We find that training to reason in the native language often leaves only a small gap to training for English reasoning. We further observe strong crosslingual transfer: training in one language often improves performance in many others. However, specific trends are highly model- and language-dependent. In some cases, training in a particular language induces severe regressions on out-of-domain capabilities in other languages. Our analysis shows that RLVR beyond English can provide broad crosslingual gains, but also requires broad evaluation to detect language-specific regressions.
cs.CL / 2 / 2608.13706
CLAIR-Fin: An Adversarial Multi-Agent Framework for Claim-Level Verification and Adaptive Debate in Cross-Modal Financial QA
Abstract
Existing defenses against hallucination in retrieval-augmented and multi-agent pipelines remain partial: evidence is trusted despite modality disagreement, debate verifies an aggregate report rather than individual claims, and such verification occurs only after drafting, leaving inter-agent errors undetected until the final text. To close this gap, we present CLAIR-Fin, a nine-agent framework that decomposes each question into atomic claims maintained in a typed Financial Claim Ledger. Each claim is resolved through Asymmetric Evidence Authority, which conditions evidence trust on claim type rather than treating all modalities as equally reliable; Chain-of-Custody Verification, which checks grounding at the hand-off between drafting and adversarial review rather than only at the pipeline's exit; an Adaptive Rebuttal Cycle, which routes contested claims through adversarial debate whose depth scales with what that debate finds; and a terminal entailment audit paired with a continuous Hallucination Risk Index that distinguishes claims that passed scrutiny from claims never contested. We evaluate CLAIR-Fin on BB-FinQA-X, a 500-question cross-modal financial evaluation set built from Bangladesh Bank Annual Report material, stratified by query type, format, and difficulty. Relative to a single-pass retrieval-augmented generation baseline, it raises faithfulness ($0.780 \rightarrow 0.889$) while abstaining on 5.4% of questions when evidence is insufficient rather than forcing an unsupported response, and it exceeds stronger retrieval-strategy baselines such as HyDE and Graph-RAG on faithfulness ($\leq 0.874$).
cs.CL / 3 / 2608.13708
TeachMateGPT: A Multi-Agent Knowledge-Grounded Framework for Pedagogical Assessment Generation from Science Curriculum Materials
Abstract
Automatically generating textbook-grounded assessment items can reduce science teachers' workload, but existing retrieval-augmented generation (RAG) systems rely on flat retrieval, support only single-question generation, lack safeguards against weak evidence, and are ill-suited to low-resource, board-exam-structured curricula. We address these limitations with TeachMateGPT, a multi-agent system contributing four advances to curriculum-grounded science-assessment authoring. (i) COPE, a hierarchical knowledge base replacing token-window chunking with a multi-resolution index that segments documents along syllabus structure and links them at three granularities via a traversable graph-based lineage, matching evidence to each topic's instructional level. (ii) A staged, fail-closed agent pipeline replacing one-shot retrieve-then-generate: routing gates search, retrieval fuses dense and lexical evidence under a coverage gate that withholds generation on insufficient evidence, and specialist agents draft objective and constructed-response items. (iii) SAVER, a source-attributed verification protocol scoring faithfulness, relevance, and hallucination risk against retrieved evidence, applying stricter grounding checks across each creative question's four sub-parts, paired with teacher-in-the-loop evaluation rather than automatic filtering. (iv) NCTB-SciGen8, a curriculum-grounded dataset of 198 items (143 multiple-choice, 55 creative questions) spanning all 14 chapters of the NCTB Class 8 science textbook, produced by the pipeline and rated by three practicing teachers. TeachMateGPT raises faithfulness (0.68 $\rightarrow$ 0.96) and answer relevancy (0.60 $\rightarrow$ 0.89) over a vanilla RAG baseline.
cs.CL / 4 / 2608.13717
StreamHear: Domain-Adapted Pseudo-Labeling for Semi-Supervised Streaming Speech Recognition
Abstract
Streaming automatic speech recognition (ASR) underperforms on domain-shifted target audio, where labeled in-domain data is costly to prepare while unlabeled audio is abundant. We present StreamHear, a semi-supervised pipeline that adapts a pretrained streaming student by fine-tuning an offline transducer teacher on the labeled training set, generating pseudo-labels on the unlabeled portion, and fine-tuning the student on the mixture. We further introduce a prior-regularized dynamic-programming realignment step that fixes chunk-level word placement using an ASR-hypothesis anchor. Across four datasets spanning financial calls, prepared read speech, and phone-quality dialogue, StreamHear consistently outperforms supervised student fine-tuning and narrows the gap to the offline teacher.
cs.CL / 5 / 2608.13722
BM25-Augmented Many-Shot Translation for Low-Resource North-Eastern Indian Languages
Abstract
This paper describes the University of Florida Gators submission to the WMT26 Low-Resource Indic Language Translation shared task. We adapt the retrieval-augmented many-shot translation pipeline from our AmericasNLP 2026 system to translate between English and eleven North-Eastern Indian languages in both directions. At inference time, BM25 retrieves the most similar parallel examples from a language-specific training bank, and Gemini 2.5 Flash translates the input conditioned on these examples. No model fine-tuning is involved. Training banks combine official WMT26 data with publicly available corpora such as Samanantar and prior WMT shared task releases. A grid search over retrieval count r and development exemplar count d across all 22 language-direction pairs selects the best configuration for each submission.
cs.CL / 6 / 2608.13741
GALA: Generation-Aware Cross-Modal Alignment for Text-to-Time-Series Synthesis
Abstract
Synthesizing time series from natural language is emerging as the most expressive form of controllable time series generation. However, existing text-conditioned generators either take caption embeddings frozen from off-the-shelf text encoders, or adapt the encoder end-to-end, letting the denoising loss shape the embeddings only as a by-product. In either case, the conditioning representation is never deliberately matched to the signal modality, leaving it ill-suited to guide generation. We address this by introducing GALA: Generation-Aware cross-modaL Alignment for text conditional time series generation. GALA is a two-stage approach that first contrastively couples a pretrained text encoder with a time-series foundation model into a shared embedding space with both encoders adapted to generation by an auxiliary generative loss, and then freezes the resulting caption embedding to drive a flow-matching generator. On TSFragment-600K, spanning four domains and three fragment lengths, GALA sets a new state of the art, ranking first in 30 of 36 metric columns and reaching an average rank of 1.08/1.08/1.42 at lengths 24/48/96 against 1.92/2.00/1.75 for the strongest baseline. We further find that generator-internal text encoders force a trade-off between fidelity and caption adherence, whereas conditioning on the aligned embedding breaks it: FID, CTTP, and JFTSD all improve at once. Ablating the auxiliary loss degrades FID, CTTP and JFTSD together, it indicates the generative term is a necessary component of the alignment rather than an add-on.
cs.CL / 7 / 2608.13760
Amplified Does Not Mean Predictive: Reasoning Behaviors in Thinking Models
Abstract
Which reasoning behaviors are associated with correct answers in reasoning models, and does reasoning-oriented training amplify those behaviors? This distinction is important because reasoning-oriented training can make traces look more deliberative without amplifying the behaviors most tied to model correctness. We quantify this mismatch with Behavioral Lift, a metric that measures how much correctness changes when a behavior is present versus absent in a model's reasoning trace. Across 15 models and 6 benchmarks spanning text-only and vision-language reasoning, we annotate 15,282 traces with a taxonomy whose core behaviors are defined for both LLM and VLM traces. We find evidence for an Amplification-Lift Gap, in which thinking models strongly amplify self-correction, hypothesis testing, and uncertainty acknowledgment, while the highest-lift behaviors are confidence calibration, knowledge alignment, and self-awareness. Confidence calibration is among the strongest positive signals of correctness in both modalities, yet is barely amplified; uncertainty acknowledgment is amplified by 3--7$\times$, yet is weakly or negatively associated with correctness. We find that reasoning-oriented training does not preferentially amplify the highest-Lift behaviors, motivating process-level objectives that reward calibrated and grounded reasoning rather than surface form alone.
cs.CL / 8 / 2608.13835
When Lexical Change Misleads: Rethinking Dynamic Topic Model Evaluation with Traditional and LLM-Based Metrics
Abstract
Dynamic topic models capture evolving word distributions, but traditional coherence metrics may fail when vocabulary changes while semantic meaning persists. We evaluate 120 topics from CoNTM and DLDA across NYT, DBLP, and arXiv, using three human annotators and Low, Medium, and High lexical-change categories. Traditional temporal coherence shows highly variable agreement with human judgments ($ρ$=-0.256 to 0.614). In contrast, LLM-based semantic similarity agrees strongly with human semantic judgments for CoNTM on NYT ($ρ$=0.609), DBLP ($ρ$=0.721), and arXiv ($ρ$=0.502), but is less consistent for DLDA. Lexical-change stratification reveals variation hidden by aggregate evaluation. We therefore advocate lexical-change-aware evaluation, jointly reporting traditional coherence and LLM-based semantic measures as complementary rather than interchangeable signals.
cs.CL / 9 / 2608.13840
ASSERT: A Measurement Pipeline for GenAI Audits
Abstract
Audits of generative AI (GenAI) systems often summarize behavior as a reported rate: how often the audited system complies with policy. Researchers and stakeholders use that rate to compare systems, track regressions, and gate deployment. A reported rate reflects both the system under audit and the measurement choices behind it, so a change in the rate can leave it unclear whether the system or those choices moved. We introduce ASSERT, a specification-driven measurement pipeline for GenAI audits that ties each reported rate to a written specification of the measurement choices used to produce it. ASSERT helps draft a behavioral rubric and test cases, then runs the audit against a GenAI system and returns a reported rate. In a case study on conversational deception, we observe that the reported rate moves substantially with the dialogue setup, the simulated user, the judge, and the evidence bar for non-compliance. These measurement choices substantially change the reported rate and can reorder GenAI system rankings. Because each reported rate is tied to an explicit specification, differences across audits are easier to attribute and interpret.
cs.CL / 10 / 2608.13854
Bootstrapping Niche Multilingual Code Translation via Reinforcement Learning with Execution-Based Verifiable Supervision
Abstract
Code translation must preserve executable behavior across many programming languages, yet neural code translation has largely focused on a few popular languages such as C++, Java, and Python. This leaves a niche, many-to-many setting where parallel supervision is sparse, producing plausible but non-executable translations. We address this setting with preference-based reinforcement learning driven by execution-based supervision. Our pipeline firstly expands verifiable seed Python programs into a multilingual pool of execution-validated codes. Using the pool, a base LLM generates translation candidates across language pairs, which we label by their execution outcomes. The resulting preferences are used to train a reward model that scores cross-language translation quality. Finally, we optimize our base LLMs with GRPO over 600 directed language pairs (25 x 24) using the reward model as a signal. To evaluate the niche translation capability, we introduce HumanEval-X++, an execution-based benchmark that extends HumanEval-X to a broad many-to-many language space. We evaluate our approach using Qwen-3.5 4B and 9B models. On HumanEval-X++ and existing benchmarks, it yields consistent gains over the untrained baselines. In particular, the 4B model achieves an average improvement of 13% across all languages on HumanEval-X++, with a gain of 21% on mid-tier languages. Our study establishes a reliable approach of data generation, training, and benchmarking, paving the way toward further bootstrapping the quality of many-to-many translation for programming languages.
cs.CL / 11 / 2608.13959
Repair, Not Improvement: Decomposing Constrained Decoding in Tool-Call Abstention
Abstract
Function calling is what the recent accounting of constrained generation explicitly sets aside: it finds the decoder's contribution small for format constraints, then warns in its Section 7 against extrapolating where a constraint encodes a correctness requirement, and names function calling as one. Tool abstention is that case at its sharpest: an enum leaves the wording of an answer alone and narrows the set of answers there are, and declining to call anything is the first it drops. We measure the excluded case. Three conditions over one byte-identical prompt separate a grammar's two jobs: it fixes where generation stops as well as which tokens may be emitted. We evaluate open-weight models from 0.6B to 4B on matched English and Korean items, so the language comparison is made within item. Against an unconstrained decoder, prior work's contrast is negative on abstention in four of six cells with intervals excluding zero, worst -29.5 points, and positive with an interval excluding zero in none. The total is a sum with opposite signs: on the smallest model in Korean the stop token costs -20.0, the enum returns +19.5, and the two leave -0.5. What it recovers is form: of 698 abstentions repaired, 545 had no readable answer and 0 were judgements the scorer refused. On tool-needed items it is positive throughout; abstention leads because it is the preregistered measure, and the pooled number being kinder to the intervention makes moving to it worse rather than better. Both preregistered language claims fail.
cs.CL / 12 / 2608.14003
Batch-wise Adaptive Pruning: Periodic Neuron Activation-Aware Weight Pruning for Language Reasoning Model
Abstract
Large Reasoning Models (LRMs) achieve strong performance on complex tasks through extended chain-of-thought generation, but incur substantial computational costs during inference. In production settings, batched inference is essential for high throughput, yet the existing training-free adaptive pruning methods we evaluate severely degrade in this regime. Because a batch must share a single pruning mask, these methods aggregate activations across samples and then apply threshold-based selection; the threshold, calibrated offline on unaggregated activations, no longer matches the aggregated distribution, so the realized sparsity ratio drifts and accuracy on reasoning tasks collapses under batched inference. In this work, we propose a training-free adaptive pruning method designed specifically for batched inference in LRMs, built on two components. First, we replace threshold-based selection with periodic top-k selection over the aggregated importance scores, which is unaffected by the shift that aggregation induces in the activation distribution, and which runs selection once per update period rather than at every token, preserving the speedup. Second, based on the observation that important neurons re-fire periodically during long reasoning generation, we introduce an activation memory that accumulates importance across update phases so that recurring neurons are retained. Experiments on diverse reasoning benchmarks demonstrate that our method outperforms the previous state-of-the-art adaptive pruning method by 39.7 percentage points in average accuracy at batch size 4 with 50% target sparsity on DeepSeek-R1-Distill-Qwen-7B, and reaches 1.40x speedup over dense inference at 50% actual sparsity.
cs.CL / 13 / 2608.14029
S2Dialog: Multimodal Dialogue Retrieval with Semantic and Acoustic-Style Modeling
Abstract
Multimodal dialogue retrieval aims to retrieve dialogues from multimodal dialogue banks that are similar to a target dialogue in terms of both textual semantics and acoustic conversational styles. Such dialogue-level retrieval is crucial for many dialogue-related tasks, including Emotion Recognition in Conversation, Spoken Dialogue Systems, and Conversational Speech Synthesis, where external dialogue examples can provide valuable semantic and stylistic references. However, existing retrieval methods are still largely limited to utterance-level or unimodal matching, and often fail to capture the global semantic coherence and stylistic consistency of an entire dialogue. To address this gap, we propose S2Dialog, a unified framework for dialogue-level semantic-style retrieval from multimodal dialogue banks. Specifically, S2Dialog consists of a Dialogue-level Textual Retriever and a Dialogue-level Acoustic Retriever, which encode the textual and acoustic modalities of a dialogue into dialogue-level representations, respectively. To further enhance multimodal retrieval, we introduce Dialogue-level Textual-Acoustic Contrastive Learning, which aligns semantically and stylistically similar dialogues while distinguishing unrelated ones. Extensive experiments on the multimodal dialogue dataset DailyTalk demonstrate that S2Dialog achieves outstanding retrieval performance.
cs.CL / 14 / 2608.14079
The conditional superiority of fast silicon sampling
Abstract
Silicon sampling can produce surprisingly good population estimates at times. Does doing it fast attenuate such fidelity? In this study, we extend and assess ongoing work in silicon sampling by comparing the algorithmic fidelity of "fast" and "slow" modes of silicon sampling among a nationally representative sample of Singaporean survey respondents. We find that silicon sampling with contemporary frontier models remains a method in early development to be used only with great caution. While silicon samples are able to produce moderately faithful estimates of population means, they continue to understate opinion variance and distort the latent contextual space behind human opinions. Conditional on such limitations, we find "fast" modes of silicon sampling to be relatively superior to traditional "slow" modes of silicon sampling. Fast silicon sampling is significantly more efficient in compute resources and run-time while being monotonically superior to slower modes of sampling in algorithmic fidelity.
cs.CL / 15 / 2608.14150
Leading-Silence Augmentation and Multi-Stage Synthetic Supervision for the Second MLC-SLM Challenge
Abstract
The second Multilingual Conversational Speech Language Model (MLC-SLM) Challenge evaluates two tasks over complete, unsegmented multilingual conversations: speaker diarization and recognition (Task 1) and conversational speech understanding (Task 2). Neither task provides oracle utterance boundaries or speaker labels at evaluation, and Task 2 provides no question-answer training set. For Task 1, we fine-tune VibeVoice-ASR-7B with random leading-silence cropping, consistent timestamp correction, and an exponential moving average (EMA) training strategy. For Task 2, we construct synthetic question-answer pairs through multimodal candidate generation, silent-audio filtering, and distribution-matched augmentation, and fine-tune Qwen3-Omni-30B-A3B-Instruct for tagged direct answering. On the Task 1 evaluation set, cropping reduces tcpMER from 18.30% to 17.27%, and EMA further reduces it to 16.73%. On the Task 2 evaluation set, jointly applying distribution-matched augmentation and tagged direct answering raises accuracy from 83.0% to 86.0%.
cs.CL / 16 / 2608.14210
How Much Do Legal RAG Systems Still Hallucinate?
Abstract
Hallucination is a major challenge for retrieval-augmented generation (RAG) systems in the legal domain, where ungrounded answers can lead to serious consequences. To better understand this problem, we conduct a fine-grained analysis of hallucination behavior in eight legal RAG systems across two legal corpora, the GDPR (in English) and a national civil law (in French). Using claim-level and answer-level evaluation, we report on hallucination density and severity, analyze performance across question categories and user personas, and validate our findings on an independent set of 142 legal-expert-authored questions. Our results show that hallucinations remain pervasive, ranging from less than 10% of responses for the best-performing systems to nearly half in the worst case. We further find that false-premise questions, containing incorrect assumptions that must be rejected, produce high hallucination rates on the manually-drafted questions.
cs.CL / 17 / 2608.14229
The More Popular, The Harder to Forget: Adaptive Popularity for LLM Unlearning
Abstract
Popular facts are memorised more deeply during pretraining and resist removal longer than rare ones, yet existing LLM unlearning methods apply uniform gradient pressure regardless of training-data frequency. We propose the AdaPop (Adaptive Popularity) method, which combines local token confidence with a per-fact popularity-dependent exponent derived from an external proxy (e.g., Wikidata sitelinks, LLM-as-Judge), and automates the forget-retain balance via a dual-ascent controller that adjusts the retain penalty each epoch. Across three model families and two benchmarks, AdaPop leaks ~5x less forgotten content than competing methods under paraphrased queries and ~1.6x less under adversarial reformulations. We support our analysis with internal metrics: under our method, forget-set hidden states move further from the pre-unlearning model's states than under other methods, while retain-set representations remain close.
cs.CL / 18 / 2608.14277
SimpleOPD: Simple Tokenizer-Agnostic On-Policy Distillation for Long-Context Reasoning
Abstract
On-policy distillation (OPD) offers a promising way to transfer reasoning capabilities from stronger teacher models, but applying it to long-context reasoning teachers and short-context students introduces practical challenges, including tokenizer mismatch, teacher-student distribution mismatch, response length explosion, and training instability. In this work, we study this setting by transferring proof-reasoning capabilities from the long-context reasoning model SU-01 to short-context student models. To handle tokenizer differences, we perform OPD in a shared text space and align only tokens that occupy identical text spans under the student and teacher tokenizers. To mitigate the problem of excessive generation length and frequent truncation, we introduce a student reference KL loss and mask the advantages of special termination tokens such as </think> and <|im_end|>. This strategy constrains the student from drifting excessively from its initial policy, thereby mitigating the teacher-student distribution mismatch problem and fostering steady length growth. Experiments on both same-family and different-family student models, including Qwen3, Qwen3.5, Intern-S2, GLM-4.7, Gemma-4, show consistent gains in mathematical reasoning, especially natural-language math proving. Notably, Intern-S2-Preview improves by 21.2 points on ProofBench, reaching 55.2 and surpassing Gemini-2.5-Pro. It also improves on science benchmarks such as HLE and HiPhO, suggesting that OPD transfers reasoning capabilities that generalize beyond the mathematical training domain.
cs.CL / 19 / 2608.14312
Envs-FORGE: Frontier-Optimized Reward-Grounded Environment Synthesis for Agent RL
Abstract
Reinforcement learning (RL) for terminal agents needs executable training environments with reliable rewards and useful difficulty. Fixed recipes such as few-shot, Self-Instruct, and Evol-Instruct apply the same prompting policy to every seed, even when the current policy would benefit from a harder, easier, or simply different task. We present Envs-FORGE, a prompting policy that converts verifier rewards into per-seed environment-synthesis actions. Envs-FORGE estimates seed pass rates, scores six projection--direction actions around a target learning frontier, and solves a per-seed mixed-integer linear program (MILP) to choose the action that conditions generation. The selected action drives synchronized rewriting of the instruction, fixtures, oracle solution, tests, and Docker environment; only gold-verified bundles enter RL training. The indexed MILP form also supports optional soft skill coverage for portfolio planning. On Qwen 3.5 35B, Envs-FORGE improves Pass@1 over Base by 9.2 percentage points on tb-core (40.0% to 49.2%) and 6.4 points on tb-2.0 (23.0% to 29.4%), exceeding the strongest fixed-recipe baseline by 2.4 and 2.1 points. It reaches 77.1% on SWE-bench Verified versus 73.4% for Base, and improves tb-core by 6.8--9.2 points across the evaluated 4B--35B models. All synthesis methods export 100 verified environments and use 2.27M--2.88M synthesis tokens, placing the comparison at the same downstream training-set size and the same operational scale. The source code is available at https://github.com/DataArcTech/DataArc-SynData-Toolkit/.
cs.CL / 20 / 2608.14361
Local and Global Regimes of Geometric Complexity in Language Model Representations
Abstract
Intrinsic dimensionality (ID) is widely used to probe the representational complexity of language models, but it remains unclear whether ID differences reflect properties of language itself or artefacts of how the underlying dataset was constructed. In this paper, we focus specifically on how lexical diversity, the number of unique last-token items present in a dataset, affects ID estimates of that dataset. We find a scale-dependent transition between two regimes: at low lexical diversity, conditions with fewer unique final words produce higher ID, while at high lexical diversity, this ordering reverses, and conditions with more unique words produce higher ID. We derive an exact, parameter-free formula for the point at which this reversal occurs, which matches the observed transition point at every scale tested. On the one hand, our results highlight how care must be taken when interpreting the intrinsic dimensionality of a set of representations as a straightforward cue of their complexity. On the other hand, our discovery of the two ID regimes reveals a general principle of organisation of linguistic data in LLMs that sheds new light on their inner manifold structures.
cs.CL / 21 / 2608.14457
Information Satisfaction: A Reader-Centered Axis for Summarization Evaluation
Abstract
The majority of work on summarization evaluation focuses on general summary quality (e.g., ROUGE, BERTScore) or specific desired properties (e.g., readability, factuality). However, these metrics fail to measure the utility of a summary to an individual user. For example, a biomedical researcher learning about the latest vaccine research will have different informational needs from a family doctor. Query-focused summarization captures part of this need, but in practice, users rarely state everything relevant in a query: a single short query is likely inadequate to distinguish the needs of a researcher from those of a physician. By contrast, a reader's background or persona (their role and expertise) is comparatively stable across queries and recovers much of this missing context, which makes it a practical signal for assessing whether a summary satisfies that reader's needs. In this work, we assess how sensitive popular summarization metrics are to both informational and persona differences, and find that many popular metrics, including strong LLM-as-judge metrics, fail basic perturbation tests of informational content. We additionally conduct an expert human evaluation, measuring summary preferences based on information satisfaction given a specific person's background and use case. We find that both traditional and LLM-based metrics are insufficient measures of information satisfaction and agree poorly with human judgment.
cs.CL / 22 / 2608.14465
You Only Pass Once: Answering and Abstaining Together in a Single Forward Pass of a Frozen Language Model
Abstract
A frozen language model on reasoning tasks has two coupled weaknesses: it under-uses evidence its own residual stream already encodes, and it fails to detect when the input is insufficient to answer, so it confabulates. This paper consolidates two research lines that address these on the same residual stream: a conditional steering probe writes the stream at mid-stack layers and recovers reasoning accuracy from a frozen backbone, and a zero-shot sufficiency direction reads the stream and abstains when information is insufficient. Deployed in one forward pass they interfere: the steering write shifts the state the direction reads, costing up to 8 AUROC points of cross-domain transfer on small models; a separate clean pass doubles inference cost. We keep the direction fixed and train a small network to reconstruct the pre-steering residual from the steered one -- mean-squared error on (steered, clean) pairs, no sufficiency labels -- and read the direction on the reconstruction. The resulting system, YOPO (You Only Pass Once), answers, steers, and abstains in one forward pass of a frozen Qwen2.5 backbone (1.5B/3B/7B). End to end, three-way accuracy more than doubles the frozen baseline (0.375->0.798 on 1.5B alphaNLI) and one pass beats the two-pass reference at every scale (0.798/0.830/0.893 vs 0.753/0.790/0.863) and on ten backbones across six model families. We chart the capacity-transfer frontier quantifying the principle that abstention should not be trained in; a source-side audit catches our own alphaNLI construction leaking a surface artifact, so architectural claims are anchored on native-label replications (SQuAD2, RepLiQA, MuSiQue); and on the standard four-domain suite we contribute, to our knowledge, the first answer-or-abstain benchmark, where our gate tops every in-domain dataset and the label-free direction is the only gate family to survive domain transfer.
cs.CL / 23 / 2608.13674
Asymmetric Discourse Homogenization and Shared Language Technology: Evidence from Reddit
Abstract
I document an ideologically asymmetric break in the pre-existing diversification trend of political discourse, emerging around late 2022, using 6 million Reddit comments from two cross-partisan forums, 2019-2025. Conservative users experienced an interruption of their prior diversification trajectory; progressive users showed no comparable change. The asymmetry is consistent across estimation strategies (ITS, DiD, RDiT, propensity-score matching) and temporal aggregations. A daily-frequency permutation test over 2,377 candidate cutoff dates shows the ChatGPT threshold produces an unremarkable estimate (49.8th percentile): the shift builds gradually instead of breaking at a single date. A continuous cumulative LLM index, tracking AI exposure across seven model releases, remains significant under a quadratic trend specification that eliminates the binary estimate. A stayer analysis narrows the mechanism: the homogenization effect disappears when the sample is restricted to authors active throughout the study period, and the stayer confidence interval excludes within-author effects even a tenth the size of the full-sample estimate. The mechanism is most parsimoniously ecological (community-level discursive convergence) rather than individual-level AI adoption, though the data cannot cleanly separate this account from concurrent secular change.
cs.CL / 24 / 2608.13831
VoiceChat-TTS: A Low-Latency Continuous Speech Synthesis Model for Interactive Agents
Abstract
Spoken dialogue is a natural form of human--computer interaction, yet most speech language models remain limited to turn-based operation and lack real-time adaptability, such as user barge-in. Recent duplex speech-to-speech and speech-to-text models reduce latency by replacing multi-stage pipelines, but often compromise speech quality because accurate ASR, interruption handling, and high-fidelity synthesis must be optimized jointly. We propose VoiceChat-TTS, a low-latency, continuous, and streamable text-to-speech model for interactive agents. VoiceChat-TTS is driven directly by LLM text-token streams, supports explicit interruption via control tokens, and produces silence when no textual input is available. The model enables always-on, responsive speech generation while preserving modularity and high speech quality, and it supports mid-utterance interruptions without resetting the KV cache.
多智能体系统 (cs.MA)
2
cs.MA / 1 / 2608.13982
XAI-Guided Conservative Decentralized Execution for Offline Multi-Agent Network Slicing
Abstract
The recent advances toward sixth-generation (6G) and beyond-6G networks have accelerated the need for intelligent resource management mechanisms capable of supporting heterogeneous services under shared infrastructures in network slicing. However, resource allocation in network slicing naturally forms a resource-coupled cooperative optimization problem with competing slice demands. Slices compete for limited resources to minimize individual latencies while coordinating to avoid conflicts and underutilization. Although multi-agent reinforcement learning (MARL) has shown promising performance in such settings, existing online formulations remain costly, unsafe, and difficult to deploy due to their reliance on environmental interactions and communication among agents. In this work, we present explainable artificial intelligence (XAI)-guided conservative decentralized execution (X-CODE). X-CODE is an explainable offline MARL that operates offline without environmental interaction, nor inter-agent communication. It exploits explainability-aware reward shaping to modify the relative preference among joint offline transitions during centralized training to improve decentralized resource-allocation behavior. In deployment, the agents operate independently without signaling exchange among the agents. Simulation results demonstrate that the proposed approach achieves zero observed resource-conflict events in the evaluated test episodes while minimizing per-slice latencies. Moreover, the proposed framework exhibits lower signaling overhead and reduces effective inference latency by 88 % under the considered communication-delay model compared to the online baselines. Source codes and datasets are available through: https://github.com/Eslam211/xcode-ran-slicing.
cs.MA / 2 / 2608.14390
Submodular Policy Learning for Distributed Task Allocation in Open Multi-Agent Systems
Abstract
This paper studies policy learning for distributed task allocation in open multi-agent systems, where agents may join and leave in a time-varying fashion, with submodular stage team utilities. At each time, the active agents select actions from local categorical policies such that the feasible joint agent-action pairs form a partition matroid. Standard continuous relaxations of submodular set functions are based on independent Bernoulli sampling, making them inconsistent with agents' policies.To solve this mismatch, we propose the \emph{partition multilinear extension} (PME), a policy-based relaxation whose continuous support matches feasible actions under categorical policies.We prove that the marginal gains of the stage utility provide an unbiased estimator of the gradient of the PME and that maximizing the PME over action distributions is equivalent to maximizing the stage utilities over agent actions, which are critical to devise principled policy gradient.Building on this, we design \emph{SubMAPL}, a centralized-training decentralized-execution KL-mirror policy-learning method that uses local marginal gains as stochastic PME gradients during training. KL-mirror updates preserve categorical feasibility without Euclidean projection.In the case where agents run tabular-softmax policies, we introduce open policy migration and an open-system KL tracking variation to handle agent arrivals and departures. Using dynamic regret analysis, we establish a lower bound on the cumulative utility which accounts for the openness of the environment and for the gap between optimal stage-wise and global utilities. Simulations on multi-agent coverage demonstrate that SubMAPL outperforms policy-gradient and online-learning baselines.
软件工程 (cs.SE)
6
cs.SE / 1 / 2608.13730
Building AI-Intensive Software with AI: Early Results and a Cautionary Tale on Measuring Development Cost
Abstract
Empirical reports on the true cost of AI-intensive software development remain scarce, and the few that exist are easy to get wrong in ways that never surface in the final number. We report early results from an ongoing case study: a six-person student team built a full conversational onboarding assistant -- RAG-based code chat, guided tours, dependency graphs, technical-debt analysis -- over one academic term using pervasive AI assistance. We instrumented development with a three-layer cost model (real AI spend, self-reported human effort, human counterfactual) and initially reported a 19.4x cost ratio. A follow-up pass revealed two independent errors -- inferring per-token cost under a flat-rate subscription, and pricing the counterfactual with the wrong regional labor rates -- that together had inflated the ratio by roughly 2x; the corrected figure is ~9.9x. We present this correction as an early, generalizable finding in its own right: both errors are easy to make, invisible in the final number, and plausibly common in similar reports. We outline next steps toward a more robust, replicable costing methodology for AI-intensive development.
cs.SE / 2 / 2608.13742
Does ISO-Grounded NFR Specification Improve LLM Code Generation? A Comparison of Rich and Structured Interventions against a Natural-Language Baseline
Abstract
In LLM-based code generation, Non-Functional Requirements (NFRs) are often specified as terse one-line phrases. We ask whether grounding those specifications in ISO/IEC 25010 Quality Model, either as rich natural-language prose (NL-rich) or as structured JSON (Structured), improves code generated on HumanEval/HumanEval-ET compared to a RobuNFR-style one-line baseline (NL-simple). We evaluate four NFRs (performance, error handling, code smell, readability) with ten prompt variations per condition under a fixed model snapshot and paired non-parametric analysis. Primary finding: ISO-grounded enrichment improves static quality proxies (unreadability density falls across all four NFRs (e.g., Performance 0.88 -> 0.69 for NL-rich)) and reduces sensitivity to prompt wording, but does not reliably improve functional correctness; for error handling, extended-test pass rate decreases, suggesting tension between defensive coding patterns and exact-output benchmarks. Secondary finding: when ISO content is held constant, NL-rich and Structured differ negligibly in correctness (|delta| <= 0.023), indicating that semantic content matters more than JSON-vs-prose format. Practitioners should invest in standard-grounded NFR content rather than serialization form. A fully traceable replication package is provided.
cs.SE / 3 / 2608.13867
Engineering Reliable Coding Agents: Evaluating and Operating the System Around the Model
Abstract
AI coding agents are commonly evaluated as models but deployed as systems. Their reliability depends not only on model capability, but on the harness, execution state, retrieval, memory and state management, permissions, review interfaces, and resource allocation. This monograph examines those boundaries and develops a framework for evaluating and operating coding agents reliably. It synthesizes 164 scholarly works, 100 practitioner records, 29 benchmark records, and 17 author-system case records through a structured multivocal review, targeted update audits, software-engineering coverage analysis, and distributed-systems evidence synthesis. Across this evidence, many apparent model failures originate elsewhere in the system, while improvements at one layer often fail to propagate to end-to-end outcomes. Evaluation and operation are treated as a dependency chain in which weaknesses in task construction, execution environments, retrieval, state management, verification, or observability can invalidate downstream conclusions. The monograph contributes a versioned catalog of 206 reliability records: 193 gated practices, including 56 developed in depth, plus 13 research leads; an evidence ledger; a framework for dependency and repair asymmetry across the agent lifecycle; measurements and failure cases from operated agent systems; runnable evaluation and reliability protocols; and five reusable agent skills with evidence maps. Together, these provide a system-level methodology for distinguishing model capability from infrastructure effects, designing defensible evaluations, and building systems that recover safely when components fail. The review is structured rather than exhaustive, evidence strength varies by topic, and results depend on workload and configuration. The methods record which search lanes were executed, which remain unexecuted, and limits on evidence-grading claims.
cs.SE / 4 / 2608.13884
Engineering Signals of Human-AI Collaboration in the Agentic Coding Era: A Longitudinal Analysis of 33,228 Pull Requests from vLLM and SGLang with Implications for Biomedical AI Agents and Bioinformatics Pipeline Developmen
Abstract
The rapid adoption of AI coding assistants and autonomous agentic development systems has coincided with major changes in the pace and structure of open-source software engineering. Yet empirical longitudinal evidence of these changes at the team level remains limited. We present a descriptive longitudinal analysis of seven engineering metrics: pull request (PR) throughput, cycle time, contributor diversity, PR comment density, merge rate, new-author participation, and PR size. Metrics were computed from all merged PRs in two high-velocity AI infrastructure repositories, vLLM (February 2023-June 2026; 18,290 PRs) and SGLang (January 2024-June 2026; 14,938 PRs). We segment development into four eras aligned with major changes in AI-assisted software development and examine human- and bot-authored activities. Both projects show substantial increases in development velocity and AI-developer collaboration signals. PR throughput increased 21x in vLLM and 17.9x in SGLang, while bot-authored PRs accounted for less than 0.2% of this growth, indicating that the increase was overwhelmingly human-driven. In the latest era, median cycle time was 1.04 days for vLLM and 0.62 days for SGLang, while P90 cycle times reached 16.8 and 14.3 days, respectively. Monthly unique authors increased steadily in both projects, suggesting broader contributor participation. PR comment density increased 4.2x in vLLM and 3.8x in SGLang, with bot comments contributing an estimated 15-20% of the increase. In contrast, PR size remained relatively stable across eras. Overall, AI-assisted development is associated with higher throughput, broader contributor participation, and increased AI-developer collaboration signals in high-velocity open-source software development.
cs.SE / 5 / 2608.14131
LegacyWorld: Atomicity-Aware Evaluation of GUI Agents for Legacy Workflows
Abstract
Legacy and legacy-like enterprise systems often remain difficult to modernize because critical workflows expose limited programmable interfaces and still require manual GUI interaction. This paper reports a pre-deployment evaluation study motivated by the development of legacy-use, an industry-oriented framework for automating such workflows with multimodal LLM agents. During framework development, domain experts helped identify stateful workflows where successful demos are not sufficient: a failed agent run may still leave persistent invalid changes in business or healthcare records. We therefore evaluate computer-use agents using atomicity: a run should either complete the intended workflow correctly or fail without unintended persistent side effects. We construct a domain-expert-informed benchmark of 28 Windows GUI workflows, each specified with an initial state, goal state, and task-specific validator. We compare expert-crafted prompts with prompts generated from screen recordings of expert golden-path executions. Across six hosted computer-use agents, our results show that useful completion, safe failure, and non-atomic side effects are distinct operational profiles. We conclude that workflow capture, state validators, and atomicity-aware acceptance tests should be first-class requirements for AI-based legacy workflow automation.
cs.SE / 6 / 2608.14315
Breaking Models to Test the Judge: A Mutation Testing Approach for Semantic Evaluators of Domain Class Diagrams
Abstract
In software engineering, many semantic modeling tasks lack a unique ground truth, as human judgments are both costly and subjective. This paper explores mutation testing as a scalable alternative for evaluating semantic judges (e.g., LLM-based) of models. We propose a mutation testing approach in which controlled semantic defects are injected into domain class diagrams. Starting from pairs of PlantUML class diagrams and textual system descriptions, we apply mutation operators (e.g., removing a class) to generate faulty variants. A candidate judge is then evaluated based on its ability to detect the injected defects. We define 11 mutation operators for the task of comparing a domain class diagram against a textual description and evaluate the proposed approach against a conventional manual assessment of judgment validity. Across six judge configurations (three LLMs and two prompt variants), the automated mutation testing approach is largely consistent with the manual assessment in identifying the better-performing configurations. The results suggest that mutation testing may serve as a scalable proxy for analyzing semantic judges.
操作系统 (cs.OS)
1
cs.OS / 1 / 2608.13689
A Bounded Reclaim Actuator for PSI-Guided Compressed Memory: A Controlled Ablation
Abstract
When the aggregate working set of active processes exceeds physical RAM capacity, the machine experiences memory pressure. Applications may therefore slow down before the kernel kills a process. Linux provides several ways to observe and respond: Pressure Stall Information (PSI) can detect memory-related task stalls, zram can provide compressed in-memory swap space, and cgroup v2 can request memory reclamation within a selected control group. These facilities are often discussed together even though they act at different points in the pressure path. This paper examines that distinction with a controlled systems study. We compare three setups: zram enabled from startup; zram enabled only after PSI indicates memory pressure; and zram enabled from startup with a one-time 96 MiB cgroup reclaim request. We first selected the request size in a 16-case pilot, then ran 180 confirmatory cases, 60 cases for each setup, on nine 1-vCPU Linux virtual machines with compute and SQLite workloads. Compared with static zram, the bounded reclaim configuration reduced compute p99 response time by 6\%, while the SQLite result was statistically indistinguishable. Delayed activation had higher median p99 latency than both alternatives. These results suggest that the benefit depends on the foreground workload and its memory-access path, rather than a general improvement across workloads.
硬件架构 (cs.AR)
5
cs.AR / 1 / 2608.13962
MoE Expert Execution in Disaggregated LLM Serving with a High-Bandwidth ReRAM Near-Memory Architecture
Abstract
Attention-FFN disaggregation maps LLM modules to specialized pools, creating an opening to keep Mixture-of-Experts (MoE) weights resident in a high-bandwidth FFN pool. Decode SLOs, however, cap the run-batch while sparse routing expands the activated-expert union, so weight traffic amortizes poorly and routing skew idles cold-expert resources. The FFN pool must therefore deliver weight-read bandwidth density under sparse unions and recover occupancy under skew without a global sharing fabric. We present a ReRAM near-memory architecture that keeps expert weights resident behind high-bandwidth local reads. The design factors actual MFU into ideal MFU and occupancy, recovers occupancy with bounded core-local multicast pooling, coactivation-aware placement, and load-aware fetch, and sizes each communication level from induced demand. A measured + modeled study on Qwen3.5-35B-A3B, Qwen3.5-397B-A17B, and GLM-5.2 shows that side-4 pooling raises occupancy from 0.328 to 0.519 and, at iso-peak compute, lowers per-token FFN-pool latency by 9.5x versus H20 with 20x lower weight-movement energy; an H20-attention + ReRAM-FFN system reduces decode TPOT by 1.25-4.0x, 2.4-10.3x, and 2.5-10.4x versus a homogeneous H20 pool.
cs.AR / 2 / 2608.14073
Experimental Study on System-Level Performance Impact of Read Disturbance in Modern SSDs
Abstract
This work investigates the system-level performance impact of read disturbance in modern NAND flash-based SSDs, aiming to provide new insights that can help develop better storage architectures and optimize system software. Continuous improvement in storage density over decades has led NAND flash memory to play a vital role in modern computing systems, but it also comes at a cost of significant reliability degradation. Among various error sources, read disturbance has gained growing attention as a major reliability concern due to its rapidly increasing impact, which can significantly affect system I/O performance by exacerbating SSD-internal reliability-management overheads. Although a large body of prior work has focused on device-level characterizations and optimizations, the system-level performance impact of read disturbance still remains largely uninvestigated. To address this gap, this work conducts a rigorous experimental study using 15 modern NVMe SSDs from 10 major vendors in two ways. First, we comprehensively analyze the system-level performance impact of read disturbance under diverse workloads and operating conditions. Second, to highlight the importance of efficient read-disturbance management, we showcase a new possible SSD-performance attack as a case study, demonstrating that an adversary can significantly degrade the I/O performance of other concurrently running processes by exploiting read disturbance alone in commodity SSDs. Based on our experimental study, we make 16 new observations and 7 takeaway lessons, which lead to 6 key directions for future improvements at the host-system and SSD-architecture levels to better cope with read disturbance.
cs.AR / 3 / 2608.14333
Beyond Capacity: Scalable MoE LLM Inference via High-Bandwidth Flash with Direct GPU and HBM Paths
Abstract
Modern mixture-of-experts (MoE) language models increasingly strain the capacity and cost efficiency of high-bandwidth memory (HBM), as rapidly growing expert weights must be provisioned close to GPUs. High-bandwidth flash (HBF) offers substantially greater capacity, but conventional designs typically deliver HBF-resident expert weights to the GPU through HBM, leaving an additional direct GPU-HBF connection underutilized. We explore an HBF organization that simultaneously exploits two independent expert-delivery routes: a direct path that transfers expert weights from HBF to the GPU and a relay path that transfers them from HBF through the HBM base die to the GPU. Whole experts are assigned to one of the two routes, and transfers over both routes proceed concurrently, increasing aggregate expert-delivery bandwidth without replicating expert weights or introducing a shared relay bottleneck. Early expert determination identifies upcoming experts ahead of their conventional execution point, allowing HBF read latency to overlap with preceding computation, while separate management of immutable expert weights and mutable KV-cache data reduces interference between the two traffic classes. We evaluate the architecture using an event-driven continuous-batching LLM serving simulator with empirically measured GPU compute latencies. Across representative MoE workloads, concurrently utilizing the direct GPU-HBF and HBF-HBM-GPU routes consistently improves expert-delivery efficiency over designs restricted to either route alone. For a representative workload, the proposed architecture can achieve 1.94$\times$ higher throughput and 1.90$\times$ end-to-end speedup over a design that delivers all HBF-resident expert weights to the GPU through the HBM base die.
cs.AR / 4 / 2608.14217
The Quartic Hessian Conjecture in Dimension Four
Abstract
The Hessian conjecture asks whether a polynomial with nonzero constant Hessian determinant has a polynomial gradient inverse. It is known in dimensions at most three, false in dimensions at least five, and open in dimension four. We prove its four-variable quartic case. The top homogeneous part has zero Hessian determinant and, by the four-dimensional homogeneous Hesse theorem, is a cone. We divide its cone representative into three exhaustive types: a genuinely ternary quartic with nonzero ternary Hessian, a genuinely binary quartic, and a fourth power of a linear form. In the first type, the degree-seven determinant equation forces the cubic part to be affine-linear in the cone direction. In the binary type, the degree-six equation gives a constant null direction in a two-variable Hessian of the cubic part. In the unary type, the degree-five equation and a constant-direction lemma give the same conclusion. Every type therefore reduces to \[ f=P(x_1,x_2,x_3)+x_4Q(x_1,x_2,x_3)+a x_4^2, \qquad °Q\leq2. \] We prove, independently of the degree or top part of \(P\), that every constant-Hessian polynomial of this form has a polynomial gradient inverse. The branch \(a\ne0\) descends from the known three-dimensional Hessian conjecture after a Schur complement. When \(a=0\), an isotropic-cone rank analysis eliminates rank two, solves the rank-one exception by an explicit triangular inverse, and reduces rank zero to the two-dimensional Hessian conjecture. The coupled degree-six identity is retained throughout; no component with respect to a fixed quadratic form is separated.
cs.AR / 5 / 2608.14285
Qu-Trefoil: Large-Scale Quantum Circuit Simulator Working on FPGA With SATA Storages
Abstract
Quantum circuits are fundamental components of quantum computing, and state-vector-based quantum circuit simulation is a widely used technique for tracking qubit behavior throughout circuit evolution. However, simulating a circuit with $n$ qubits requires $2^{n+4}$ bytes of memory, making simulations of more than 40 qubits feasible only on supercomputers. To address this limitation, we propose the Qu-Trefoil, a system designed for large-scale quantum circuit simulations on an FPGA-based platform called Trefoil. Trefoil is a multi-FPGA system connected to eight storage subsystems, each equipped with 32 SATA disks. Qu-Trefoil integrates a suite of HLS-based universal quantum gates, including Clifford gates (Hadamard (H), Pauli-Z (Z), Phase (S), Controlled-NOT (CNOT)), the T gate, and unitary matrix computation, along with HDL-designed modules for system-wide integration. Our extensive evaluation demonstrates the system's robustness and flexibility, covering quantum gate performance, chunk size, disk extensibility, and efficiency across different SATA generations. We successfully simulated quantum circuits with over 43 qubits, which required more than 128 TB of memory, in approximately 3.72 to 13.06 hours on a single storage subsystem equipped with one FPGA. This achievement represents a significant milestone in the advancement of quantum computing simulations. Furthermore, thanks to its unique architecture, Qu-Trefoil is more accessible, flexible, and cost-efficient than other existing simulators for large-scale quantum circuit simulations, making it a viable option for researchers with limited access to supercomputers.
密码学与安全 (cs.CR)
20
cs.CR / 1 / 2608.13659
"I Thought You Were The Uncensored Place": Norms, Rules, and Moderation in AI-Generated Sexual Content Communities
Abstract
As AI-generated sexual content (AIG-SC) is increasingly produced, online communities have emerged to support creators' needs. To understand whether and how community governance attempts work to prevent abuse while supporting free expression, we interviewed 24 members and moderators of large AIG-SC online communities (10,000+ members) with stated rules against creating and sharing abusive content (e.g., AI-generated CSAM). Through in-depth interviews, we offer insight into: (1) how and why these communities form; (2) implicit community norms; (3) explicitly stated rules---and their operationalization via content moderation; and (4) tensions between community values and moderation that leave space for abusive behavior. Our findings reveal a complex picture: while many creators and communities have personal boundaries against abuse, advice and resources for creating any form of AI-generated sexual content are accessible to users regardless of their intentions. Further complicating community moderation are norms that center anti-censorship and non-judgment, which leave moderators to justify their actions using the limits of the law and terms of service. We end by reflecting on the ways in which technical, community, and legal governance may most effectively mitigate the production of abusive content.
cs.CR / 2 / 2608.13685
Weird Machines in Transport Layer Security
Abstract
Weird machines are latent computational capabilities that emerge from the composition of architectural components. Prior work has studied this phenomenon extensively in software systems, including x86 instructions, ELF metadata, and page tables, and more recently in cyber-physical systems such as industrial control networks. This paper extends weird machine theory to a new domain: the Transport Layer Security (TLS) handshake and its two dominant implementations, OpenSSL and BoringSSL. We show that legitimate TLS primitives, including session cache entries, renegotiation logic, extension parsing, and certificate verification steps, compose into Turing-complete systems whose computation is coupled to authentication and trust decisions rather than physical actuation. We formalize this coupling, which we call trust actuation, and argue that any TLS implementation providing session storage, arithmetic on sequence counters, conditional branching on handshake state, and iteration through resumption or retry loops satisfies the conditions for arbitrary computation. We validate this theory with two working demonstrations built on real OpenSSL code paths. The first, a sentinel system, composes standard TLS primitives into a defensive mechanism that detects anomalous handshake behavior. The second, an authentication bypass, composes the same class of primitives into an attack that defeats a cipher-strength policy check through mid-connection renegotiation, without any memory corruption or external malware. Both demonstrations run against real server and client binaries in Docker.
cs.CR / 3 / 2608.13784
A Reproducibility Protocol for Cross-Implementation Evaluation of Post-Quantum ACVP Test Vectors
Abstract
Independent implementations of a cryptographic standard should reproduce the same known-answer results, yet agreement is meaningful only when the corpus, revisions, public interfaces, exclusions, and evidence are precisely stated. This study defines a product-neutral reproducibility protocol for three public implementations of NIST ML-KEM against a pinned public Automated Cryptographic Validation Protocol corpus. Protocol v2 freezes provider-specific capabilities, applies one validation-error taxonomy symmetrically, preserves every selected case, and separates bytes, validation verdicts, unsupported operations, and adapter errors. The source-built experiment evaluated @noble/post-quantum 0.7.0, liboqs 0.16.0, and Go 1.26.4. Across three repetitions, the required Cartesian product comprised 2,160 base records: all 1,650 declared executable evaluations matched the NIST oracle, and all 510 unsupported records matched Go's predeclared capability boundary. Pairwise agreement was complete on every executable overlap: 720 of 720 noble-liboqs records and 210 of 210 records for each Go pairing. A separate keyGen-ek-projection diagnostic matched all 150 Go encapsulation-key projections without counting them as full key generation. Three frozen controls independently exercised byte comparison, verdict comparison, and malformed-response error separation; each produced its exact predeclared outcome. No base failure, adapter error, or status instability occurred. The evidence establishes bounded author-run repeatability and exposes a practical standards gap: public ML-KEM packages provide materially different deterministic and validation-test surfaces. Independent external reproduction remains unobserved. The results do not establish certification, exhaustive correctness, side-channel resistance, secure integration, or production assurance.
cs.CR / 4 / 2608.13792
The ack3 H1 2026 DeFi Incident Dataset: Audit Scope Across 135 Security Incidents
Abstract
Smart-contract audits cover defined artifacts at a specific time, but the label audited is often treated as project-wide assurance. We analyze audit history and incident-path scope across 135 DeFi security incidents using the H1 2026 DeFi Incident Dataset published by cybersecurity company ack3 (https://ack3.ai), covering 1 January to 29 June 2026. The corpus reports USD 939.86 million in attributed loss. Audit history was identified for 68 incidents. Of these, 46 attack paths were outside all identified public pre-incident audit scopes, 20 were inside at least one scope, and 2 were unresolved. Within this 68-incident subset, outside-scope paths represented 67.6% by count and 94.4% of reported loss. The loss-weighted result was concentrated in two large incidents; excluding both reduced the share to 72.1%, while preserving the direction of the result. We also describe audit age, temporal loss distribution, and affected project types. The results show that project-level audit history and incident-path scope are distinct variables.
cs.CR / 5 / 2608.13803
Mobile Apps vs. Web Browsers: A User Perception Study with Android Apps and Google Chrome
Abstract
This study examines user perceptions of mobile applications (apps) versus web browsers for accessing online services, with an emphasis on security, privacy, and usability aspects. Through a combination of an experiment and a survey with Android smartphone users, the research seeks to identify the key concerns and preferences that influence their choice between mobile apps and web browsers. The findings will offer valuable insights for developers to improve the security, privacy, and usability of both platforms by addressing user concerns and misconceptions.
cs.CR / 6 / 2608.13806
Vaulted Passkeys: A Device-Bound Proposal for Authenticated Credential Export and Import
Abstract
Hardware authenticators deliberately resist private-key extraction, yet replacement, disaster recovery, and controlled migration create a legitimate need for portability. Existing guidance for device-bound credentials commonly reduces recovery risk by registering an additional authenticator before failure. That creates an independent credential registration and requires replacement hardware to exist in advance; it is redundancy, not a backup of the original credential. This paper addresses the resulting recovery gap by exporting protected credential state while the source is available and restoring it to hardware acquired later, without cloning a complete authenticator or exposing plaintext private keys to routine desktop software. We propose Vaulted Passkeys, a device-bound architecture in which a random 256-bit Kvault protects authenticated PKV1 credential envelopes through HKDF-separated keys and four explicit AEAD profiles. The design separates enrollment from export/import and the required vault from optional identity. We contribute a role-separated system model, wire format, threat analysis, implementation mapping, and falsifiable evaluation plan. The prototype demonstrates feasibility but is neither a formal security proof nor a proposed final standard.
cs.CR / 7 / 2608.13815
TLF: Rapid Characterization of RF Transceiver Parameters in Embedded Systems via Bus-Level Interception
Abstract
We present TLF (Transceiver Lifter Framework), a tool for recovering RF transceiver configuration and runtime behavior from bus-level traces captured between a microcontroller and its transceiver IC. A stateful protocol decoder, built against the transceiver's register and data interface, reconstructs operating RF parameters and behavior from intercepted register writes and FIFO transfers. For bus-attached transceivers whose hardware-cryptography keys are loaded through the intercepted host interface, key material is also recoverable. Where the firmware drives frequency hopping -- either through a hardware-assisted engine or a custom schedule -- the decoder extracts the channel table, hop sequence, and timing. We evaluate the approach on two targets from different Semtech families: an SX1233-based UAV C2 modem employing firmware-level FHSS with per-packet sync word rotation, and an SX1276-based Meshtastic node exercising the LoRa register overlay. From a single bus capture, processed in seconds, TLF recovers the complete register-exposed RF configuration (modulation, band plan, phase behavior) without prior knowledge of the target firmware -- sufficient to configure a matched receiver or develop targeted countermeasures. Above the chip layer, a pluggable protocol decoder interprets recovered FIFO payloads as application PDUs, demonstrated end-to-end on Meshtastic. Firmware-level cryptographic state remains, as expected, opaque. The approach requires physical access or emulation of the target hardware, and its recovery depth is bounded by the transceiver's register interface: parameters implemented entirely in firmware (custom FEC, whitening, encryption) are observable only as opaque FIFO payloads.
cs.CR / 8 / 2608.13846
Verified Pythagorean Composition for Adaptive Cryptographic Games: Noise Flooding in Homomorphic Encryption
Abstract
Noise flooding is a standard defense against decryption attacks on approximate homomorphic encryption, but its security proof is unusually sensitive to composition. Replacing each of $q$ adaptive decryption answers with a statistically close simulation and applying an ordinary hybrid argument loses linearly in $q$. The cryptographic proof instead accumulates conditional Kullback-Leibler (KL) costs and converts to statistical distance once, giving the parameter-critical square-root loss. We machine-check this argument using Rocq and SSProve. Given any fully homomorphic encryption scheme that is approximately correct and IND-CPA secure, we formalize a reduction for every $q$-query IND-CPAD adversary and prove \[ \Pr[\mathsf{IND\text{-}CPAD}_{\mathsf{NF}}^{\mathcal A}=1] \leq β_{\mathsf{CPA}}(\mathcal B_{\mathcal A,q}) + \frac{\sqrt{qn}}{2γ}. \] where $n$ is the plaintext dimension and $γ$ is the flooding-width multiplier. Our proof constructs a new relational program logic over SSProve semantics. Its Pythagorean judgment composes conditional KL budgets without converting them to statistical distance, and a verified trace compiler lifts a local oracle rule to arbitrary adaptive programs with a single final conversion.
cs.CR / 9 / 2608.13905
CipherSight: Robust Website Fingerprinting via Record-Resource Semantic Supervision under Distribution Shifts
Abstract
HTTPS website fingerprinting (WF) aims to identify visited websites from metadata observable in encrypted traffic. However, real-world deployments introduce a significant out-of-distribution (OOD) problem caused by temporal and geographic changes, while previously unseen websites are common in open-world scenarios. Existing methods primarily learn from raw TCP packet sequences and struggle to capture stable and generalizable website representations, resulting in performance degradation under practical conditions. We propose CipherSight, a TLS-record-based hierarchical framework for robust HTTPS WF. Unlike existing approaches that rely on TCP packet sequences and are sensitive to transport-layer artifacts, CipherSight learns website representations from TLS records by jointly encoding multiple record-level attributes. It introduces a hierarchical architecture that captures both intra-flow dependencies among TLS records and inter-flow interactions across concurrent flows, enabling the model to exploit structural patterns in HTTPS traffic. Besides, to learn robust representations, CipherSight employs a masked record modeling (MRM) task to capture contextual traffic semantics and leverages fine-grained record-resource annotations as privileged supervision through structure-aware objectives and semantic distillation. Experiments show that CipherSight achieves 95.41% accuracy across more than 2,000 website classes in the closed-world setting and maintains over 90% accuracy under both temporal and geographic drift, consistently outperforming all evaluated baselines.
cs.CR / 10 / 2608.13920
Characterizing the Variance Envelope: A Multi-Dimensional Analysis of Spectre Telemetry Across Architectures and Workloads
Abstract
Hardware attacks like Spectre exploit built-in processor vulnerabilities, leaving anomalous footprints in Hardware Performance Counter (HPC) metrics. While machine learning can detect these footprints in controlled settings, static models fail in the real world when confronted with background system noise, diverse attack variants, and adversarial traffic pacing. To close this gap, this paper characterizes the "variance envelope"-the full range of how attack signatures shift- across Intel, ARM, and AMD architectures. We evaluate an extensive experimental matrix encompassing three attack variants, four pacing modes, and four background-noise conditions. Our analysis proves that HPC signatures are highly fragile and easily warped by their execution environment. Furthermore, we expose a critical microarchitectural bottleneck: the persistent, hardware-level failure of Prime+Probe attacks on the AMD Jaguar. Ultimately, this comprehensive characterization demonstrates that reliable runtime detection requires architecture-aware, adaptive monitoring rather than static models.
cs.CR / 11 / 2608.13930
Extracting and Verifying Illicit Bitcoin Addresses from Underground Forum Discussions
Abstract
Existing labeled Bitcoin datasets are largely derived from community-reported abuse, blockchain heuristics, incident-specific collections, or proprietary labeling processes. Their construction methods are rarely publicly reproducible and often provide limited evidence that an address was directly involved in illicit activity. We present a reproducible pipeline for constructing evidence-backed Bitcoin labels from HackForums, an underground cybercrime forum with fifteen years of archived activity. The pipeline combines LLM-assisted screening, expert review, and on-chain validation to identify Bitcoin addresses explicitly associated with illicit transactions discussed on the forum. Each released label is supported by contextual evidence from underground discussions and validated on-chain. The resulting dataset contains 2,438 manually verified illicit Bitcoin addresses spanning 2010-2024 and twelve cybercrime categories assigned during LLM screening. We release the dataset, temporal metadata, and the complete extraction pipeline to support reproducible research on cryptocurrency-facilitated cybercrime.
cs.CR / 12 / 2608.13948
Vectorized SQIsign Implementation Using AVX-512
Abstract
SQIsign is the sole isogeny-based digital signature scheme submitted to the NIST Post-Quantum Cryptography standardization process, distinguished by its foundation on the hardness of the endomorphism ring problem for supersingular elliptic curves. While offering compact key and signature sizes, SQIsign's practical deployment is hindered by computationally intensive signing procedures. This paper presents the first comprehensive vectorized implementation of SQIsign utilizing the AVX-512 Integer Fused Multiply-Add (IFMA) instruction set architecture. By systematically redesigning the computational stack---encompassing prime-field and extension-field arithmetic, elliptic curve operations including batched point doubling and scalar multiplication, as well as pairing computations via cubical arithmetic and two-dimensional isogeny evaluations---we achieve substantial performance improvements over the reference implementation. When combined with Qlapoti technology, our implementation attains a $2.69\times$ speedup for signing and a $3.18\times$ improvement for verification at NIST security level I. Contrary to misconceptions regarding the obsolescence of AVX-512, we emphasize that Intel's AVX10 instruction set architecture (revision 10.2, scheduled for widespread deployment in late 2026) will standardize AVX-512 capabilities---including IFMA instructions---across both performance and efficiency cores, ensuring long-term viability of these optimization techniques. Furthermore, our vectorization strategies are architecture-agnostic and provide a methodological foundation applicable to broader isogeny-based cryptographic constructions. This work demonstrates that SIMD vectorization represents a critical yet underexplored optimization dimension for post-quantum isogeny-based schemes, independent of recent algorithmic advances.
cs.CR / 13 / 2608.13981
Structural Leakage in Graph Encryption: Attacks and Defenses
Abstract
Graph encryption schemes (GES) enable secure outsourcing of graph data while supporting efficient queries. This report provides a comprehensive analysis of structural leakage in GES for single-pair shortest path (SPSP) queries, integrating findings from two recent works. First, we analyze PathGES, a scheme designed to resist query recovery attacks through heavy-light decomposition (HLD) and canonical fragment encoding. Our analysis reveals that PathGES suffers from significant imbalances in HLD decomposition, with over 99% of token-path mappings being one-to-one on real-world datasets, enabling both the Falzon-Paterson attack and side-channel inference of path lengths. Second, we present Fragment Tree attack that exploits these structural weaknesses to recover query contents, achieving up to 10.24% exact recovery on sparse graphs. Third, we introduce BlindGES, an enhanced scheme incorporating a Merge-and-Divide mechanism and two-level multimap index that reduces one-to-one mappings to below 20%, cuts setup time by 50%, reduces storage overhead by 32%, and limits path length leakage to under 1%. This report systematically presents attack methodologies, defense mechanisms, security proofs, and experimental evaluations on seven real-world datasets.
cs.CR / 14 / 2608.14094
P2Skill: Privacy Preserving Skill Distillation for Cloud-Local LLM Inference Systems
Abstract
Cloud-local LLM inference systems have the potential to use the reasoning capability of large cloud models while protecting sensitive user data on personal devices. Cloud-bound requests must exclude personally identifiable information (PII) to prevent external data leakage. Existing privacy-preserving methods rely on prompt perturbation, entity masking, or model fine-tuning, but these approaches may distort contextual semantics or require additional training. This paper proposes P2Skill, a prompt-based skill distillation method in which a local small language model (SLM) autonomously performs decomposition, PII-aware routing, paraphrasing, and reconstruction by following the skill prompts. Skills are iteratively refined from execution failures by a cloud LLM, enabling the local SLM to generalize beyond memorized PII patterns, and therefore P2Skill requires no privacy-specific fine-tuning or learned auxiliary detectors. Evaluation on a four-domain benchmark shows that P2Skill achieves $1.69\times$ and $3.66\times$ higher privacy-preserved inference quality than previous baselines.
cs.CR / 15 / 2608.14126
BGA: A noise-immune neural distillation framework for malicious signature extraction in high-entropy encrypted flows
Abstract
To mitigate attention dilution in high-entropy TLS 1.3 flows, we propose BGA, a noise-immune neural distillation framework for encrypted threat intelligence.The methodology first employs Analysis of Variance (ANOVA) to decouple high-discriminatory control-plane features - specifically industrial setpoints - from stochastic cryptographic noise. To resolve the extreme class imbalance within a corpus of 86,878 flow records, a Wasserstein GAN with Gradient Penalty (WGAN-GP) module, enforcing the 1-Lipschitz constraint, is integrated to synthesize high-fidelity minority samples, elevating the detection recall of rare Malicious State Command Injections(MSCI) attacks by 43.2%. At its core, the BGA architecture integrates Bidirectional Long Short-Term Memory (BiLSTM) for temporal dependency extraction and an Adaptive Gated Multi-Head Attention mechanism. This gated unit functions as a neural filter to dynamically suppress encryption artifacts while amplifying malicious signatures. Extensive evaluations on CIC-IDS-2018 and Edge-IIoT benchmarks demonstrate a performance ceiling exceeding 95.2% across all key metrics. Furthermore, noise-injection stress tests confirm BGAs superior structural resilience with a 8.57% performance margin over vanilla Transformers, while its ultra-low inference latency of 0.2820 ms (estimated 1.6920 ms via theoretical scaling for ARM) indicates a high potential for real-time feasibility on heterogeneous industrial edge gateways, providing a promising architectural baseline for future hardware implementation.
cs.CR / 16 / 2608.14329
A Four-Axis Trustworthiness Benchmark for LLM-as-Judge in Principle-Based Regulation
Abstract
Principle-based regulation, with evaluative standards such as "fair, clear, and not misleading" or "deliver good outcomes", cannot be reduced to binary predicates, and LLM-as-judge is increasingly used as the substitute. Our position is that any such judge must be evaluated on four axes: accuracy, paraphrase robustness, adversarial robustness, and calibration. We release Principle-Bench, 168 cryptoasset financial-promotion scenarios mapped to two UK FCA principles, with paraphrase, adversarial keyword-stuffing, and boundary perturbations authored under a pre-registered rubric; the first benchmark covering all four axes for principle-based regulation. We also introduce Ceca (Calibrated Exemplar-Cluster Assessment): a calibrated, auditable assessor that emits exact per-exemplar counterfactual attributions. Across keyword counting, three sentence-transformer embedders, an open-weight LLM-judge, and a calibrated cascade, no method dominates all four axes. A 120B LLM-judge, strongest on benign inputs, loses 47 accuracy points (0.74 to 0.27) on keyword-stuffed Consumer Duty inputs: "compliance theatre." A second judge from a different model family agrees only at Cohen's kappa = 0.16 on that split, localising the failure to the model rather than the corpus. Any deployment-grade LLM-judge for principle-based regulation must report per-principle adversarial deception and post-hoc calibration alongside aggregate accuracy.
cs.CR / 17 / 2608.14501
Lower Bounds on Black-Box Constructions of Pseudorandom Functions
Abstract
In their seminal work, Goldreich, Goldwasser, and Micali [CRYPTO 1984] constructed a pseudorandom function (PRF) using a black-box access to a pseudorandom generator (PRG). When combined with Levin's domain extension technique, the GGM construction invokes the PRG $ω(\log n)$ times, where $n$ denotes the input length to the PRG. To this day, no black-box construction achieving fewer calls is known. Recently, Beimel, Malkin, and Mazor [CRYPTO 2024] showed that for a certain family of constructions, which they termed \emph{tree constructions}, the GGM construction is optimal. However, the basic challenge of whether a PRF can be built with just \emph{one invocation} of the PRG still remains open. In this work, we consider fully black-box constructions of PRFs from PRGs, where both the construction and the reduction are required to be black-box, and the number of interactions the reduction makes with the adversary is independent of the number of oracle calls the adversary makes to its underlying function within each interaction. Our main result shows that no such construction can have $o(n/\log n)$ and $o(\mathsf{in}/\log\mathsf{in})$ \emph{non-adaptive} calls to the PRG, where $\mathsf{in}$ is the input length of the PRF. This impossibility holds even for weak PRFs with one-bit output, where the adversary is restricted to making i.i.d. uniformly random queries. In addition, we prove a lower bound for weak PRFs with sufficiently long outputs that holds even when the construction is allowed to make adaptive queries to the PRG.
cs.CR / 18 / 2608.14532
Trust Without Boundaries: An Architectural Analysis of Satellite Flight Software
Abstract
As spacecraft become more software-driven and interconnected, onboard flight software is an increasingly important security boundary. Popular flight software architectures often treat onboard components as trusted peers, simplifying integration while limiting internal isolation and access control. We analyze NASA's Core Flight Software (cFS) to examine how authority, identity, communication, observability, and persistence are distributed across onboard components. Using NASA's flight-representative NOS3 simulator, we validate these weaknesses through five experiments implemented with a malicious onboard component that abuses legitimate architectural privileges. We then compare cFS with other modular flight software frameworks to identify recurring trust assumptions and architectural weaknesses. Our results show that a single compromised component can exploit broadly shared authority in ways that are difficult to distinguish from legitimate behavior. We conclude with architectural implications and discuss mechanisms for strengthening internal trust boundaries in future flight software systems.
cs.CR / 19 / 2608.14356
Designing Inclusive Crypto-Asset Dispute Resolution A Hybrid AI and Smart Contract Online Dispute Resolution Framework for Vulnerable Users
Abstract
The growing use of crypto-assets has generated disputes that sit uneasily within existing legal redress mechanisms. Their resolution is complicated by the technical features of blockchain transactions, the cross-border nature of many relationships, and fragmented legal rules. These difficulties are particularly acute for users without legal or technical expertise, who may struggle to identify their rights, organise relevant evidence, or pursue an effective remedy. Crypto-assets may therefore produce new forms of digital exclusion, despite their association with financial accessibility. This paper examines whether a hybrid Online Dispute Resolution framework combining artificial intelligence and blockchain-based smart contracts could mitigate these barriers. In the proposed model, AI operates primarily off-chain, supporting natural-language interaction, dispute classification, evidence organisation, and accessible legal information. Smart contracts perform limited on-chain functions, including dispute registration, timestamping, verification, escrow management, and, where appropriate, execution of agreed outcomes. The framework is human-centred, with emphasis on explainability, procedural guidance, and human oversight in complex cases. The model is examined through disputes involving crypto-asset seizure, exchange failures, and cross-border scams. The paper argues that a carefully delimited hybrid architecture may improve accessibility, transparency, efficiency, and enforceability, provided that automated and human roles are clearly defined. It also addresses key legal and governance concerns, including algorithmic bias, liability, data protection, smart-contract enforceability, and compliance with the EU AI Act and the Markets in Crypto-Assets Regulation. The paper thus proposes a legally grounded framework for more inclusive and robust dispute resolution in crypto-asset environments.
cs.CR / 20 / 2608.14331
Equivalence Between Average-Case Hardness of Learning and Cryptography for Mixed Quantum States
Abstract
The relationship between cryptography and learning theory has long been a central theme in the foundations of theoretical computer science: cryptographic primitives can imply hardness of learning, while hardness of learning can in turn be used to construct cryptographic schemes. Recent works have begun exploring analogous connections in the quantum setting, relating the average-case hardness of learning quantum states (AHL) to cryptographic primitives such as one-way state generators (OWSG). Despite recent progress exploring this for pure states, the relationship for mixed states has remained an open question. In this work, we prove that the existence of AHL for mixed quantum states is equivalent to the existence of inefficiently verifiable one-way state generators (IV-OWSGs). As a consequence, this relates mixed-state AHL to EFI pairs. Moreover, as a corollary of existing results, we obtain a separation between IV-OWSGs and OWSGs relative to the SWAP oracle.