arXiv Daily Digest

323

Papers

HERMES: Heterogeneous Edge-Relational Multi-Head Embedded SSM Attention for Traffic Conflict Prediction at Signalized Intersections

HERMES：用于信号交叉口交通冲突预测的异构边关系多头嵌入式 SSM 注意力机制

Islam, Md Monzurul, Das, Subasish

Abstract

Surrogate safety measures (SSMs) enable proactive traffic safety assessment, but many existing methods evaluate pairwise interactions independently or flatten multi-agent scenes into fixed feature vectors, limiting their ability to represent heterogeneous interaction structure and evolving scene-level risk. This study formulates traffic conflict assessment as temporal heterogeneous scene-graph classification and proposes HERMES, a heterogeneous edge-relational graph neural network with SSM-informed multi-head attention. Vehicles and pedestrians are represented as heterogeneous nodes, while vehicle-vehicle, vehicle-pedestrian, and pedestrian-pedestrian interactions are encoded as relation-specific edges with continuous kinematic and surrogate-safety descriptors. Relation-specific attention, dynamic node-edge updates, safety-aware graph pooling, and temporal sequence learning are jointly used to estimate scene-level conflict probability. HERMES was evaluated using 109,028 trajectory-derived sequences from a signalized urban intersection and tested on an independently collected comparable intersection dataset. Enhanced HERMES achieved an AUC-ROC of 0.9898 +/- 0.0013, an AUC-PR of 0.9412 +/- 0.0067, and an F1 score of 0.8449 +/- 0.0103. At a 5% false-alarm rate, it detected 95.7% of conflict sequences, outperforming the strongest Transformer baseline and XGBoost. In zero-shot external evaluation, HERMES achieved an AUC-ROC of 0.9752 and an AUC-PR of 0.7829. Joint source-target training further improved target-site performance with limited target-site data. These findings show that preserving heterogeneous interaction topology, safety-informed edge semantics, and short-term temporal evolution improves scene-level conflict classification and supports transferable roadside safety monitoring at signalized intersections.

Chinese Translation

替代安全措施（SSMs）使得主动交通安全评估成为可能，但许多现有方法独立评估成对交互，或将多智能体场景压缩为固定特征向量，从而限制了它们表示异构交互结构和不断演变的场景级风险的能力。本研究将交通冲突评估形式化为时间异构场景图分类，并提出了 HERMES，一种具有 SSM 信息的多头注意力机制的异构边关系图神经网络。车辆和行人被表示为异构节点，而车辆-车辆、车辆-行人和行人-行人交互则被编码为具有连续运动学和替代安全描述符的关系特定边。关系特定注意力、动态节点-边更新、安全感知图池化和时间序列学习共同用于估计场景级冲突概率。HERMES 使用来自一个信号化城市交叉口的 109,028 条轨迹派生序列进行了评估，并在一个独立收集的可比较交叉口数据集上进行了测试。增强版 HERMES 实现了 0.9898 +/- 0.0013 的 AUC-ROC，0.9412 +/- 0.0067 的 AUC-PR 和 0.8449 +/- 0.0103 的 F1 分数。在 5% 的误报率下，它检测到了 95.7% 的冲突序列，超越了最强的 Transformer 基线和 XGBoost。在零样本外部评估中，HERMES 实现了 0.9752 的 AUC-ROC 和 0.7829 的 AUC-PR。联合源-目标训练进一步改善了目标站点在有限目标站点数据下的性能。这些发现表明，保留异构交互拓扑、安全感知边语义和短期时间演变能够改善场景级冲突分类，并支持在信号交叉口的可转移路边安全监测。

View on arXiv Download PDF AI Translation

cs.RO / 2 / 2607.20653

PhysCoRe: Physics-Corrected Residual World Models for Material-Aware Deformable Dynamics

PhysCoRe：物理校正的残差世界模型用于材料感知的可变形动力学

Yin, Haocheng, Tao, Shuohan, Chen, Yongsheng, Gan, Lu

Abstract

Predicting how deformable objects evolve under robotic manipulation is a longstanding challenge. Existing approaches typically rely on per-object optimization to fit material parameters, which can be slow and cannot generalize, while end-to-end learned alternatives extrapolate poorly and often violate basic physical structure. We present PhysCoRe, a physics-corrected residual world model that couples a differentiable Material Point Method (MPM) simulator with two feed-forward neural networks. A material refinement module, Material from Motion (MfM), infers per-particle elasticity from visual observations, grounding the simulator in object-specific physics. A residual correction module, Residual from Dynamics (RfD), learns the discrepancy and predicts corrections to the simulator's internal dynamics, absorbing systematic biases that the analytical model cannot capture. This design also supports online material identification on novel objects. MfM adapts from limited interactions, and its predictive uncertainty steers further exploration toward the regions where its estimate is least confident. Experiments on real deformable-object manipulation sequences show that PhysCoRe outperforms state-of-the-art baselines in prediction accuracy, and that its predicted confidence forms a reliable distribution across the object's geometry, providing a natural signal for future confidence-guided exploration.

Chinese Translation

预测可变形物体在机器人操控下的演变一直是一个长期存在的挑战。现有的方法通常依赖于针对每个物体的优化来拟合材料参数，这种方法可能较慢且无法泛化，而端到端学习的替代方案则往往外推效果不佳，并且常常违反基本的物理结构。我们提出了PhysCoRe，一种物理校正的残差世界模型，它将可微分的材料点方法（Material Point Method, MPM）模拟器与两个前馈神经网络相结合。材料细化模块“运动中的材料”（Material from Motion, MfM）从视觉观测中推断每个粒子的弹性，将模拟器与特定物体的物理特性相结合。残差校正模块“动力学中的残差”（Residual from Dynamics, RfD）学习差异并预测对模拟器内部动力学的修正，吸收分析模型无法捕捉的系统性偏差。这一设计还支持对新物体的在线材料识别。MfM能够从有限的交互中适应，其预测的不确定性引导进一步探索到其估计信心最低的区域。对真实可变形物体操控序列的实验表明，PhysCoRe在预测准确性上优于最先进的基线，并且其预测的置信度在物体几何形状上形成了可靠的分布，为未来的置信度引导探索提供了自然信号。

View on arXiv Download PDF AI Translation

cs.RO / 3 / 2607.20662

Scalable Low-Cost Laboratory Automation: A Digital Twin-Integrated Robotic Platform for Autonomous Liquid Handling (RAINBOTTM)

可扩展的低成本实验室自动化：一种集成数字双胞胎的自主液体处理机器人平台（RAINBOTTM）

Ayeche, Mohamed Rami, Sid, Souhil, Mostofa, Ahyen, Hussain, Rehaan, Shayesteh, Ali, Mellouhi, Fadwa El

Abstract

Laboratory automation accelerates discovery, yet its adoption is constrained by the high cost, proprietary design, and limited remote supervisability of commercial liquid-handling systems. This work presents RAINBOT\textsuperscript{TM}, a low-cost, openly reproducible liquid-handling robot built by converting a consumer-grade Cartesian 3D printer (Elegoo Neptune 4 Max). The printer extruder is replaced by a precision single-channel pipette actuated through the printer's own G-code-driven X--Y--Z gantry, with plunger and tip-eject motions effected by two compact linear actuators under Python control. To make experiments transparent and remotely supervisable, a browser-based digital twin is implemented to synchronise bidirectionally with the physical platform, mirroring kinematics and pipetting states in real time and exposing remote monitoring, intervention, and an emergency stop from any web browser. As a proof of concept, RAINBOT\textsuperscript{TM} performed sequential exchanges of differently coloured aqueous solutions while an integrated colour sensor quantified the resulting mixtures; measured red, yellow, and blue (RYB) responses agreed with expected mixing behaviour to within a mean absolute error of two percentage points, validating correct execution and real-time tracking. Closing the loop, the platform is coupled to the CEID\textsuperscript{TM} (Cooperative Explorer for Inverse Design) framework, which recasts experimentation from iterative manual guessing into a goal-directed inverse-design search while keeping a human in the loop. The complete hardware costs under US\$1300, which is roughly an order of magnitude below entry-level commercial handlers, thereby establishing an accessible physical--virtual framework for self-driving laboratory automation.

Chinese Translation

实验室自动化加速了科学发现，但其采用受到商业液体处理系统高成本、专有设计和有限远程监督能力的限制。本研究提出了RAINBOTTM，这是一款低成本、可公开复现的液体处理机器人，通过将消费级卡式3D打印机（Elegoo Neptune 4 Max）改造而成。打印机的挤出头被一个精确的单通道移液器所替代，该移液器通过打印机自身的G-code驱动的X-Y-Z龙门架进行驱动，活塞和尖端弹出动作由两个紧凑的线性执行器在Python控制下实现。为了使实验透明且可远程监督，实施了一个基于浏览器的数字双胞胎，与物理平台双向同步，实时反映运动学和移液状态，并允许通过任何网页浏览器进行远程监控、干预和紧急停止。作为概念验证，RAINBOTTM在集成的颜色传感器的帮助下，顺序交换了不同颜色的水溶液，并量化了所得到的混合物；测得的红色、黄色和蓝色（RYB）响应与预期的混合行为一致，平均绝对误差为两个百分点，验证了正确执行和实时跟踪。通过闭环，该平台与CEIDTM（逆向设计合作探索者）框架相结合，将实验从迭代的手动猜测转变为目标导向的逆向设计搜索，同时保持人类在环中。完整的硬件成本低于1300美元，约为入门级商业处理器的一个数量级，从而建立了一个可接入的物理-虚拟框架，用于自主实验室自动化。

View on arXiv Download PDF AI Translation

cs.RO / 4 / 2607.20665

Safe and Scalable Multi-Drone Payload Transport via CBF-based Reinforcement Learning with Zero-Shot Sim-to-Real Transfer

基于CBF的强化学习与零-shot模拟到现实转移的安全可扩展多无人机载荷运输

Choi, Jaeyoun, So, Oswin, Zhang, Songyuan, Taylor, Cooper, Fan, Chuchu

Abstract

Multi-drone payload transportation has emerged as a promising research paradigm with potential applications in construction, logistics, and disaster response. However, the complex coupled dynamics among drones, cables, and payloads pose significant challenges, and existing approaches remain limited in safety and scalability, particularly in dynamic and unstructured environments. In this work, we propose a learning-based framework for safe and scalable multi-drone cooperative payload transport. We introduce a minimal 2D abstraction that preserves the task-relevant drone-payload coupling required for coordination and safety, while remaining computationally efficient for large-scale learning. Using domain randomization over team size and physical parameters, we train a fully distributed policy via Discrete Graph Control Barrier Function Proximal Policy Optimization (DGPPO), enabling robust zero-shot sim-to-real transfer without fine-tuning. Extensive real-world evaluations demonstrate that a single learned policy generalizes across varying team sizes and task scenarios. Furthermore, multi-group hardware experiments show that the same policy can safely operate in dynamic environments, where other drone teams act as moving obstacles. These results indicate that the proposed framework enables efficient, safe, and scalable multi-drone payload transportation with strong generalization to complex real-world conditions.

Chinese Translation

多无人机载荷运输已成为一种有前景的研究范式，具有在建筑、物流和灾害响应等领域的潜在应用。然而，无人机、缆绳和载荷之间复杂的耦合动态带来了显著挑战，现有方法在安全性和可扩展性方面仍然有限，尤其是在动态和非结构化环境中。在本研究中，我们提出了一种基于学习的框架，用于安全和可扩展的多无人机协作载荷运输。我们引入了一个最小的二维抽象，保留了协调和安全所需的任务相关无人机-载荷耦合，同时在大规模学习中保持计算效率。通过对团队规模和物理参数进行领域随机化，我们通过离散图控制障碍函数近端策略优化（DGPPO）训练了一个完全分布式策略，实现了强大的零-shot模拟到现实转移，无需微调。大量的现实世界评估表明，单一学习策略能够在不同团队规模和任务场景中进行泛化。此外，多组硬件实验显示，同一策略能够在动态环境中安全运行，其中其他无人机团队作为移动障碍物。这些结果表明，所提出的框架能够实现高效、安全和可扩展的多无人机载荷运输，并在复杂的现实世界条件下具有强泛化能力。

View on arXiv Download PDF AI Translation

cs.RO / 5 / 2607.20679

Towards Capability-Aware Traversability Navigation for Unstructured Environments

面向能力感知的非结构化环境可通行性导航

Capezzuto, Gianluca, Tommaselli, Felipe, Angarola, Matheus P., Godoy, Ricardo V., Becker, Marcelo

Abstract

Estimating traversability in unstructured environments requires conditioning on robot embodiment, as the same terrain can be traversable for one platform and unsafe for another. Existing methods often transfer predictions across morphologies through late-stage trajectory filtering rather than encoding platform constraints in the learned representation. We propose Capability-Aware Traversability (CAT), a framework that embeds physical limits directly into the spatial feature space. CAT grounds dense supervision masks in physical trajectories through an interactive annotation pipeline and modulates semantic terrain maps with robot-specific traversability vectors through Spatially-Adaptive Denormalization (SPADE) blocks. Across human-annotated and trajectory-aligned datasets, CAT leads all ranking-based metrics, improving AUROC by 11.0% on physically executed trajectories and AUPRC by 15.8% on human traces over the strongest baseline. Ablations show that spatial conditioning and per-robot prototypes produce capability sensitivity beyond generic path prediction. Deployments on a legged quadruped and a wheeled skid-steer demonstrate embodiment-aware obstacle avoidance on embedded hardware at 4.8 Hz.

Chinese Translation

在非结构化环境中估计可通行性需要考虑机器人形态，因为同一地形对于一种平台可能是可通行的，而对另一种平台则可能是不安全的。现有方法通常通过后期轨迹过滤在不同形态之间转移预测，而不是在学习的表示中编码平台约束。我们提出了能力感知可通行性（Capability-Aware Traversability, CAT），这是一个将物理限制直接嵌入空间特征空间的框架。CAT通过交互式注释管道将密集监督掩码与物理轨迹结合，并通过空间自适应去归一化（Spatially-Adaptive Denormalization, SPADE）模块调节具有机器人特定可通行性向量的语义地形图。在人类注释和轨迹对齐的数据集上，CAT在所有基于排名的指标中表现最佳，在物理执行的轨迹上提高了11.0%的AUROC，在人类轨迹上提高了15.8%的AUPRC，相较于最强基线。消融实验表明，空间条件和每个机器人的原型产生了超越通用路径预测的能力敏感性。在一台腿式四足机器人和一台轮式滑移转向机器人上的部署展示了在嵌入式硬件上以4.8 Hz的频率实现的形态感知障碍物规避。

View on arXiv Download PDF AI Translation

cs.RO / 6 / 2607.20683

FELT: Generating Tactile Signals from Vision for Visuo-Tactile Manipulation

FELT：从视觉生成触觉信号以实现视觉-触觉操控

Li, Zinan, Ling, Yiyang, Gu, Yuming, Huang, Binghao, Liang, Chenhao, Islam, Sharfin, Bedri, Hisham, Chirikjian, John, Li, Yunzhu, Nikolaidis, Stefanos, Seita, Daniel

Abstract

The sense of touch is central to manipulation, especially when vision is occluded or ambiguous. Although combining vision and touch improves manipulation, learning robust visuo-tactile policies requires substantial tactile data. Such data remains scarcer than visual data, because tactile sensors are fragile, specialized, and hard to standardize. To address this, we present Feature-Extracted Latent Tactile (FELT), a learning-based framework that synthesizes per-finger pressure tactile images from RGB observations, reducing the need for tactile-equipped data collection. FELT uses a large frozen visual encoder and a lightweight query decoder to predict tactile signals in a single feed-forward pass. To respect the physical topology of dual-finger tactile sensors, FELT decodes the left and right tactile sensor panels through separate branches, capturing the asymmetric contact patterns during interactions such as wiping, insertion, and in-hand rotation. At inference time, FELT only requires RGB data, allowing us to augment existing vision-only data with tactile observations, either as generated tactile images or as latent tactile features. Experiments on four contact-rich manipulation tasks demonstrate that both generated tactile images and latent tactile features improve policy success over vision-only baselines, with latent feature requiring no real tactile sensor during policy training or deployment. Supplementary material is available on our anonymous website: https://felt-tactile.github.io/.

Chinese Translation

触觉在操控中至关重要，尤其是在视觉受到遮挡或模糊时。尽管结合视觉和触觉可以改善操控，但学习稳健的视觉-触觉策略需要大量的触觉数据。这类数据相较于视觉数据仍然稀缺，因为触觉传感器脆弱、专业且难以标准化。为了解决这个问题，我们提出了特征提取潜在触觉（Feature-Extracted Latent Tactile, FELT），这是一个基于学习的框架，可以从RGB观测中合成每个手指的压力触觉图像，从而减少对触觉设备数据收集的需求。FELT使用一个大型的冻结视觉编码器和一个轻量级的查询解码器，在单次前馈传递中预测触觉信号。为了尊重双指触觉传感器的物理拓扑，FELT通过独立的分支解码左右触觉传感器面板，捕捉在擦拭、插入和手中旋转等交互过程中的不对称接触模式。在推理时，FELT只需要RGB数据，使我们能够用触觉观测（无论是生成的触觉图像还是潜在触觉特征）增强现有的仅视觉数据。在四个接触丰富的操控任务中的实验表明，生成的触觉图像和潜在触觉特征都提高了策略成功率，相较于仅视觉基线，潜在特征在策略训练或部署期间不需要真实的触觉传感器。补充材料可在我们的匿名网站上获取：https://felt-tactile.github.io/

View on arXiv Download PDF AI Translation

cs.RO / 7 / 2607.20710

Decentralized UAV Swarms for Ground Target Protection in GPS- and Communication-Denied Environments

在GPS和通信受限环境下的去中心化无人机群体用于地面目标保护

Silveria, Dimitria, de Araujo, Paulo Ricardo Marques, Nascimento, Tiago, Givigi, Sidney

Abstract

The presence of UAVs in military operations has recently increased, also increasing the demand for defense systems against UAV attacks. UAVs can also be used as countermeasures. Most available methods rely on UAV-to-UAV communication and global positioning. However, such resources may not be available in modern warfare scenarios. To address these limitations, we propose a pipeline for ground-target protection against UAV attacks that employs autonomous swarms of UAVs. We assume a communication- and GPS-denied environment in which the UAVs use onboard sensors to track the target and coordinate as a swarm. We developed Kalman filters to estimate the states of unknown targets and the positions of UAVs in the swarm using only relative measurements. Also, our strategy is to encircle the target of interest to maximize coverage. To achieve that, we propose a decentralized swarm encirclement technique that adapts to the target's motion. Our approach was extensively validated using real robots, demonstrating its effectiveness in detecting, encircling, and intercepting hostile UAVs.

Chinese Translation

无人机在军事行动中的使用最近有所增加，这也提高了对防御系统以抵御无人机攻击的需求。无人机同样可以作为反制手段。现有的大多数方法依赖于无人机之间的通信和全球定位。然而，在现代战争场景中，这些资源可能不可用。为了解决这些限制，我们提出了一种针对无人机攻击的地面目标保护方案，该方案采用自主无人机群体。我们假设在一个没有通信和GPS的环境中，无人机利用机载传感器跟踪目标并作为一个群体进行协调。我们开发了卡尔曼滤波器，仅使用相对测量来估计未知目标的状态和无人机在群体中的位置。此外，我们的策略是围绕感兴趣的目标进行包围，以最大化覆盖范围。为此，我们提出了一种去中心化的群体包围技术，该技术能够适应目标的运动。我们的方案通过真实机器人进行了广泛验证，证明了其在检测、包围和拦截敌对无人机方面的有效性。

View on arXiv Download PDF AI Translation

cs.RO / 8 / 2607.20743

Self-Supervised Bio-Inspired Robotic Trajectory Planning with Obstacle Avoidance

自监督生物启发式机器人轨迹规划与避障

Krupa, Miroslav, Cibula, Miroslav, Malinovská, Kristína

Abstract

Trajectory planning is a fundamental problem in robotics, requiring the generation of collision-free and efficient trajectories in a potentially complex environment. While sampling-based planners remain the dominant approach, they are often computationally expensive, particularly in high-dimensional spaces and obstacle-rich environments. Methods based on model learning offer a promising alternative, enabling efficient planning through a bounded number of forward passes through a neural trajectory planner, but commonly suffer from low sample efficiency or limited generalisation due to their reliance on exploration or expert demonstrations. This follow-up work tests our neuro-inspired self-supervised learning framework for trajectory planning that leverages forward and inverse models as the internal supervisory mechanism in an environment that contains an obstacle. Experimental results demonstrate the feasibility of the approach while revealing a tendency of our planner to exploit the learning signal provided by the forward and inverse models. To address this issue, additional training regimes and mitigation strategies are proposed and evaluated.

Chinese Translation

轨迹规划是机器人学中的一个基本问题，需要在潜在复杂的环境中生成无碰撞且高效的轨迹。尽管基于采样的规划器仍然是主流方法，但它们通常计算开销较大，尤其是在高维空间和障碍物丰富的环境中。基于模型学习的方法提供了一种有前景的替代方案，通过有限次数的前向传递在神经轨迹规划器中实现高效规划，但通常由于依赖于探索或专家演示，导致样本效率低或泛化能力有限。本研究的后续工作测试了我们的神经启发式自监督学习框架，该框架利用前向和逆向模型作为内部监督机制，在包含障碍物的环境中进行轨迹规划。实验结果展示了该方法的可行性，同时揭示了我们的规划器倾向于利用前向和逆向模型提供的学习信号。为了解决这一问题，提出并评估了额外的训练方案和缓解策略。

View on arXiv Download PDF AI Translation

cs.RO / 9 / 2607.20748

A real-time RGB-D perception pipeline for autonomous impact hammers in mining: self-filtering, rock segmentation and rock-breaking poses generation

用于矿业的实时RGB-D感知管道：自我过滤、岩石分割和岩石破碎姿态生成的自主冲击锤

Gallegos, Martín, Leiva, Francisco, Loncomilla, Patricio, Cortés, Michelle, Ruiz-del-Solar, Javier

Abstract

Impact hammers, also known as rock-breakers, are essential machines in mining operations, where they perform secondary reduction. In underground mining, these machines are typically teleoperated, limiting operational efficiency. This paper presents a real-time RGB-D perception pipeline as a step towards automating the operation of hydraulic impact hammers used in mining. The proposed system simultaneously generates operationally feasible rock-breaking poses and a robot-free 3D representation of the workspace. The proposed approach combines image-based instance segmentation with geometric point cloud processing, and operates on embedded hardware at approximately 10 Hz with a total latency of around 675 ms, enabling responsive closed-loop behavior when integrated with a control system. Experimental results in a representative scaled scenario demonstrate that the proposed system is suitable for real-time autonomous impact hammer operation.

Chinese Translation

冲击锤，也称为岩石破碎机，是矿业操作中必不可少的机器，主要用于二次破碎。在地下矿业中，这些机器通常由遥控操作，限制了操作效率。本文提出了一种实时RGB-D感知管道，作为自动化矿业中液压冲击锤操作的一个步骤。所提出的系统同时生成可操作的岩石破碎姿态和无机器人工作空间的3D表示。该方法结合了基于图像的实例分割和几何点云处理，并在嵌入式硬件上以约10 Hz的频率运行，总延迟约为675毫秒，使其在与控制系统集成时能够实现响应迅速的闭环行为。在一个具有代表性的缩放场景中的实验结果表明，所提出的系统适用于实时自主冲击锤操作。

View on arXiv Download PDF AI Translation

cs.RO / 10 / 2607.20771

Emergent Compositional Skills in Mixture-of-Experts VLAs

混合专家模型中的新兴组合技能

Shah, Shlok, Jhaveri, Rhiaan, Kalidoss, Tharun Kumar Tiruppali, Nimonkar, Chirayu, Javali, Ishaan

Abstract

We consider the problem of learning compositional robot policies end-to-end from expert demonstrations, without any pre-specified notion of task decomposition or hierarchy. We ask whether a VLA trained with a simplified Mixture-of-Experts (MoE) action head can emergently learn to decompose tasks into reusable, interpretable primitives. We find that learned experts are heavily reused across tasks and consistently correspond to qualitatively distinct low-level behaviors, suggesting that the router implicitly learns to perform high-level sequencing while experts serve as compositional primitives. Our MoE matches the task performance of a monolithic baseline while demonstrating meaningful expert specialization, a step toward modular, interpretable robot policies that emerge from data alone.

Chinese Translation

我们考虑从专家示范中端到端学习组合机器人策略的问题，而不需要任何预先指定的任务分解或层次结构的概念。我们探讨了一个使用简化的混合专家（Mixture-of-Experts, MoE）动作头训练的变换器（VLA）是否能够新兴地学习将任务分解为可重用的、可解释的原语。我们发现，学习到的专家在不同任务中被大量重用，并且始终对应于质上不同的低级行为，这表明路由器隐式地学习执行高级序列，而专家则作为组合原语。我们的MoE在任务性能上与单一基线相匹配，同时展示了有意义的专家专业化，这是朝着从数据中单独产生的模块化、可解释的机器人策略迈出的一步。

View on arXiv Download PDF AI Translation

cs.RO / 11 / 2607.20772

Socially Consistent Multi-Robot Navigation Using Decoupled Planning and Trajectory Coordination

基于解耦规划与轨迹协调的社会一致性多机器人导航

Sato, Matthew M., Law, Kincho H.

Abstract

The successful integration of mobile robots in human-centric environments requires navigation that is not only safe and efficient, but also predictable and aligned with social conventions, key precursors for human comfort and acceptance. While significant research addresses short-term human-aware planning, these methods often lack mechanisms for ensuring consistent and predictable behaviors across long horizons. Without socially aware long-term planners, local planners are overburdened, resulting in inefficient and locally reactive movements that undermine predictability. This paper introduces a partially decentralized framework that generates predictable and socially consistent multi-robot motion by decoupling global path planning from trajectory coordination. First, we propose a modified A* planner that embeds macroscopic social norms into the planner cost function. Planned paths are shared across mobile robots to collaboratively build a social graph of established routes, which enforces path consistency and reduces future planning effort. Second, we leverage the emergent structure of the socially constrained paths to formulate the multi-robot trajectory coordination problem as a mixed-integer convex program. The convex program enables efficient computation of conflict- free trajectories, scaling effectively to large fleets and supporting dynamic task assignment. Our results demonstrate that enforcing social consistency at the path planning stage produces predictable, socially compliant mobile robot paths and simplifies the otherwise complex problem of multi-robot coordination.

Chinese Translation

在以人为中心的环境中成功整合移动机器人需要导航不仅安全高效，还需可预测并符合社会规范，这是人类舒适度和接受度的关键前提。尽管已有大量研究关注短期人类感知规划，但这些方法往往缺乏确保长期一致和可预测行为的机制。没有社会意识的长期规划者，局部规划者的负担加重，导致低效和局部反应式的运动，从而削弱了可预测性。本文提出了一种部分去中心化的框架，通过将全局路径规划与轨迹协调解耦，生成可预测和社会一致的多机器人运动。首先，我们提出了一种修改过的 A* 规划器，将宏观社会规范嵌入规划成本函数中。规划路径在移动机器人之间共享，以协同构建已建立路线的社会图，从而强制路径一致性并减少未来的规划工作。其次，我们利用社会约束路径的涌现结构，将多机器人轨迹协调问题表述为混合整数凸规划。该凸规划能够高效计算无冲突的轨迹，有效扩展到大型车队，并支持动态任务分配。我们的结果表明，在路径规划阶段强制执行社会一致性可以生成可预测的、符合社会规范的移动机器人路径，并简化本来复杂的多机器人协调问题。

View on arXiv Download PDF AI Translation

cs.RO / 12 / 2607.20785

Robostral Navigate

Majumdar, Arjun, Sooriyarachchi, Avinash, Tibi, Benjamin, Bamford, Chris, Chane-Sane, Elliot, Lample, Guillaume, Chandu, Khyathi Raghavi, Fuh, Ludovic Ho, Poiree, Mathieu, Duchenne, Olivier, Millner, Rosalie, Mishra, Srijan, Cachet, Theo, Chabal, Thomas

Abstract

Deploying navigation systems at scale requires a recipe that minimizes sensor assumptions, generalizes across robot embodiments, and trains efficiently. Yet, today's best systems depend on depth sensors, multi-camera rigs, or pre-built maps, limiting the hardware they support and increasing deployment cost. We introduce Robostral Navigate, an 8B vision-language model built around this scalability objective. The model consumes only a stream of monocular RGB images - the most ubiquitous sensor across robotic platforms and predicts waypoints by pointing to the next target location in the current camera view. Operating purely in image space, rather than robot-specific coordinates, makes the policy naturally robust to changes in camera intrinsics and scene scale, enabling deployment across wheeled, legged, and aerial robots without recalibration. We generate 2.4 million trajectories across 350k simulated scenes to reduce the reliance on real-world data collection and scale easily. We further introduce a prefix-caching training recipe that packs entire episodes into single training sequences, reducing training tokens by 22x and cutting training time from months to days. A tree-based attention mask prevents conditioning on previous ground-truth actions, encouraging visually grounded action prediction, and reinforcement learning is used to further improve exploration and recovery capabilities. On the Room-to-Room and Room-Across-Room in Continuous Environments (R2R-CE and RxR-CE) benchmarks, Robostral Navigate sets a new state of the art. On R2R-CE, it achieves a 77.4% success rate, surpassing the best monocular method by 10.5 points and the strongest depth- or multi-camera system by 5.3 points despite using only a single RGB camera. On RxR-CE, it reaches 75.1% success rate, outperforming all monocular baselines.

Chinese Translation

在大规模部署导航系统时，需要一种能够最小化传感器假设、跨机器人形态进行泛化并高效训练的方案。然而，当前最佳系统依赖于深度传感器、多摄像头设备或预构建地图，这限制了它们支持的硬件并增加了部署成本。我们提出了 Robostral Navigate，这是一个围绕可扩展性目标构建的8B视觉-语言模型。该模型仅使用单目RGB图像流——这是机器人平台上最普遍的传感器，并通过在当前相机视图中指向下一个目标位置来预测航点。该模型纯粹在图像空间中操作，而不是在特定于机器人的坐标系中，这使得策略对相机内部参数和场景规模的变化自然具有鲁棒性，从而能够在轮式、腿式和空中机器人上部署而无需重新校准。我们在35万个模拟场景中生成了240万条轨迹，以减少对现实世界数据收集的依赖并实现轻松扩展。我们进一步提出了一种前缀缓存训练方案，将整个剧集打包成单一训练序列，从而将训练标记减少了22倍，并将训练时间从数月缩短到数天。基于树的注意力掩码防止对先前真实动作的条件化，鼓励视觉基础的动作预测，同时使用强化学习进一步提高探索和恢复能力。在 Room-to-Room 和 Room-Across-Room in Continuous Environments (R2R-CE 和 RxR-CE) 基准测试中，Robostral Navigate 创造了新的最先进水平。在 R2R-CE 中，它达到了77.4%的成功率，超越了最佳单目方法10.5个百分点，并且尽管仅使用单个RGB相机，仍比最强的深度或多摄像头系统高出5.3个百分点。在 RxR-CE 中，它达到了75.1%的成功率，超越了所有单目基线。

View on arXiv Download PDF AI Translation

cs.RO / 13 / 2607.20912

URF: A Unified Robot Control-Policy Framework for Stable Contact Aware Manipulation

URF：一种统一的机器人控制策略框架用于稳定的接触感知操作

Shin, Jiyou, Seo, Youngjin, Won, Jaeseog, Seo, Sungwon, Kim, Hyunjun, Yoon, Seokmin, Luong, Tuan, Moon, Hyungpil

Abstract

Learning-based manipulation policies usually predict robot actions from sensory observations and leave their execution to a separate low-level controller. In rigid contact, this separation can be problematic: the same motion to a virtual target or compliant motion command can lead to unstable contact, tracking error, excessive loading, or tool damage, depending on the low-level controller. In this paper, we propose a \textit{Unified Robot Control-Policy Framework} (URF), which connects compliant action prediction with unified impedance-admittance control. Given multimodal observations, URF predicts a virtual target, a stiffness matrix, and an impedance-admittance switch ratio. The switch ratio determines when the controller should behave more like admittance control for accurate motion tracking and when it should move toward impedance control for safer rigid contact. Because demonstration data do not provide ground-truth environment stiffness, we construct switch-ratio labels from measured contact forces and use them to supervise controller-mode prediction. Across box-flipping and line-pressing tasks, URF achieves higher task success rates while reducing failure modes observed with admittance-only execution, including rapid force buildup, large force oscillations, tool breakage, and robot safety stops. These results suggest that contact-aware policies benefit from predicting not only compliant actions but also the controller behavior used to execute them. Project page: https://jiyou384.github.io/urf_project_page/

Chinese Translation

基于学习的操作策略通常从传感器观测中预测机器人动作，并将其执行交给一个单独的低级控制器。在刚性接触中，这种分离可能会导致问题：相同的运动指向虚拟目标或顺应运动指令可能会导致不稳定的接触、跟踪误差、过载或工具损坏，这取决于低级控制器。在本文中，我们提出了一种 extit{统一机器人控制策略框架}（URF），该框架将顺应性动作预测与统一的阻抗-导纳控制相连接。给定多模态观测，URF预测一个虚拟目标、一个刚度矩阵和一个阻抗-导纳切换比。切换比决定了控制器何时应更像导纳控制以实现精确的运动跟踪，以及何时应转向阻抗控制以实现更安全的刚性接触。由于演示数据未提供真实环境刚度，我们从测量的接触力构建切换比标签，并利用这些标签来监督控制器模式的预测。在翻转盒子和压线任务中，URF实现了更高的任务成功率，同时减少了仅使用导纳执行时观察到的失败模式，包括快速的力积累、大幅度的力振荡、工具破损和机器人安全停机。这些结果表明，接触感知策略不仅从预测顺应性动作中受益，还从预测用于执行这些动作的控制器行为中获益。项目页面：https://jiyou384.github.io/urf_project_page/

View on arXiv Download PDF AI Translation

cs.RO / 14 / 2607.20973

Deep Reinforcement-Learning-Guided Model Predictive Control for Preventing Overtakes in Autonomous Racing

深度强化学习引导的模型预测控制用于防止自动驾驶赛车超车

Xi, Yufei, Liao, Yijie, Ersal, Tulga

Abstract

This paper addresses defensive blocking in autonomous racing, where a vehicle must prevent a faster opponent from overtaking while operating near its dynamic limits. Different from lap-time minimization, we formulate defense as a spatial occupancy regulation problem via a hierarchical reinforcement-learning guided model predictive control framework. A Soft Actor-Critic strategic layer operates in the Frenet domain to generate geometry-aware defensive references, which are embedded into the nonlinear model predictive control formulation as spatial regularization under friction constraints. Evaluated on the Thunderhill West circuit in simulation, the framework increases average overtake time from 8.8 s to 14.6 s while significantly reducing opponent progress. Meanwhile, it allows the vehicle to utilize 83.4% of available tire force. The framework achieves a 33.3 ms mean solve time (13.9 ms std), supporting real-time high-speed adversarial interaction.

Chinese Translation

本文探讨了自动驾驶赛车中的防御性阻挡问题，其中车辆必须在接近其动态极限的情况下防止更快的对手超车。与圈速最小化不同，我们将防御问题表述为一个空间占用调节问题，通过一个分层的强化学习引导的模型预测控制框架来实现。Soft Actor-Critic战略层在Frenet域中操作，以生成具有几何感知的防御参考，这些参考被嵌入到非线性模型预测控制的公式中，作为摩擦约束下的空间正则化。在Thunderhill West赛道的仿真评估中，该框架将平均超车时间从8.8秒提高到14.6秒，同时显著减少了对手的进展。同时，它允许车辆利用83.4%的可用轮胎力。该框架实现了33.3毫秒的平均求解时间（标准差为13.9毫秒），支持实时高速对抗交互。

View on arXiv Download PDF AI Translation

cs.RO / 15 / 2607.20992

Distributed Model-Based Diffusion For Scalable Multi-Robot Trajectory Optimization

可扩展多机器人轨迹优化的分布式模型基础扩散

Lee, Haejoon, Wang, Xinyi, Kim, Taekyung, Panagou, Dimitra

Abstract

Trajectory optimization for multi-robot systems remains a critical challenge, particularly when navigating highly non-convex, non-linear, and non-differentiable environments. While Model-Based Diffusion (MBD) has recently emerged as a promising sampling-based optimization paradigm for single-robot trajectory generation, extending it to multi-robot systems results in a centralized, high-dimensional inference problem that (i) suffers from poor sample efficiency due to the curse of dimensionality and (ii) requires global access to all robots' dynamics, constraints, and objectives. To address this, we propose Distributed Model-Based Diffusion (DMBD), a distributed server-robot framework that decomposes the reverse diffusion process into local conditional reverse diffusion processes. This decomposition enables each robot to iteratively perform denoising independently within its own control subspace while conditioning on the current trajectory estimates of the other robots that are aggregated and broadcast by the server. Extensive simulations in goal swapping, multi-floor coverage, parking, and rush-hour scenarios demonstrate that DMBD achieves strong scalability, solving many challenging coordination tasks in sub-seconds and significantly outperforming existing baselines.

Chinese Translation

多机器人系统的轨迹优化仍然是一个关键挑战，尤其是在高度非凸、非线性和不可微分的环境中进行导航时。尽管模型基础扩散（Model-Based Diffusion, MBD）最近作为一种有前景的基于采样的优化范式出现在单机器人轨迹生成中，但将其扩展到多机器人系统会导致一个集中式的高维推断问题，该问题（i）由于维度诅咒而导致样本效率低下，以及（ii）需要对所有机器人的动态、约束和目标进行全局访问。为了解决这个问题，我们提出了分布式模型基础扩散（Distributed Model-Based Diffusion, DMBD），这是一种分布式服务器-机器人框架，将反向扩散过程分解为局部条件反向扩散过程。这种分解使每个机器人能够在自己的控制子空间内独立地迭代执行去噪，同时以服务器聚合和广播的其他机器人的当前轨迹估计为条件。广泛的仿真结果表明，在目标交换、多层覆盖、停车和高峰时段场景中，DMBD实现了强大的可扩展性，在亚秒级内解决了许多具有挑战性的协调任务，并显著优于现有基准。

View on arXiv Download PDF AI Translation

cs.RO / 16 / 2607.21017

TableVerse: A Large-scale Tabletop Dataset with Real-world Grounded Layouts for Generalizable Manipulation

TableVerse：一个具有真实世界基础布局的大规模桌面数据集，用于可推广的操控

Wang, Boyuan, Zhang, Yue, Xue, Xutao, Song, Xueyu, Sun, Yu

Abstract

The development of generalizable robotic manipulation policies is inherently bounded by the availability of large-scale, high-fidelity scene data. While recent automated synthesis methods attempt to bridge this gap via text-to-layout hallucination or simplified procedural generation, they frequently suffer from physical implausibility and fail to capture the complex, dense clutter of actual human environments. In this paper, we introduce TableVerse, a fully automated Real2Sim pipeline that shifts the paradigm from imaginative layout generation to deterministic reconstruction from unstructured, in-the-wild image data. Our framework seamlessly processes unscripted internet media into high-fidelity, simulation-ready tabletop environments with accurate metric scales, authentic topologies, and verified mechanical stability. Furthermore, an automated task-conditioned trajectory generation framework is integrated to synthesize high-quality, collision-free pick-and-place demonstrations. Leveraging this complete pipeline, we construct the TableVerse-100K Dataset, a large-scale corpus comprising 100,000 unique, physically consistent environments paired with interactive manipulation trajectories. By capturing diverse asset compositions, realistic spatial distributions, and high-quality demonstrations, TableVerse-100K establishes a highly scalable and high-fidelity data foundation, providing significant value to facilitate future research in generalizable robotic manipulation tasks.

Chinese Translation

可推广的机器人操控策略的发展本质上受到大规模、高保真场景数据可用性的限制。尽管最近的自动合成方法试图通过文本到布局的幻觉或简化的程序生成来弥补这一差距，但它们常常遭遇物理不合理性，并未能捕捉实际人类环境中的复杂、密集杂乱。在本文中，我们介绍了TableVerse，一个完全自动化的Real2Sim管道，它将范式从想象中的布局生成转变为从非结构化的真实世界图像数据中进行确定性重建。我们的框架无缝地将无脚本的互联网媒体处理为高保真、适合模拟的桌面环境，具有准确的度量尺度、真实的拓扑结构和经过验证的机械稳定性。此外，我们集成了一个自动化的任务条件轨迹生成框架，以合成高质量、无碰撞的取放演示。利用这一完整的管道，我们构建了TableVerse-100K数据集，这是一个包含100,000个独特、物理一致环境及其交互操控轨迹的大规模语料库。通过捕捉多样的资产组合、现实的空间分布和高质量的演示，TableVerse-100K建立了一个高度可扩展和高保真的数据基础，为促进未来可推广的机器人操控任务的研究提供了重要价值。

View on arXiv Download PDF AI Translation

cs.RO / 17 / 2607.21025

ZONDA: Zero-shot Object Navigation with Dynamic Avoidance in Multi-floor Environments

ZONDA：在多层环境中进行零样本目标导航与动态避障

Liang, Shaomin, Liao, Xuanhong, Zhang, Shiyao

Abstract

In Object Goal Navigation task, existing methods are typically restricted to static and single-floor environments, ignoring cross-floor topologies and dynamic pedestrian, which limits their real-world deployment. To address these limitations, we propose ZONDA, a zero-shot object navigation with dynamic avoidance framework. In particular, ZONDA integrates three core components: (i) Heuristic multi-floor planning: from height-difference traversable maps, enables stair traversal and cross-floor exploration without a platform-specific learned controller; (ii) Multi-view target verification: cross-checks multi-scale observations with a vision-language model, significantly reducing false positives; and (iii) Dynamic pedestrian avoidance: explicitly tracks and predicts moving pedestrians to generate anticipatory behaviors. Evaluated on a real Direct Drive Tech TITA biped robot and extensive simulations on HM3D and MP3D, ZONDA achieves significantly improved results. Moreover, ZONDA can maintain robust navigation on the dynamic benchmark HM3D-DYNA compared to the existing baseline.

Chinese Translation

在目标物体导航任务中，现有的方法通常局限于静态和单层环境，忽视了跨层拓扑和动态行人，这限制了它们在现实世界中的应用。为了解决这些局限性，我们提出了ZONDA，一种具有动态避障能力的零样本目标导航框架。具体而言，ZONDA集成了三个核心组件：(i) 启发式多层规划：基于高度差可通行地图，支持楼梯通行和跨层探索，而无需特定平台的学习控制器；(ii) 多视角目标验证：通过视觉-语言模型交叉验证多尺度观察，显著减少误报；以及(iii) 动态行人避障：明确跟踪和预测移动行人，以生成预期行为。在真实的Direct Drive Tech TITA双足机器人和HM3D及MP3D的广泛模拟评估中，ZONDA取得了显著的改进结果。此外，与现有基线相比，ZONDA在动态基准HM3D-DYNA上能够保持稳健的导航能力。

View on arXiv Download PDF AI Translation

cs.RO / 18 / 2607.21043

A Real-Time Generalized Nash Equilibrium Framework for Interaction-Aware Autonomous Driving in Mixed Traffic

一种实时广义纳什均衡框架用于混合交通中的互动感知自动驾驶

Naidja, Nouhed, Rahal, Mohamed-Cherif, Pechberti, Steve, Font, Stéphane, Sandou, Guillaume, Revilloud, Marc

Abstract

Safe and efficient navigation in mixed-traffic environments remains a critical challenge for Autonomous Vehicles (AVs), primarily due to the complex interdependence between the AV's decisions and the unpredictable reactions of human drivers. This paper introduces a comprehensive decision-making framework that formulates the driving interaction as a Generalized Nash Equilibrium Problem (GNEP). Unlike decoupled optimization approaches, this framework explicitly models shared safety and geometric constraints, ensuring that the feasibility of the AV's strategy is dynamically linked to the opponent's actions. To solve this non-convex problem in real-time, we propose a dedicated solver based on Particle Swarm Optimization (PSO). The complete architecture was validated on a test track using a real autonomous Renault Zo\'e interacting with a human driver. Experimental results demonstrate the system's ability to handle critical scenarios by generating comfortable, human-like trajectories. Benchmarks confirm the solver's operational feasibility, achieving convergence in under 50 ms.

Chinese Translation

在混合交通环境中安全高效的导航仍然是自动驾驶汽车（AVs）面临的一个关键挑战，主要是由于自动驾驶汽车的决策与人类驾驶员的不可预测反应之间的复杂相互依赖关系。本文提出了一种全面的决策框架，将驾驶互动形式化为广义纳什均衡问题（GNEP）。与解耦优化方法不同，该框架明确建模了共享的安全和几何约束，确保自动驾驶汽车策略的可行性与对手的行为动态相关联。为了解决这一实时非凸问题，我们提出了一种基于粒子群优化（PSO）的专用求解器。完整的架构在测试赛道上进行了验证，使用真实的自动驾驶雷诺Zoe与人类驾驶员进行互动。实验结果表明，该系统能够通过生成舒适的人类驾驶轨迹来处理关键场景。基准测试确认了求解器的操作可行性，收敛时间在50毫秒以内。

View on arXiv Download PDF AI Translation

cs.RO / 19 / 2607.21049

GuidedAttention: Interpretable and Correctable Visual Attention for OOD-Robust Robot Manipulation via Imitation Learning

引导注意力：通过模仿学习实现可解释和可纠正的视觉注意力，以增强机器人操作的OOD鲁棒性

Murooka, Masaki, Nakajo, Ryoichi, Shirai, Keisuke, Motoda, Tomohiro, Oh, Hanbit, Hanai, Ryo, Domae, Yukiyasu

Abstract

End-to-end visuomotor policies provide little opportunity for humans to understand or correct the policy's visual attention. We propose GuidedAttention, a visuomotor imitation learning framework that introduces interpretable and correctable visual attention as an explicit intermediate representation. Task-relevant attention keypoints are predicted from camera images and condition a diffusion-based action policy. Users can inspect and optionally correct selected keypoints once at rollout initialization, after which the corrected attention is automatically propagated throughout execution by a tracking module. Experiments in simulation and the real world demonstrate that GuidedAttention consistently improves robot manipulation performance, particularly under positional and appearance out-of-distribution (OOD) conditions.

Chinese Translation

端到端的视觉运动策略几乎没有机会让人类理解或纠正策略的视觉注意力。我们提出了引导注意力（GuidedAttention），这是一种视觉运动模仿学习框架，旨在引入可解释和可纠正的视觉注意力作为一种显式的中间表示。任务相关的注意力关键点从摄像头图像中预测，并作为条件输入到基于扩散的动作策略中。用户可以在执行初始化时检查并选择性地纠正所选关键点，之后纠正后的注意力会通过跟踪模块在执行过程中自动传播。模拟和实际实验表明，引导注意力（GuidedAttention）在机器人操作性能上始终有所提升，尤其是在位置和外观的分布外（OOD）条件下。

View on arXiv Download PDF AI Translation

cs.RO / 20 / 2607.21058

Human-Inspired Framework for Robotic Craniotomy: Integrating Multimodal Fusion and Adaptive Trajectory Adjustment

人类启发的机器人开颅框架：集成多模态融合与自适应轨迹调整

Le, Renzhen, Zhang, Xiao, Wu, Di, Wei, Yuanyu, Zhu, Jiachen, Ying, Zhenzhi, Zhang, Pengfei, Shu, Liming

Abstract

Manual craniotomy is a high-risk, skill-dependent procedure associated with surgeon fatigue and potential dural injury. While robotic approaches have improved safety, existing open-loop systems rely solely on preoperative images and cannot compensate for intraoperative registration errors or tissue deformation. To address this, we propose a human-inspired closed-loop robotic craniotomy framework that intelligently integrates preoperative planning with intraoperative execution. An adaptive dual-contour fusion algorithm is employed to generate trajectories that conform to complex cranial geometries while maintaining a consistent tool-bone relative pose. For intraoperative perception, a multimodal two-stage cross-modal attention block (CMA)-temporal convolutional network (TCN)-Transformer network combined with an adaptive Bayesian filter fuses force and acoustic signals to achieve robust breakthrough detection under varying bone conditions. Upon detection, an in-situ projection-based trajectory adjustment strategy dynamically compensates for depth deviations, enabling safe residual bone isolation. Experiments on bovine ribs show a breakthrough prediction accuracy of 97%, a detection latency of 0.048 +/- 0.097 s, and a maximum overshoot of 0.29 mm. All four ex vivo cranial experiments were successfully completed without dural injury. These results demonstrate that the proposed cybernetic framework enables safe and autonomous craniotomy with highly effective closed-loop control.

Chinese Translation

手动开颅术是一项高风险、依赖技能的手术，伴随外科医生疲劳和潜在的硬膜损伤。尽管机器人方法提高了安全性，但现有的开环系统仅依赖于术前影像，无法补偿术中注册误差或组织变形。为了解决这一问题，我们提出了一种人类启发的闭环机器人开颅框架，智能地将术前规划与术中执行相结合。采用自适应双轮廓融合算法生成符合复杂颅骨几何形状的轨迹，同时保持工具与骨骼之间的相对姿态一致。对于术中感知，结合自适应贝叶斯滤波器的多模态两阶段跨模态注意力块（CMA）-时间卷积网络（TCN）-变换器网络融合了力和声学信号，以在不同骨骼条件下实现稳健的突破检测。在检测到突破后，基于原位投影的轨迹调整策略动态补偿深度偏差，从而实现安全的残余骨隔离。在牛肋骨上的实验显示，突破预测准确率达到97%，检测延迟为0.048 +/- 0.097秒，最大超调为0.29毫米。所有四个体外颅骨实验均成功完成，未发生硬膜损伤。这些结果表明，所提出的控制论框架能够实现安全和自主的开颅术，并具备高效的闭环控制能力。

View on arXiv Download PDF AI Translation

cs.RO / 21 / 2607.21113

RL-MACRO: A Cybernetic Closed-Loop Intelligence Framework for Multimodal Adaptive Robotic Craniotomy

RL-MACRO：一种用于多模态自适应机器人颅骨切除的控制论闭环智能框架

Zhang, Xiao, Li, Jiaxuan, Le, Renzhen, Wu, Di, Sun, Chao, Zhu, Jiachen, Zhang, Haoyuan, Li, Xiang, Liu, Jian, Ying, Zhenzhi, Zhang, Pengfei, Shu, Liming

Abstract

Autonomous robotic craniotomy requires continuous regulation of tool-tissue interactions to mitigate mechanical overload and thermal damage while maintaining surgical efficiency. However, this process is inherently partially observable due to unknown, time-varying tissue properties and the inability to directly measure cutting temperatures under physical occlusion. To address these challenges, we propose RL-MACRO, a cybernetic closed-loop intelligence framework that couples multimodal perception, adaptive decision-making, and robotic execution. This framework empowers the surgical robot to autonomously perceive inaccessible states from partial sensory feedback and dynamically optimize its behaviors under uncertain environment. A CNN-LSTM observer first fuses force and sound feedback to reconstruct the hidden temperature state (R^2=0.939, MAE = 1.717 deg C). This reconstructed temperature, alongside multi-sensor features, forms the belief state for an offline Implicit Q-Learning (IQL) policy. A novel dual-head Actor dynamically coordinates the feed rate, spindle speed, and cutting depth to optimize efficiency within strict safety bounds. These decisions are seamlessly translated into spatial motions via online trajectory re-planning and velocity servoing. Experiments on bovine ribs and six ex vivo goat skulls validate the system's robust perception, adaptive recovery from force/temperature excursions, and smooth execution on irregular surfaces, establishing a data-driven cybernetic paradigm for safe and efficient autonomous bone cutting.

Chinese Translation

自主机器人颅骨切除需要持续调节工具与组织之间的相互作用，以减轻机械过载和热损伤，同时保持手术效率。然而，由于未知的、随时间变化的组织特性以及在物理遮挡下无法直接测量切割温度，这一过程本质上是部分可观察的。为了解决这些挑战，我们提出了RL-MACRO，这是一种控制论闭环智能框架，结合了多模态感知、自适应决策和机器人执行。该框架使手术机器人能够自主感知来自部分传感反馈的不可达状态，并在不确定的环境中动态优化其行为。首先，CNN-LSTM观察者融合力和声音反馈，以重建隐藏的温度状态（R^2=0.939，MAE = 1.717摄氏度）。重建的温度与多传感器特征一起形成离线隐式Q学习（Implicit Q-Learning, IQL）策略的信念状态。一个新颖的双头Actor动态协调进给速率、主轴速度和切割深度，以在严格的安全范围内优化效率。这些决策通过在线轨迹重新规划和速度伺服无缝转化为空间运动。在牛肋骨和六个离体山羊颅骨上的实验验证了系统的稳健感知、自适应从力/温度波动中恢复的能力，以及在不规则表面上的平滑执行，建立了一个数据驱动的控制论范式，以实现安全高效的自主骨切割。

View on arXiv Download PDF AI Translation

cs.RO / 22 / 2607.21227

FORGE-plus: Force-Budgeted Recovery for Contact-Rich Assembly with a Frozen LLM Supervisor

FORGE-plus：基于力预算的接触丰富装配恢复方法，辅以冻结的大型语言模型监督

Rah, Kyupaeck Jeff, Oh, Midum

Abstract

Force-conditioned reinforcement learning (RL) enables tight-clearance assembly under a commanded force ceiling, but practical deployment requires determining an appropriate force limit for each object and recovering from insertion failures without exceeding it. We present a two-layer framework in which a frozen, text-only large language model (LLM) assigns a per-object force ceiling before execution and selects recovery maneuvers from a fixed action menu using compact textual force signatures. The LLM never controls force directly: a low-level controller enforces the force ceiling, the recovery policy cannot increase it, and the hidden breaking-force threshold is known only to the evaluator. We evaluate the framework on fragile bottle placement and 0.4 mm diametral-clearance gear insertion using two grippers (Robotiq 2F-140 and Franka Panda hand). A single policy passes 256/256 evaluation episodes on both fragile and robust objects without breakage, correctly predicts release timing, and completes a full table-pick-and-insert pipeline with a mean peak force of 5.4 N. Under injected in-grip slip, the force-signature recovery strategy resolves 40% and 64% of failures on the two grippers, whereas a press-harder baseline is either ineffective or causes frequent breakage. We also report negative results, including the failure of PPO to solve the task under strict force constraints and unsuccessful learned release strategies. All experiments are conducted in rigid-body simulation with hidden force-threshold breakage; no sim-to-real claim is made.

Chinese Translation

力条件强化学习（RL）在设定的力上限下实现了紧密间隙装配，但实际应用需要为每个物体确定合适的力限制，并在不超过该限制的情况下从插入失败中恢复。我们提出了一个两层框架，其中一个冻结的、仅文本的大型语言模型（LLM）在执行前为每个物体分配一个力上限，并使用紧凑的文本力特征从固定的动作菜单中选择恢复操作。LLM并不直接控制力：低级控制器执行力上限，恢复策略不能增加力上限，而隐藏的破坏力阈值仅为评估者所知。我们在脆弱瓶子放置和0.4毫米直径间隙齿轮插入任务上评估了该框架，使用了两种夹具（Robotiq 2F-140和Franka Panda手）。单一策略在脆弱和坚固物体上均通过了256/256的评估回合，没有发生破损，正确预测了释放时机，并以平均峰值力5.4 N完成了完整的桌面拾取和插入流程。在施加的夹持滑移情况下，力特征恢复策略解决了两种夹具上40%和64%的失败，而加压基线要么无效，要么导致频繁破损。我们还报告了一些负面结果，包括PPO在严格力约束下无法解决任务以及学习的释放策略未成功。所有实验均在具有隐藏力阈值破坏的刚体仿真中进行；未提出从仿真到现实的主张。

View on arXiv Download PDF AI Translation

cs.RO / 23 / 2607.21309

Factorized Spatio-Temporal Convolutions for Human Pose Estimation from Planar Lidar

基于分解时空卷积的人体姿态估计方法：来自平面激光雷达的研究

Arreghini, Simone, Nava, Mirko, Carlotti, Nicholas, Paolillo, Antonio, Giusti, Alessandro

Abstract

Localizing nearby humans and estimating their facing direction are key capabilities for safe navigation and socially aware human-robot interaction. Many pose-estimation pipelines target cameras and 3D LiDAR or assume GPU-class compute, whereas service robots are often equipped only with omnidirectional planar LiDARs and modest onboard processors. We address omnidirectional human detection and relative 2D pose estimation from planar LiDAR sequences with a lightweight network based on Space-Time Blocks, which explicitly separate spatial processing along scan rays from temporal aggregation across scans. Our network processes 360{\deg} LiDAR sequences to output per-ray human presence, distance, and relative orientation. We train it via cross-modal self-supervision from a narrow RGB-D body tracker in the sensors' overlap region, removing the need for manual LiDAR labels. Quantitative experiments show that our approach consistently outperforms a parameter-matched baseline model, reducing errors in distance (-38%), position (-28%), and orientation (-15%). We further benchmark on the public FROG dataset, report real-time CPU inference on a service robot, and validate with in-field demonstrations, supporting its suitability for spatial perception on computationally constrained service robots.

Chinese Translation

定位附近的人类并估计其朝向是安全导航和社会意识人机交互的关键能力。许多姿态估计管道针对相机和3D激光雷达，或假设使用GPU级计算，而服务机器人通常仅配备全向平面激光雷达和适度的机载处理器。我们提出了一种基于时空块（Space-Time Blocks）的轻量级网络，解决了从平面激光雷达序列中进行全向人类检测和相对2D姿态估计的问题，该网络明确将沿扫描光线的空间处理与跨扫描的时间聚合分开。我们的网络处理360°激光雷达序列，以输出每条光线的人类存在、距离和相对方向。我们通过在传感器重叠区域内使用窄RGB-D身体跟踪器的跨模态自我监督进行训练，消除了对手动激光雷达标签的需求。定量实验表明，我们的方法在距离（减少38%）、位置（减少28%）和方向（减少15%）上始终优于参数匹配的基线模型。我们进一步在公共FROG数据集上进行基准测试，报告在服务机器人上的实时CPU推理，并通过实地演示进行验证，支持其在计算受限的服务机器人上的空间感知适用性。

View on arXiv Download PDF AI Translation

cs.RO / 24 / 2607.21341

Grasp, Handover, Rotate: Bimanual Object Reorientation via Compositional Diffusion and Energy-Based Optimization

抓取、交接、旋转：通过组合扩散和基于能量的优化实现双手物体重新定向

Yeung, Wun Lam, Liu, Wenjun, Yu, Yui Cheung, Qin, Zhengyan Lambo, She, Qijin, Li, Heng, Wang, Ziqi, Tan, Ping

Abstract

Bimanual object reorientation - picking an object, handing it over between two arms, and placing it in a desired target pose - is valuable when direct placement from the initial grasp is infeasible due to collisions, kinematic constraints, or poor final orientation. However, achieving this under multiple competing objectives remains challenging. We introduce BiCompoDiff, a compositional diffusion and energy-based framework that jointly optimizes grasp selection, handover, regrasp, and motion planning under multiple constraints. By combining a pretrained grasp diffusion model with bimanual planning energy-based models (EBMs), our method injects gradient guidance during reverse diffusion to enforce collision avoidance, trajectory smoothness (via differentiable inverse kinematics), handover feasibility, and regrasp safety. Annealed MCMC sampling further refines grasp poses over the composite energy landscape. Experiments across diverse simulated household reorientation tasks demonstrate that BiCompoDiff achieves over 20% higher success rates and up to 37% smoother trajectories (measured by joint displacement) compared to strong sampling-based baselines. Real-world validation confirms effective sim-to-real transfer and robust performance on challenging scenes.

Chinese Translation

双手物体重新定向——抓取物体、在两只手臂之间交接并将其放置在期望的目标姿态——在由于碰撞、运动学约束或最终方向不佳而无法直接从初始抓取位置放置时具有重要价值。然而，在多个相互竞争的目标下实现这一点仍然具有挑战性。我们提出了BiCompoDiff，一个组合扩散和基于能量的框架，能够在多个约束下联合优化抓取选择、交接、重新抓取和运动规划。通过将预训练的抓取扩散模型与双手规划的基于能量的模型（EBMs）相结合，我们的方法在反向扩散过程中注入梯度引导，以强制执行碰撞避免、轨迹平滑性（通过可微分的逆运动学）、交接可行性和重新抓取安全性。退火马尔可夫链蒙特卡罗（MCMC）采样进一步在复合能量景观中细化抓取姿态。在多样化的模拟家庭重新定向任务中的实验表明，与强大的基于采样的基线相比，BiCompoDiff的成功率提高了超过20%，轨迹平滑度（以关节位移测量）提高了多达37%。现实世界的验证确认了有效的仿真到现实转移以及在挑战场景中的稳健表现。

View on arXiv Download PDF AI Translation

cs.RO / 25 / 2607.21400

VoLN: Vision-Only Long-Horizon Navigation---Paradigm, Benchmark, and Method

VoLN：仅基于视觉的长距离导航——范式、基准与方法

Lou, Jiabin, Wang, Haopeng, Wang, Yuanshuai, Liu, Xinyu, Lv, Xuxin, Guo, Yuxin, Huang, Lei, Shi, Rongye, Wu, Wenjun

Abstract

Vision-and-Language Navigation (VLN) enables embodied agents to follow natural-language instructions. However, route-level instructions commonly encode spatial priors, such as orientation, distance, and layout, that are not explicitly available from onboard sensing at deployment in open, GPS-denied environments. Benchmark performance under such interfaces therefore jointly reflects visual navigation ability and the use of route structure explicitly supplied by the task description. As a complementary formulation, we propose Vision-Only Long-Horizon Navigation (VoLN), which shifts route-relevant information from externally supplied instructions and global guidance to locally observable in-scene cues. In VoLN, goal views specify the destination, while route-relevant information is available only through locally observable in-scene cues that the agent must detect, interpret, and select online. We instantiate VoLN for aerial navigation through VoLN-UAV, a 7,210-episode benchmark that combines long-horizon goal-directed flight, continuous 3D motion, large viewpoint changes, and context-dependent beacon selection. We further provide VoLN-MLLM as an initial reference baseline. It aligns self-supervised visual features with a structured semantic space and predicts short-horizon waypoint segments from observation history, goal views, retrieved visual--semantic tokens, and proprioception. On the five-environment Test-Unseen split, it obtains success rates of 7.4%, 4.5%, and 1.8% on Easy, Normal, and Hard episodes, respectively. These results provide an initial evaluation of VoLN and reveal substantial remaining challenges in long-horizon evidence integration, cross-view goal matching, and closed-loop stability. Project page: https://admire-ljb.github.io/VoLN-UAV/

Chinese Translation

视觉与语言导航（VLN）使得具身代理能够遵循自然语言指令。然而，路线级指令通常编码空间先验信息，如方向、距离和布局，这些信息在开放的、无GPS环境中并不能通过车载传感器明确获取。因此，在这种接口下的基准性能共同反映了视觉导航能力和任务描述中明确提供的路线结构的使用。作为一种补充的表述，我们提出了仅基于视觉的长距离导航（VoLN），它将与路线相关的信息从外部提供的指令和全局指导转移到可在场景中局部观察到的线索。在VoLN中，目标视图指定了目的地，而与路线相关的信息仅通过代理必须在线检测、解释和选择的可局部观察到的场景线索获得。我们通过VoLN-UAV实例化VoLN，这是一个7210集的基准，结合了长距离目标导向飞行、连续的三维运动、大视角变化和上下文相关的信标选择。我们进一步提供VoLN-MLLM作为初步参考基线。它将自监督视觉特征与结构化语义空间对齐，并从观察历史、目标视图、检索的视觉-语义标记和本体感觉中预测短距离的航点段。在五个环境的Test-Unseen分割中，它在简单、普通和困难集上分别获得了7.4%、4.5%和1.8%的成功率。这些结果为VoLN提供了初步评估，并揭示了在长距离证据整合、跨视图目标匹配和闭环稳定性方面仍然存在的重大挑战。项目页面：https://admire-ljb.github.io/VoLN-UAV/

View on arXiv Download PDF AI Translation

cs.RO / 26 / 2607.21416

GLAM-SLAM: Real-time Gaussian Large-scale Mapping via Flow Densification and Spatial Decomposition

GLAM-SLAM：通过流密集化和空间分解实现实时高斯大规模映射

Mermigkas, Panagiotis, Manetas, Argyris, Maragos, Petros

Abstract

Existing Gaussian-splatting-based monocular Simultaneous Localization and Mapping (SLAM) systems are either tailored to short sequences, are not real-time, or suffer from prohibitive GPU memory requirements, limiting their applicability in realistic, long-horizon scenarios. To address this, we present GLAM-SLAM, a real-time, decoupled Gaussian-splatting SLAM system designed for large-scale outdoor scenes. We ensure lightweight tracking using a robust, feature-based SLAM frontend, while for mapping, we adopt a structured, sparse anchor grid representation that ensures scalable operation and maintains scene coherence across long-term sequences. To satisfy the dense initialization requirements of 3D Gaussian Splatting (3DGS), we introduce a geometry-based flow-densification anchoring strategy using epipolar constraints. Furthermore, by treating mapping as a multi-scene problem, we propose a scene-partitioning strategy that introduces a strong spatial inductive bias via MLP initializations to generate localized Gaussians. We evaluate our system on the challenging, long-sequence KITTI Odometry, Oxford RobotCar, and M'alaga datasets. Extensive ablations and comparisons demonstrate a 15% improvement in reconstruction quality over the second-best performer, while maintaining real-time performance and the ability to scale to longer sequences. Code is publicly available for the benefit of the community.

Chinese Translation

现有基于高斯喷溅的单目同时定位与地图构建（SLAM）系统要么针对短序列进行定制，要么无法实时运行，或者面临高昂的GPU内存需求，这限制了它们在现实长时间场景中的应用。为了解决这一问题，我们提出了GLAM-SLAM，一个为大规模户外场景设计的实时解耦高斯喷溅SLAM系统。我们通过一个稳健的基于特征的SLAM前端确保轻量级跟踪，而在映射方面，我们采用了一种结构化的稀疏锚点网格表示，确保可扩展的操作并在长期序列中保持场景一致性。为了满足3D高斯喷溅（3DGS）的密集初始化要求，我们引入了一种基于几何的流密集化锚定策略，利用极线约束。此外，通过将映射视为一个多场景问题，我们提出了一种场景划分策略，通过多层感知器（MLP）初始化引入强大的空间归纳偏置，以生成局部高斯。我们在具有挑战性的长序列KITTI里程计、牛津机器人汽车和马拉加数据集上评估了我们的系统。大量消融实验和比较表明，在保持实时性能和扩展到更长序列的能力的同时，我们的重建质量比第二名提高了15%。代码已公开，以造福社区。

View on arXiv Download PDF AI Translation

cs.RO / 27 / 2607.21522

GS-Agent: Creating 4D Physical Worlds With Generative Simulation

GS-Agent：利用生成模拟创建四维物理世界

Zhang, Hongxin, Lin, Chunru, Li, Junyan, Xian, Zhou, Wang, Tsun-Hsuan, Gan, Chuang

Abstract

Creating dynamic and physically realistic 4D worlds from natural language descriptions is both fascinating and challenging. Traditional computer graphics methods rely on manual creation, requiring extensive human effort to fine-tune materials, motions, and visual fidelity. Recent advances in generative foundation models have sparked interest in learning to generate such 4D worlds from large-scale data; however, existing methods still struggle to ensure physical plausibility and controllability. In this work, we take a different path by leveraging foundation models to construct an agentic system that emulates how humans traditionally create 4D worlds, yet automates the entire process. We present GS-Agent, an end-to-end multi-agent framework that integrates physics engines in the loop to generate realistic, dynamic, and controllable 4D physical worlds from natural language. Inspired by how humans build 4D worlds, GS-Agent decomposes the task into entity management, covering 3D asset curation, material tuning, placement, and motion control, and rendering configuration, including camera and lighting manipulation. Multiple agents with distinct expertise interact with the physics engine via code, seek multimodal feedback, and collaborate to iteratively construct 4D worlds that align with the given descriptions. Experimental results show that GS-Agent effectively converts natural language into diverse and physically plausible 4D worlds exhibiting rich interactions among liquids, deformable objects, and rigid bodies, while achieving cinematic camera and lighting control. We envision GS-Agent as a foundation for a new paradigm in 4D world generation, empowering creative content creation and physical AI. Project page at https://umass-embodied-agi.github.io/gs-agent/

Chinese Translation

从自然语言描述中创建动态且物理真实的四维世界既令人着迷又充满挑战。传统计算机图形学方法依赖于手动创建，需要大量人力来微调材料、运动和视觉逼真度。最近在生成基础模型方面的进展引发了对从大规模数据中学习生成此类四维世界的兴趣；然而，现有方法仍然难以确保物理的合理性和可控性。在本研究中，我们采取了一条不同的路径，通过利用基础模型构建一个代理系统，模拟人类传统创建四维世界的方式，同时自动化整个过程。我们提出了GS-Agent，一个端到端的多代理框架，将物理引擎集成到生成过程中，以从自然语言生成真实、动态且可控的四维物理世界。受到人类构建四维世界方式的启发，GS-Agent将任务分解为实体管理，包括三维资产策划、材料调节、放置和运动控制，以及渲染配置，包括相机和照明操作。多个具有不同专业知识的代理通过代码与物理引擎互动，寻求多模态反馈，并协作迭代构建与给定描述一致的四维世界。实验结果表明，GS-Agent有效地将自然语言转换为多样且物理合理的四维世界，展现了液体、可变形物体和刚体之间丰富的交互，同时实现了电影级的相机和照明控制。我们设想GS-Agent作为四维世界生成新范式的基础，赋能创意内容创作和物理人工智能。项目页面：https://umass-embodied-agi.github.io/gs-agent/

View on arXiv Download PDF AI Translation

cs.RO / 28 / 2607.21571

Beyond Episodic Evaluation: Memory Architectural Bottlenecks in Sequential Embodied Question Answering

超越情节评估：顺序具身问答中的记忆架构瓶颈

Cai, Zikui, Janga, Kaushal, Dao, Tan Dat, Lee, Seungjae, Dass, Shivin, Seo, Mingyo, Yue, Kaiyu, Kang, Mintong, Pillai, Nandhu, Hoover, Monte, Palnitkar, Aadi, Rawal, Ruchit, Zheng, Ruijie, Li, Bo, Zhu, Yuke, Martín-Martín, Roberto, Goldstein, Tom, Huang, Furong

Abstract

Embodied question answering (EQA) is traditionally evaluated under an episodic formulation, where agents solve each task independently and reset internal state between episodes. However, real-world robots operate continuously and must accumulate, retain, and selectively reuse information acquired from prior interactions. Despite this practical requirement, the architectural mechanisms needed to support sequential memory in EQA remain underexplored. In this work, we investigate how different memory architectures behave when EQA agents are evaluated sequentially, with multiple questions answered in the same scene while memory is carried forward across queries. We find that simply preserving existing memory is often insufficient. Agents that retain only traversability information, such as 2D occupancy maps, remember where the robot has explored but not the visual-semantic evidence needed for later questions. Agents trained on short-horizon episodic data face a different challenge: when exposed to continuous, multi-query histories, their inherited context suffers from severe temporal mismatch, rather than forming a reusable scene representation. To overcome this architectural bottleneck, we highlight the necessity of structured, spatially grounded memory: architectures that map persistent visual observations onto metric 3D geometry preserve visual-semantic evidence in a coherent scene representation. Extensive experiments in simulated environments reveal that this form of memory breaks the accuracy-efficiency tradeoff in sequential settings, simultaneously achieving higher answer accuracy and lower navigation costs. We further validate these findings on a real-world mobile robot, demonstrating that spatially grounded visual memory is critical for enabling continuous, intelligent operation in physical environments.

Chinese Translation

具身问答（EQA）传统上在情节框架下进行评估，其中代理独立解决每个任务，并在情节之间重置内部状态。然而，现实世界中的机器人是连续操作的，必须积累、保留并选择性地重用从先前交互中获得的信息。尽管这一实际需求存在，但支持EQA中顺序记忆所需的架构机制仍然未被充分探索。在本研究中，我们调查了不同记忆架构在顺序评估EQA代理时的表现，即在同一场景中回答多个问题，同时在查询之间保持记忆。我们发现，简单地保留现有记忆往往是不够的。仅保留可遍历性信息（如2D占用图）的代理能够记住机器人探索过的地方，但无法记住后续问题所需的视觉-语义证据。接受短期情节数据训练的代理面临着不同的挑战：当暴露于连续的多查询历史时，其继承的上下文遭遇严重的时间不匹配，而不是形成可重用的场景表示。为了克服这一架构瓶颈，我们强调了结构化、空间基础记忆的必要性：将持久视觉观察映射到度量3D几何体的架构能够在一致的场景表示中保留视觉-语义证据。在模拟环境中进行的广泛实验表明，这种形式的记忆打破了顺序设置中的准确性-效率权衡，同时实现了更高的回答准确性和更低的导航成本。我们进一步在真实世界的移动机器人上验证了这些发现，证明空间基础视觉记忆对于在物理环境中实现连续智能操作至关重要。

View on arXiv Download PDF AI Translation

cs.RO / 29 / 2607.21582

Scale Up Strategically: Learning Compositional Generalization via Bias-Aware Evaluation and Data Collection for Robotic Manipulation

战略性地扩大规模：通过偏见意识评估和数据收集学习组合泛化以实现机器人操作

Qi, Yu, Ye, Zhang, Xu, Xinyi, Lu, Yuxuan, Sandhu, Amitoj, Hu, Boce, Huang, Haojie, Tremblay, Jonathan, Wong, Lawson L. S.

Abstract

Compositional generalization is essential for robot to follow diverse instructions. However, pretrained policies are known to take shortcuts, deferring to salient cues rather than grounding language. We introduce a diagnostic framework that localizes this failure to individual \textit{instruction factors}, \textit{e.g.,} reusable semantic components such as color, verb, object, size, and spatial attribute. Our framework formalizes instruction factor bias, the tendency of fine-tuned policies to over-rely on dominant factors as shortcuts, and quantifies it through two metrics: Factor Dominance Rate (FDR), capturing pairwise bias between factors, and Factor Dominance Hierarchy (FDH), aggregating these into a global ranking. Evaluation on six foundation policies reveals broadly consistent ordering, \textit{i.e.}, color $\geq$ object $\geq$ spatial $\geq$ verb $\geq$ size, with color dominant, and verb and size most under-grounded. We further show the diagnosis is actionable: a bias-aware data collection strategy that reallocates a fixed budget toward under-grounded factors outperforms baselines in simulation and on a real robot using half the demonstrations, thereby enabling more sample-efficient and generalizable policy learning.

Chinese Translation

组合泛化对于机器人遵循多样化指令至关重要。然而，已知预训练策略往往采取捷径，依赖显著线索而非扎根于语言。我们提出了一个诊断框架，将这一失败归因于个别的指令因素，例如可重用的语义组件，如颜色、动词、物体、大小和空间属性。我们的框架形式化了指令因素偏见，即微调策略过度依赖主导因素作为捷径的倾向，并通过两个指标量化这一偏见：因素主导率（Factor Dominance Rate, FDR），捕捉因素之间的成对偏见，以及因素主导层级（Factor Dominance Hierarchy, FDH），将这些因素汇总为一个全球排名。对六个基础策略的评估显示出普遍一致的排序，即颜色 ≥ 物体 ≥ 空间 ≥ 动词 ≥ 大小，其中颜色占主导地位，而动词和大小则最不扎根。我们进一步展示了这一诊断的可操作性：一种偏见意识的数据收集策略，通过将固定预算重新分配给不够扎根的因素，能够在模拟和真实机器人上使用一半的演示超越基线，从而实现更高的样本效率和更具泛化能力的策略学习。

View on arXiv Download PDF AI Translation

cs.RO / 30 / 2607.21588

AXIS: A Growable Community-Driven Data Engine for Scalable Robot Manipulation

AXIS：一个可扩展的社区驱动数据引擎，用于可扩展的机器人操作

Zhao, Mengfei, Huang, Dihong, Tang, Yikai, Li, Peihao, Yan, Mingxuan, Zhuang, Ruiqi, Huang, Yanjia, Wang, Jie, Zhai, Hai, Zhou, Tony, Zhang, Rui, Luo, Zhexi, Huang, Yuchen, Yang, Jianfei, Li, Jiachen

Abstract

Learning effective robot manipulation policies requires diverse, high-quality demonstrations, yet existing data pipelines are often difficult to scale because they rely on specialized hardware, centralized operators, or fixed task suites. We present AXIS, a growable community-driven data engine and benchmark for scalable robot learning, which enables browser-based teleoperation for large-scale demonstration collection, automatically generates and validates new manipulation tasks, and transforms community-collected demonstrations into training-ready data through automated success checking, quality filtering, trajectory smoothing, and visual and physics-based augmentation. The AXIS dataset currently contains 207 diverse tasks and 50K+ trajectories. Meanwhile, AXIS organizes data into task snapshots and evaluates policies with a systematic held-out protocol. We compare vision-language-action (VLA) policies under a unified AXIS evaluation suite and analyze scaling behavior across different data volumes. Continual pretraining on AXIS substantially improves the overall success rate of $\pi_{0.5}$ by 5.8%, outperforms the model pretrained on RoboCasa365 by 37.3%, and exhibits consistent scaling with increasing data volume, with the largest gains observed under layout, sensor-noise, and camera perturbations.

Chinese Translation

学习有效的机器人操作策略需要多样化的高质量示范，然而现有的数据管道往往难以扩展，因为它们依赖于专用硬件、集中式操作员或固定的任务套件。我们提出了AXIS，一个可扩展的社区驱动数据引擎和基准，用于可扩展的机器人学习，它支持基于浏览器的远程操作以进行大规模示范收集，自动生成和验证新的操作任务，并通过自动成功检查、质量过滤、轨迹平滑以及视觉和物理增强，将社区收集的示范转化为可用于训练的数据。目前，AXIS数据集包含207个多样化的任务和超过50,000条轨迹。同时，AXIS将数据组织为任务快照，并通过系统的保留协议评估策略。我们在统一的AXIS评估套件下比较了视觉-语言-动作（VLA）策略，并分析了不同数据量下的扩展行为。在AXIS上进行持续预训练显著提高了$ ext{π}_{0.5}$的整体成功率5.8%，超越了在RoboCasa365上预训练的模型37.3%，并在数据量增加时表现出一致的扩展性，最大增益发生在布局、传感器噪声和相机扰动下。

View on arXiv Download PDF AI Translation

计算机视觉 (Computer Vision)

cs.CV / 1 / 2607.20628

RealVDeblur: One-Step Diffusion for Generalizable Real-World Video Deblurring

RealVDeblur：用于可泛化现实世界视频去模糊的一步扩散方法

Jin, Renbiao, Yang, Mingxin, Chen, Yutian, Zhuang, Junhao, Cai, Xin, Yu, Mulin, Xu, Linning, Yu, Wenxian, Zou, Danping, Guo, Shi, Xue, Tianfan

Abstract

Real-world video deblurring remains challenging due to diverse motion patterns, complex degradations, and the scarcity of realistic training data, yet robust restoration is critical for downstream pipelines such as mobile imaging and 3D reconstruction. This work presents \textbf{RealVDeblur}, an efficient generative framework designed to improve in-the-wild robustness under diverse real capture conditions. First, a large-scale, physically grounded blur synthesis pipeline is constructed from scene-level 3D Gaussian Splatting (3DGS) assets and high-frame-rate videos, providing realistic training data covering both camera-induced and object-motion blur. Second, a video diffusion prior is leveraged for restoration; to better accommodate frame-dependent blur variations, temporal compression in the VAE is disabled and a frame-wise encoding scheme is adopted. For practical deployment on long videos, multi-step diffusion sampling is distilled into an efficient one-step generator, and a training-free Temporal Window Mask stabilizes inference beyond the training horizon with constant memory usage. Extensive experiments on diverse real-world benchmarks demonstrate strong perceptual quality, semantic fidelity, and temporal consistency on unseen videos, as well as improved robustness in downstream 3D reconstruction under severe motion blur. Project page: https://rbjin.github.io/RealVDeblur

Chinese Translation

现实世界视频去模糊仍然面临挑战，原因在于多样的运动模式、复杂的退化以及现实训练数据的稀缺，然而，稳健的恢复对于移动成像和三维重建等下游流程至关重要。本研究提出了 extbf{RealVDeblur}，这是一个高效的生成框架，旨在提高在多样化真实捕捉条件下的鲁棒性。首先，构建了一个大规模、基于物理的模糊合成管道，该管道利用场景级三维高斯点云（3D Gaussian Splatting, 3DGS）资产和高帧率视频，提供覆盖相机引起的模糊和物体运动模糊的真实训练数据。其次，利用视频扩散先验进行恢复；为了更好地适应帧依赖的模糊变化，禁用了变分自编码器（VAE）中的时间压缩，并采用了逐帧编码方案。为了在长视频上实现实际部署，多步扩散采样被提炼为高效的一步生成器，并且无训练的时间窗口掩码在训练范围之外稳定推理，同时保持恒定的内存使用。针对多样化现实世界基准的广泛实验表明，在未见视频上具有强大的感知质量、语义保真度和时间一致性，并在严重运动模糊下提高了下游三维重建的鲁棒性。项目页面：https://rbjin.github.io/RealVDeblur

View on arXiv Download PDF AI Translation

cs.CV / 2 / 2607.20642

Masked Topology Modeling for Self-Supervised Learning on Parametric CAD

用于参数化计算机辅助设计的自监督学习的掩蔽拓扑建模

Jiang, Heinrich, Jang, Jennifer

Abstract

Computer aided design (CAD) is ubiquitous: virtually any modern object was designed using editable CAD tools. However, with the shortage of available CAD datasets in its native editable and parametric format, boundary representation (B-Rep), it is ever more important to develop data-efficient methods for this domain. We present a new self-supervised pretraining task, Masked Topology Modeling (MTM), that leverages the face-adjacency graph, an induced structure unique to B-reps that the encoder can be asked to reconstruct. MTM masks a fraction of edges and trains a small head to predict each masked edge's convexity and curve type from the encoder's post-message-passing face features. We combine MTM with a MoCo-style momentum-queue contrastive learning over B-rep-aware augmentations, a BFS-connected face-region masked-reconstruction objective, and pretraining on the ABC dataset and our new procedurally generated dataset to show strong performance on a number of benchmarks.

Chinese Translation

计算机辅助设计（CAD）无处不在：几乎所有现代物体都是使用可编辑的CAD工具设计的。然而，由于缺乏以其原生可编辑和参数化格式（边界表示，B-Rep）提供的CAD数据集，开发数据高效的方法在该领域变得愈发重要。我们提出了一种新的自监督预训练任务——掩蔽拓扑建模（Masked Topology Modeling, MTM），该任务利用了面邻接图，这是一种独特于B-Rep的诱导结构，编码器可以被要求重构。MTM掩蔽了一部分边，并训练一个小的头部来预测每个掩蔽边的凸性和曲线类型，基于编码器的消息传递后面特征。我们将MTM与MoCo风格的动量队列对比学习、基于B-Rep的增强、广度优先搜索（BFS）连接的面区域掩蔽重建目标以及在ABC数据集和我们新生成的数据集上的预训练相结合，以在多个基准测试中展示出强劲的性能。

View on arXiv Download PDF AI Translation

cs.CV / 3 / 2607.20660

Axolotl3D: a Unified Framework for Faithful 3D Shape Completion

Axolotl3D：一个统一的忠实3D形状补全框架

Hu, Anita, Shugrina, Maria

Abstract

Recent 3D generative models produce high-quality geometry from a single image using large-scale priors and diffusion architectures. However, they assume complete visibility and single-view inputs, limiting applicability in multi-view, occluded, or editing scenarios. Although prior works address these challenges individually, they lack a unified framework for controllable 3D completion under diverse conditioning signals. We present Axolotl3D, a multi-modal and occlusion-aware 3D generation model that jointly conditions on images, visibility masks, camera parameters, and a partial point cloud. The point cloud serves as a geometric anchor promoting faithful shape completion, while camera parameters ensure consistent multi-view alignment in a shared 3D coordinate system. A unified training strategy synthesizes diverse conditioning regimes from large-scale 3D data, enabling robust cross-modal reasoning. Experiments on Toys4K and OmniObject3D demonstrate state-of-the-art performance under both clean and occluded settings, as well as strong results in real-world reconstruction and geometry-consistent editing.

Chinese Translation

最近的3D生成模型利用大规模先验和扩散架构，从单幅图像中生成高质量的几何形状。然而，它们假设输入为完全可见的单视图，这限制了在多视图、遮挡或编辑场景中的应用。尽管已有研究分别解决了这些挑战，但缺乏一个统一的框架以在多样的条件信号下进行可控的3D补全。我们提出了Axolotl3D，这是一种多模态且考虑遮挡的3D生成模型，能够同时基于图像、可见性掩码、相机参数和部分点云进行条件生成。点云作为几何锚点，促进忠实的形状补全，而相机参数则确保在共享的3D坐标系统中实现一致的多视图对齐。统一的训练策略从大规模3D数据中合成多样的条件模式，增强了跨模态推理的鲁棒性。在Toys4K和OmniObject3D上的实验展示了在干净和遮挡设置下的最先进性能，以及在真实世界重建和几何一致编辑中的强大结果。

View on arXiv Download PDF AI Translation

cs.CV / 4 / 2607.20670

ODeform: Learning Continuous 4D Motion for Shape Deformation with Neural ODEs

ODeform：利用神经常微分方程学习形状变形的连续4D运动

Velikova, Yordanka, Saleh, Mahdi, Kuang, Liming, Busam, Benjamin

Abstract

Modeling continuous object deformation is important for many computer vision and robotics tasks, such as manipulation and simulation. Existing approaches rely on learning-based methods or physics simulators to model shape deformations. However, these approaches either use discrete time steps or are too computationally intensive for real-time applications. We present ODeform, a novel extension of Neural Ordinary Differential Equations to continuous 4D dynamics of deformable objects in 3D space. Our method transforms 3D point clouds and physical conditions (like material properties) into a unified latent space. By solving the resulting ordinary differential equations over time, we model deformations as continuous flows within this learned embedding, eliminating the need for discrete time steps while maintaining computational efficiency. We evaluate our approach on unseen physical parameter configurations, showing improved motion prediction accuracy over baseline methods. Our experiments further demonstrate a successful transfer to real 3D captured objects with novel shapes, along with effective interpolation and extrapolation of the learned dynamics. Our code and data will be made publicly available.

Chinese Translation

建模连续物体变形对于许多计算机视觉和机器人任务（如操作和仿真）至关重要。现有方法依赖于基于学习的方法或物理模拟器来建模形状变形。然而，这些方法要么使用离散时间步，要么在实时应用中计算开销过大。我们提出了ODeform，这是神经常微分方程（Neural Ordinary Differential Equations）在三维空间中对可变形物体连续4D动态的一个新扩展。我们的方法将三维点云和物理条件（如材料属性）转化为统一的潜在空间。通过随时间求解得到的常微分方程，我们将变形建模为在这一学习嵌入中的连续流动，消除了对离散时间步的需求，同时保持了计算效率。我们在未见过的物理参数配置上评估了我们的方法，显示出比基线方法更高的运动预测准确性。我们的实验进一步证明了成功转移到具有新形状的真实3D捕获物体，并有效地进行学习动态的插值和外推。我们的代码和数据将公开发布。

View on arXiv Download PDF AI Translation

cs.CV / 5 / 2607.20691

Spatially Grounded Concept Bottleneck Models for Trustworthy Breast Ultrasound Diagnosis

空间基础的概念瓶颈模型用于可信的乳腺超声诊断

Tonmoy, Moshiur Rahman, Che, Dunren, Adarbah, Haitham Y., Noore, Afzel

Abstract

Concept Bottleneck Models provide interpretable-by-design predictions by mediating diagnosis through human-understandable concepts, but in medical imaging, their trustworthiness is often limited by the quality and granularity of available supervision. In particular, predicted concept activations can be driven by irrelevant regions, leading to spatially unfaithful explanations. We study a data-centric spatially grounded Concept Bottleneck Model (SG-CBM) that leverages coarse lesion delineations as weak supervision to encourage anatomically plausible concept evidence. For breast ultrasound, we derive two clinically motivated zones from each lesion mask: (i) an in-lesion region of interest for morphology-related concepts and (ii) a posterior acoustic band for posterior phenomena. We train concept maps using a grouped spatial grounding objective and preserve semantic faithfulness with a linear bottleneck classifier. Across five-fold stratified group cross-validation, the proposed SG-CBM improves diagnostic AUROC and concept macro-AUROC while markedly increasing spatial alignment of concept evidence. We also perform a Train-corrupt/Test-clean annotation-quality stress test to quantify the impact of supervision quality on diagnosis and spatial faithfulness. Overall, the results underscore the need for data-quality-aware supervision design and systematic trustworthiness validation for deployable healthcare AI systems.

Chinese Translation

概念瓶颈模型通过人类可理解的概念中介诊断，提供设计上可解释的预测，但在医学影像中，其可信度常常受到可用监督质量和粒度的限制。特别是，预测的概念激活可能受到无关区域的驱动，导致空间上不真实的解释。我们研究了一种以数据为中心的空间基础概念瓶颈模型（SG-CBM），利用粗略的病灶轮廓作为弱监督，以鼓励解剖上合理的概念证据。对于乳腺超声，我们从每个病灶掩膜中推导出两个临床动机区域：（i）与形态相关概念的病灶内感兴趣区域和（ii）用于后部现象的后部声带。我们使用分组空间基础目标训练概念图，并通过线性瓶颈分类器保持语义的真实性。在五折分层组交叉验证中，所提出的SG-CBM提高了诊断的AUROC和概念宏观AUROC，同时显著增加了概念证据的空间对齐。我们还进行了一项训练-损坏/测试-清晰的注释质量压力测试，以量化监督质量对诊断和空间真实性的影响。总体而言，结果强调了对数据质量敏感的监督设计和可部署医疗AI系统的系统性可信度验证的必要性。

View on arXiv Download PDF AI Translation

cs.CV / 6 / 2607.20692

DS@GT ARC at ImageCLEFmed GANs 2026: Geometric Filtering for Privacy-Preserving CT Slice Generation

DS@GT ARC 在 ImageCLEFmed GANs 2026：用于隐私保护的 CT 切片生成的几何过滤

Regina, Eric, Arnaud, Richard, Cisneros, Samir Hadi

Abstract

We present a privacy-preserving framework for synthetic lung CT slice generation developed for the Image-CLEFmed GANs 2026 challenge. The approach combines Optimal Transport Conditional Flow Matching with privacy-oriented training and a post-generation "Supervisor" pipeline that filters generated candidates in learned geometric latent spaces using autoencoder embeddings, Determinantal Point Processes, and Stein Kernel Thinning. Official results show a strong realism-privacy trade-off, with the best-performing model achieving a Privacy Preservation Score of 0.549 and competitive visual fidelity with an FID of 0.3290. While the proposed geometric filtering substantially reduces nearest-neighbor memorization and membership-inference leakage, persistent patient re-identification scores indicate that preventing direct image copying is not sufficient to remove deeper patient-specific anatomical identity, highlighting an important frontier for future privacy-preserving medical image generation.

Chinese Translation

我们提出了一种隐私保护框架，用于合成肺部 CT 切片的生成，旨在参与 Image-CLEFmed GANs 2026 挑战。该方法结合了最优传输条件流匹配（Optimal Transport Conditional Flow Matching）、面向隐私的训练以及一个后生成的“监督者”管道，该管道利用自编码器嵌入、行列式点过程（Determinantal Point Processes）和斯坦核稀疏（Stein Kernel Thinning）在学习的几何潜在空间中过滤生成的候选项。官方结果显示出强烈的现实性与隐私之间的权衡，表现最佳的模型达到了 0.549 的隐私保护评分，并且在视觉保真度方面具有竞争力，FID 值为 0.3290。尽管所提出的几何过滤显著减少了最近邻记忆和成员推断泄漏，但持续的患者重新识别评分表明，仅仅防止直接图像复制不足以消除更深层次的患者特定解剖身份，这突显了未来隐私保护医疗图像生成的重要前沿。

View on arXiv Download PDF AI Translation

cs.CV / 7 / 2607.20705

U-CFR: Uncertainty-Guided Cascade Forward Refinement for Interactive Segmentation

U-CFR：基于不确定性引导的级联前向细化用于交互式分割

Darko, Elijah Danquah, Xian, Min, Soule, Terence, Yao, Tiankai, Anderson, Matthew William

Abstract

Interactive image segmentation is critical for efficient image annotation; however, existing methods often require many corrective clicks or rely on passive refinement schemes that converge slowly. We propose Uncertainty-Guided Cascade Forward Refinement (U-CFR), a novel inference-time framework that enables models to autonomously self-correct after each user interaction. U-CFR introduces a boundary-aware uncertainty score that fuses segmentation uncertainty, contour gradients, and explicit edge predictions to guide the placement of internal pseudo-clicks. These self-generated clicks target the most ambiguous boundary regions, providing strong corrective signals without additional manual input. To support this process, we design a dual-head network with a shared encoder-decoder backbone: a segmentation head ensures region consistency, while an edge head sharpens boundary alignment. In inference, U-CFR launches a cascade of refinement steps, where each stage leverages the uncertainty-driven pseudo-clicks to refine the mask progressively. Experiments on standard benchmark datasets demonstrate that the proposed U-CFR improves click efficiency, initial mask quality, and boundary accuracy. It reduces the required clicks by over 10% on challenging datasets like Berkeley and offers a more intelligent and efficient interactive annotation.

Chinese Translation

交互式图像分割对于高效的图像标注至关重要；然而，现有方法往往需要大量的纠正点击，或依赖于收敛缓慢的被动细化方案。我们提出了不确定性引导的级联前向细化（U-CFR），这是一种新颖的推理时框架，使模型能够在每次用户交互后自主进行自我修正。U-CFR引入了一种边界感知的不确定性评分，该评分融合了分割不确定性、轮廓梯度和显式边缘预测，以指导内部伪点击的放置。这些自生成的点击目标是最模糊的边界区域，提供强有力的纠正信号，而无需额外的手动输入。为了支持这一过程，我们设计了一个双头网络，具有共享的编码器-解码器主干：分割头确保区域一致性，而边缘头则增强边界对齐。在推理过程中，U-CFR启动一系列细化步骤，每个阶段利用不确定性驱动的伪点击逐步细化掩膜。在标准基准数据集上的实验表明，所提出的U-CFR提高了点击效率、初始掩膜质量和边界准确性。在挑战性数据集如伯克利（Berkeley）上，所需点击减少超过10%，并提供了更智能和高效的交互式标注。

View on arXiv Download PDF AI Translation

cs.CV / 8 / 2607.20780

Rethinking Open-World Video Anomaly Detection: Diagnosing Definition Blindness

重新思考开放世界视频异常检测：诊断定义盲点

Song, Inpyo, Lee, Jangwon

Abstract

Open-world video anomaly detection (OWVAD) is expected to detect events that match a user-specified definition of abnormality. This requirement is stronger than generic anomaly localization: in the same video, changing the definition should change which temporal regions are scored as anomalous. We show that current OWVAD evaluation largely fails to isolate this conditional behavior. Standard VAD metrics and the dynamic-definition protocol can be dominated by target-versus-normal separation, allowing models to obtain strong scores while remaining nearly insensitive to the queried definition. We call this failure mode definition blindness. To explain why it is missed, we decompose dynamic-definition evaluation into target-versus-normal detection and target-versus-other-anomaly discrimination, and find that the former receives 7.2-26.8$\times$ more weight across common VAD benchmarks. Motivated by this diagnosis, we introduce three definition-conditioned evaluation metrics, DC-Disc, DC-Det$\Delta$, and DC-Sel$\Delta$, which progressively remove normal-frame, generic-anomaly, and multi-event selection shortcuts. Experiments on UCF-Crime, XD-Violence, and MSAD reveal that several strong VAD, OWVAD, and general vision language model baselines localize anomalous moments but exhibit weak definition following, often with near-zero definition-response margins. To validate that the failure is actionable, we further introduce DeCoS, a definition-contrastive scoring rule that subtracts anomaly evidence shared across definitions. DeCoS improves the strongest baseline by 7.3-16.0 AUROC points on DC-Disc and 15.5-28.3 points on DC-Det$\Delta$. Overall, our results argue that OWVAD should be evaluated as definition-conditioned anomaly scoring, not as anomaly detection under different prompt labels.

Chinese Translation

开放世界视频异常检测（OWVAD）旨在检测符合用户指定的异常定义的事件。这一要求比一般的异常定位更为严格：在同一视频中，改变定义应当改变被评分为异常的时间区域。我们展示了当前的OWVAD评估在很大程度上未能孤立这种条件行为。标准的视频异常检测（VAD）指标和动态定义协议容易受到目标与正常样本分离的主导影响，使得模型能够在几乎对查询定义不敏感的情况下获得较高的评分。我们将这种失败模式称为定义盲点。为了说明为何这一点被忽视，我们将动态定义评估分解为目标与正常检测以及目标与其他异常的区分，发现前者在常见的VAD基准中占据了7.2-26.8倍的权重。基于这一诊断，我们引入了三种基于定义的评估指标：DC-Disc、DC-DetΔ和DC-SelΔ，逐步消除正常帧、一般异常和多事件选择的捷径。在UCF-Crime、XD-Violence和MSAD上的实验表明，多个强大的VAD、OWVAD和通用视觉语言模型基线能够定位异常时刻，但在定义跟随上表现较弱，通常定义响应边际接近于零。为了验证这一失败是可操作的，我们进一步引入了DeCoS，这是一种定义对比评分规则，通过减去跨定义共享的异常证据来进行评分。DeCoS在DC-Disc上提高了最强基线7.3-16.0 AUROC点，在DC-DetΔ上提高了15.5-28.3点。总体而言，我们的结果表明，OWVAD应作为基于定义的异常评分进行评估，而不是作为不同提示标签下的异常检测。

View on arXiv Download PDF AI Translation

cs.CV / 9 / 2607.20789

3D-GIMP: When 3D Gaussian Inpainting Meets PatchMatch

3D-GIMP：当3D高斯修复遇上PatchMatch

Tian, Xuening, Schmalstieg, Dieter, Mori, Shohei

Abstract

Recent advances in 3D scene editing have leveraged iterative diffusion models to update input views. However, this process is computationally expensive and struggles to produce sharp details. Meanwhile, ``hallucination drift'' frequently introduces multi-view inconsistencies, leading to structural artifacts when rendering novel viewpoints. To address this problem, we present 3D-GIMP (3D Gaussian Inpainting Meets Patch Matching), a novel hybrid paradigm designed for high-fidelity object removal in 3D Gaussian Splatting. Instead of diffusing every view, 3D-GIMP performs a single generative inpainting on a key reference view, which serves as an appearance prior. We then introduce a 3D-aware PatchMatch algorithm to propagate these reference textures across all remaining views via correspondence matching, effectively bypassing the stochastic nature of frame-by-frame diffusion. By prioritizing reconstructive consistency over iterative generation, 3D-GIMP maintains high-frequency details across arbitrary resolutions while ensuring a mathematically consistent 3D reconstruction. Our experiments demonstrate that 3D-GIMP not only achieves competitive inpainting quality as previous methods using diffusion in multiple views, but also outperforms these methods in rendering speed and view consistency.

Chinese Translation

最近在3D场景编辑方面的进展利用了迭代扩散模型来更新输入视图。然而，这一过程计算成本高昂，并且难以产生清晰的细节。同时，“幻觉漂移”常常引入多视图不一致性，导致在渲染新视点时出现结构性伪影。为了解决这个问题，我们提出了3D-GIMP（3D Gaussian Inpainting Meets Patch Matching），一种新颖的混合范式，旨在实现高保真度的3D高斯点云中的物体移除。3D-GIMP并不是对每个视图进行扩散，而是在一个关键参考视图上执行单次生成性修复，该视图作为外观先验。然后，我们引入了一种3D感知的PatchMatch算法，通过对应匹配在所有剩余视图中传播这些参考纹理，有效地绕过了逐帧扩散的随机性。通过优先考虑重建一致性而非迭代生成，3D-GIMP在任意分辨率下保持高频细节，同时确保数学上一致的3D重建。我们的实验表明，3D-GIMP不仅在多视图扩散的质量上与之前的方法具有竞争力，而且在渲染速度和视图一致性方面超越了这些方法。

View on arXiv Download PDF AI Translation

cs.CV / 10 / 2607.20790

Ocular Verification for Virtual Reality

虚拟现实中的眼部验证

Singh, Husanpreet, Tran, Robert, Kharel, Ayushree, Banerjee, Sudipta

Abstract

Virtual reality (VR) headsets (e.g., Meta Quest, Apple Vision Pro) provide a seamless user experience due to their fast, frictionless interaction with the physical world in a simulated environment. User authentication relies on biometric cues such as iris in such headsets. However, traditional iris recognition protocols may not be adequate in cases of unconstrained acquisition, which is typical of VR-based data. In this work, we examine three crucial aspects: (1) evaluating ISO/IEC 29794-6 iris quality metrics on VRBiom dataset and analyzing their limitations, (2) addressing data-specific challenges such as off-axis gaze, non-uniform illumination, and specular reflection using generative models, and (3) performing unimodal (iris, periocular) recognition and multimodal score-level fusion (iris + periocular). We observe that some metrics (e.g., margin adequacy) fail on VR-acquired data; whereas, image adjustments primarily benefit periocular recognition, and multimodal fusion lowers EER by ~11% over unimodal iris recognition performance. We will release the evaluation scripts upon acceptance for reproducibility.

Chinese Translation

虚拟现实（VR）头戴设备（如 Meta Quest、Apple Vision Pro）由于其与物理世界在模拟环境中的快速、无摩擦交互，提供了无缝的用户体验。在这些头戴设备中，用户身份验证依赖于生物特征线索，例如虹膜。然而，传统的虹膜识别协议在非约束获取的情况下可能不够充分，而这在基于 VR 的数据中是典型的。在本研究中，我们考察了三个关键方面：（1）评估 ISO/IEC 29794-6 虹膜质量指标在 VRBiom 数据集上的表现并分析其局限性，（2）利用生成模型解决数据特定的挑战，如偏轴注视、非均匀照明和镜面反射，以及（3）进行单模态（虹膜、眼周）识别和多模态得分级融合（虹膜 + 眼周）。我们观察到一些指标（如边距充分性）在 VR 获取的数据上表现不佳；而图像调整主要有利于眼周识别，多模态融合使得等错误率（EER）比单模态虹膜识别性能降低约 11%。我们将在论文接受后发布评估脚本以便于复现。

View on arXiv Download PDF AI Translation

cs.CV / 11 / 2607.20813

SubSplat: High-Resolution Pixel-aligned 3DGS via Sub-pixel Gaussian Reparameterization

SubSplat：通过亚像素高斯重参数化实现高分辨率像素对齐的3DGS

Lee, Jiun, Kim, Jaekwang, Lee, Sangmin

Abstract

Pixel-aligned Gaussian splatting enables efficient and generalizable novel-view synthesis. However, high-resolution rendering faces a critical trade-off where increasing input resolution improves detail at the expense of quadratically rising network computational cost. Conversely, maintaining low-resolution inputs stabilizes this cost but results in insufficient Gaussian density and artifacts. To address this, we propose SubSplat, which introduces Sub-pixel Gaussian Reparameterizer(SPGR) to subdivide primary Gaussians into fine-grained primitives, restoring structural density directly from low-resolution features. We further enhance the reparameterization quality through feature aggregation, which effectively captures high-frequency details across multiple views. Experiments on RealEstate10K and ACID demonstrate that SubSplat achieves high-fidelity rendering with superior efficiency. Our results validate that the proposed framework successfully resolves the trade-off between reparameterization fidelity and network computational cost inherent in pixel-aligned Gaussian Splatting.

Chinese Translation

像素对齐的高斯溅射使得高效且可泛化的新视图合成成为可能。然而，高分辨率渲染面临着一个关键的权衡，即提高输入分辨率虽然可以改善细节，但会导致网络计算成本呈平方级上升。相反，保持低分辨率输入可以稳定这一成本，但会导致高斯密度不足和伪影。为了解决这个问题，我们提出了SubSplat，引入了亚像素高斯重参数化器（Sub-pixel Gaussian Reparameterizer, SPGR），将主要高斯分解为细粒度的原始元素，直接从低分辨率特征中恢复结构密度。我们进一步通过特征聚合提升重参数化质量，有效捕捉多个视图中的高频细节。在RealEstate10K和ACID上的实验表明，SubSplat实现了高保真渲染且效率优越。我们的结果验证了所提出的框架成功解决了像素对齐高斯溅射中重参数化保真度与网络计算成本之间的权衡。

View on arXiv Download PDF AI Translation

cs.CV / 12 / 2607.20819

Explainable graph attention network for stress recognition (StressGAT) via differential action units

基于差异动作单元的可解释图注意力网络用于压力识别（StressGAT）

Kassiotis, Thomas, Gkikas, Stefanos, Smyrnis, Nikolaos, Giannakakis, Giorgos

Abstract

Stress is a dynamic process characterized by significant individual variability in facial expression. Traditional architectures, such as Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs), often overlook person-specific baselines or lack the representational capacity to model the non-linear temporal progression of distress due to sequential bottlenecks and rigid grid-based constraints. Furthermore, many deep learning models lack the interpretability required for clinical deployment. This study introduces StressGAT, a Graph Attention Network that leverages the relational inductive bias of graph modeling to capture complex facial dynamics that indicate acute stress. By using Differential Action Units, the framework normalizes individual responses relative to neutral baselines to achieve personalized recognition. The proposed model achieves 88.62\% accuracy on a diverse stress-induction cohort (58 participants) using a subject-independent, Leave-One-Subject-Out (LOSO) cross-validation protocol. Beyond predictive accuracy, the architecture integrates a Multiple Instance Learning (MIL) attention mechanism to identify peak stress intervals and reveal distinct expressivity phenotypes. By simultaneously optimizing for accuracy and interpretability, this framework provides a robust, explainable solution for personalized affective monitoring.

Chinese Translation

压力是一种动态过程，其特征在于面部表情的个体差异显著。传统架构，如递归神经网络（RNNs）和卷积神经网络（CNNs），往往忽视个体特定的基线，或缺乏建模因序列瓶颈和刚性网格约束而导致的痛苦非线性时间进展的表示能力。此外，许多深度学习模型缺乏临床应用所需的可解释性。本研究提出了StressGAT，一种图注意力网络，利用图建模的关系归纳偏置来捕捉指示急性压力的复杂面部动态。通过使用差异动作单元，该框架相对于中性基线规范化个体反应，以实现个性化识别。所提模型在一个多样化的压力诱导队列（58名参与者）上使用独立于受试者的留一法交叉验证（Leave-One-Subject-Out, LOSO）协议达到了88.62%的准确率。除了预测准确性外，该架构还集成了多实例学习（Multiple Instance Learning, MIL）注意力机制，以识别高峰压力间隔并揭示不同的表现型特征。通过同时优化准确性和可解释性，该框架为个性化情感监测提供了一种稳健且可解释的解决方案。

View on arXiv Download PDF AI Translation

cs.CV / 13 / 2607.20866

Agentic Designer: Progressive Multi-Agent Collaboration for Structure-Aware Interior Layout Generation

自主设计师：面向结构的室内布局生成的渐进式多智能体协作

Yang, Zhijing, Lin, Haocheng, Xu, Zhihua, Li, Haojie, Wang, Keze, Lin, Liang, Chen, Tianshui

Abstract

Generating realistic interior furniture layouts that strictly adhere to architectural constraints (e.g., walls, doors, and windows) remains a fundamental challenge in automated spatial design. Existing approaches, primarily based on one-shot generation using diffusion models or Large Language Models (LLMs), lack explicit mechanisms for intermediate geometric constraint verification, often resulting in structural collisions and functionally infeasible arrangements under complex room constraints. To address these challenges, we propose Agentic Designer, a progressive, multi-agent framework that formulates structure-aware interior layout generation as an iterative and constraint-verified decision process. By decomposing layout synthesis into modular stages of proposal, verification, and adjustment, the framework coordinates three specialized agents, a Generator, an Evaluator, and a Refiner, through a Progressive Consensus Mechanism. This mechanism enforces stepwise geometric validation and correction before each placement is committed, thereby preventing error accumulation. To facilitate this structure-aware paradigm and standardize evaluation, we establish InStruct, a comprehensive benchmark that integrates a dataset comprising over 18,000 high-quality, parametrically annotated samples with a novel suite of structure-centric metrics. Extensive quantitative evaluations, qualitative analyses, and user studies show that Agentic Designer significantly outperforms state-of-the-art methods, demonstrating substantial improvements in strict structural adherence and functional design coherence.

Chinese Translation

生成严格遵循建筑约束（例如墙壁、门和窗户）的现实室内家具布局仍然是自动化空间设计中的一个基本挑战。现有方法主要基于使用扩散模型或大型语言模型（LLMs）的一次性生成，缺乏明确的中间几何约束验证机制，常常导致在复杂房间约束下出现结构碰撞和功能上不可行的布局。为了解决这些挑战，我们提出了自主设计师（Agentic Designer），这是一个渐进式的多智能体框架，将面向结构的室内布局生成形式化为一个迭代和约束验证的决策过程。通过将布局合成分解为提案、验证和调整的模块化阶段，该框架通过渐进共识机制协调三个专业智能体：生成器（Generator）、评估器（Evaluator）和精炼器（Refiner）。该机制在每次放置之前强制进行逐步的几何验证和修正，从而防止错误积累。为了促进这种面向结构的范式并标准化评估，我们建立了InStruct，这是一个综合基准，整合了超过18,000个高质量、参数化注释样本的数据集，并配备了一套新颖的以结构为中心的指标。广泛的定量评估、定性分析和用户研究表明，自主设计师显著优于最先进的方法，在严格的结构遵循和功能设计一致性方面表现出显著改善。

View on arXiv Download PDF AI Translation

cs.CV / 14 / 2607.20868

ViSTR-Bench: Can MLLMs Reason from Continuous Visual Cues in Dynamic Scenes?

ViSTR-Bench：大型多模态语言模型能否从动态场景中的连续视觉线索进行推理？

Li, Han, Liu, Si, Huang, Zehao, Lyu, Dongxin, Xu, Longfei, Fu, Jiahui, Tian, Daxin, Xiu, Yuliang, Wang, Naiyan

Abstract

Multimodal Large Language Models (MLLMs) have achieved remarkable success across diverse expert-level tasks, but they still struggle with fundamental abilities that humans naturally develop through continuous observation of the real world, such as spatial perception and dynamic reasoning. Recent studies have recognized this gap and introduced dedicated benchmarks to evaluate the spatial-temporal capabilities of MLLMs. However, existing benchmarks mostly focus on static scenes or require exact quantitative predictions, leaving intuitive reasoning from temporal cues largely underexplored. In this paper, we introduce the Visual Spatial-Temporal Reasoning Benchmark (ViSTR-Bench), a novel evaluation suite designed to systematically assess whether MLLMs can perform qualitative reasoning from continuous visual cues in dynamic scenes. Guided by the principles of temporal emphasis, reasoning orientation, and qualitative evaluation, ViSTR-Bench establishes a comprehensive four-dimensional evaluations covering Motion Perception, Spatial Relations, Outcome Prediction, and Physical Dynamics. The benchmark comprises 15 distinct subtasks and 1,340 high-quality video question-answer pairs spanning diverse tabletop, indoor, and outdoor scenarios. Extensive evaluations of a broad spectrum of state-of-the-art proprietary, open-source, and specialized spatial MLLMs reveal that, despite their strong general video understanding capabilities, current models still face substantial bottlenecks in complex spatial-temporal reasoning and remain far below human performance.

Chinese Translation

多模态大型语言模型（MLLMs）在各种专家级任务中取得了显著成功，但在一些人类通过对现实世界的持续观察自然发展出的基本能力上仍然存在困难，例如空间感知和动态推理。近期研究已认识到这一差距，并引入了专门的基准测试来评估MLLMs的时空能力。然而，现有基准测试大多集中于静态场景或要求精确的定量预测，导致从时间线索进行直观推理的研究尚未得到充分探索。本文介绍了视觉时空推理基准（ViSTR-Bench），这是一个新颖的评估工具，旨在系统性地评估MLLMs是否能够从动态场景中的连续视觉线索进行定性推理。在时间强调、推理导向和定性评估的原则指导下，ViSTR-Bench建立了一个涵盖运动感知、空间关系、结果预测和物理动态的全面四维评估。该基准包括15个不同的子任务和1,340对高质量的视频问答对，涵盖了多种桌面、室内和室外场景。对广泛的最先进的专有、开源和专业空间MLLMs的广泛评估表明，尽管它们在视频理解能力上表现强劲，但当前模型在复杂的时空推理方面仍面临显著瓶颈，远未达到人类的表现水平。

View on arXiv Download PDF AI Translation

cs.CV / 15 / 2607.20874

Webly Supervised Multi-Label Recognition: Evaluation Benchmark and Dual-Branch Multi-Label Contrastive Learning

网络监督的多标签识别：评估基准与双分支多标签对比学习

Xu, Zhihua, Yang, Zhijing, Yang, Yufeng, Chen, Tianshui

Abstract

Training deep learning models with freely available web images can reduce their dependence on costly manual annotations. Although webly supervised learning has been widely studied for single-label recognition, its multi-label counterpart remains underexplored, partly due to the lack of unified benchmarks and fair comparison protocols. To address this gap, we construct a benchmark for webly supervised multi-label recognition (WS-MLR), including Web-COCO and Web-Pascal, and re-implement representative baselines under a unified setting. The two datasets cover the same 80 and 20 categories as MS-COCO and Pascal VOC, respectively, and contain about 300 thousand images retrieved from the Internet using category-word combinations as search keywords. We further propose a Dual-Branch Multi-Label Contrastive Learning (DBMLCL) framework, which learns category-specific instance-level and category-level representations together with their similarities to identify and correct noisy labels. Extensive experiments on the benchmark demonstrate that DBMLCL achieves superior performance compared to representative baselines.

Chinese Translation

利用自由获取的网络图像训练深度学习模型可以减少对昂贵人工标注的依赖。尽管网络监督学习在单标签识别方面得到了广泛研究，但其多标签对应的研究仍然相对较少，部分原因是缺乏统一的基准和公平的比较协议。为了解决这一问题，我们构建了一个网络监督多标签识别（Webly Supervised Multi-Label Recognition, WS-MLR）基准，包括Web-COCO和Web-Pascal，并在统一设置下重新实现了代表性的基线。两个数据集分别覆盖与MS-COCO和Pascal VOC相同的80个和20个类别，并包含约30万张通过类别词组合作为搜索关键词从互联网检索的图像。我们进一步提出了一种双分支多标签对比学习（Dual-Branch Multi-Label Contrastive Learning, DBMLCL）框架，该框架学习类别特定的实例级和类别级表示及其相似性，以识别和纠正噪声标签。在基准上的大量实验表明，DBMLCL相比于代表性基线实现了更优的性能。

View on arXiv Download PDF AI Translation

cs.CV / 16 / 2607.20883

WhereEdit: Mask-aware Local Latent Editing for One-Step Image Editing

WhereEdit：面向掩膜的局部潜在编辑用于一步图像编辑

Hu, Ming, Dou, Mingyu, Yin, Jianfu, Zhang, Miaomiao, Hu, Cong, Wang, Yao, Hu, Bingliang, Wang, Quan

Abstract

Recent one-step text-to-image (T2I) models enable efficient image synthesis and provide new opportunities for real-time image editing. However, existing one-step editing methods primarily rely on text conditioning for semantic transformation, lacking explicit spatial control over \textit{where} to edit. More importantly, even when spatial constraints are introduced, these methods often struggle to achieve strong and stable semantic modifications within the target regions. In this work, we revisit one-step image editing from a spatially controlled perspective and identify two key challenges: discovering editable regions and achieving effective localized semantic transformation. We reveal that existing methods perform global semantic transport, which limits high-intensity local editing under the one-step setting. To address this issue, we propose \textbf{WhereEdit}, a framework that reformulates one-step editing as localized adaptive editing. WhereEdit automatically identifies semantically relevant regions from internal model features and applies adaptive local modulation to enhance target-region editing while preserving non-target areas and structural consistency. Experiments on the PIE-Bench benchmark demonstrate that WhereEdit consistently outperforms existing one-step image editing methods, achieving superior editing quality while maintaining the efficiency of one-step generation. Additional experiments with region-level supervision further highlight the importance of explicit spatial reasoning for high-quality one-step image editing.

Chinese Translation

近期的一步文本到图像（T2I）模型实现了高效的图像合成，并为实时图像编辑提供了新的机会。然而，现有的一步编辑方法主要依赖文本条件进行语义转换，缺乏对编辑位置的明确空间控制。更重要的是，即使引入了空间约束，这些方法在目标区域内往往难以实现强而稳定的语义修改。在本研究中，我们从空间控制的角度重新审视了一步图像编辑，并确定了两个关键挑战：发现可编辑区域和实现有效的局部语义转换。我们揭示了现有方法执行全局语义传输，这限制了在一步设置下进行高强度局部编辑。为了解决这个问题，我们提出了 extbf{WhereEdit}，一个将一步编辑重新定义为局部自适应编辑的框架。WhereEdit自动从内部模型特征中识别出语义相关区域，并应用自适应局部调制来增强目标区域的编辑，同时保持非目标区域和结构的一致性。在PIE-Bench基准上的实验表明，WhereEdit始终优于现有的一步图像编辑方法，实现了更优的编辑质量，同时保持了一步生成的效率。针对区域级监督的额外实验进一步强调了明确空间推理在高质量一步图像编辑中的重要性。

View on arXiv Download PDF AI Translation

cs.CV / 17 / 2607.20889

Engine-Native Editable 3D World Reconstruction with Objects and Lighting

引擎原生可编辑的三维世界重建：对象与光照

Chen, Junhao, Chen, Xinghao, Zhang, Henghaofan, Qiao, Zihao, Zhang, Saining, Li, Yongzhi, Huang, Ruqi, Li, Sisi, Sheng, Yimin, Zhu, Jianyi, Zhao, Hao

Abstract

Editable 3D scene creation requires object instances and lights that can be inspected, moved, and imported into standard engines, yet existing single-image methods largely stop at room-scale geometry, baked/global illumination, or text-driven generation. We introduce Lumera (Light-aware Unified Engine-native Reconstruction and Assembly), a benchmark and reference pipeline for engine-native, light-aware 3D scene parsing from a single image. Lumera-2K is built from 2,513 UE5 projects and provides 3.73M components, 63M object instances, 102.6K engine-native parametric lights, and 95.1K camera views. On this data, Lumera-Box and Lumera-Light adapt VLM to parse object boxes and parametric light tuples (x,y,z,r,g,b,I), which are assembled with per-object mesh reconstruction, HDR environment estimation, and a bounded agentic refinement loop. In a sanitized box benchmark against DetAny3D, SpatialLM, N3D-VLM, and WildDet3D, Lumera-Box obtains the strongest overall detection, geometry, semantic, and layout scores (merged mAP 0.1141, IoU-B 0.2472, F-score 0.2762), while WildDet3D remains stronger on anchor recall. For lights, Lumera-Light recovers almost all non-empty scenes (recall 0.998) but remains limited at individual-light localization (F1 0.209 at 0.5 m); matched lights have median position error 0.261 m, median {\Delta}E2000 4.59, and intensity Pearson r=0.628. These results establish parametric lights as a measurable editable-scene target and expose remaining bottlenecks in relation structure, light recall/intensity, and cross-engine generalization.

Chinese Translation

可编辑的三维场景创建需要能够被检查、移动并导入到标准引擎中的对象实例和光源，而现有的单图像方法在房间规模的几何体、烘焙/全局光照或文本驱动生成方面大多停滞不前。我们介绍了Lumera（光感知统一引擎原生重建与组装），这是一个用于从单一图像解析引擎原生、光感知三维场景的基准和参考管道。Lumera-2K由2513个UE5项目构建，提供了373万个组件、6300万个对象实例、102.6K个引擎原生参数光源和95.1K个相机视角。在这些数据上，Lumera-Box和Lumera-Light适应了VLM以解析对象框和参数光源元组（x,y,z,r,g,b,I），这些元组与每个对象的网格重建、HDR环境估计以及一个有界的代理精炼循环相结合。在与DetAny3D、SpatialLM、N3D-VLM和WildDet3D的清理箱基准测试中，Lumera-Box在整体检测、几何体、语义和布局得分上获得了最强的表现（合并mAP 0.1141，IoU-B 0.2472，F-score 0.2762），而WildDet3D在锚点召回上仍然更强。对于光源，Lumera-Light几乎恢复了所有非空场景（召回率0.998），但在单个光源定位上仍然有限（F1 0.209在0.5米处）；匹配光源的中位位置误差为0.261米，中位{ abla}E2000为4.59，强度的Pearson相关系数r=0.628。这些结果确立了参数光源作为可测量的可编辑场景目标，并揭示了与结构关系、光源召回/强度以及跨引擎泛化相关的剩余瓶颈。

View on arXiv Download PDF AI Translation

cs.CV / 18 / 2607.20897

MAGE-Vein: Multi-Instance Age and Gender Estimation from Finger Vein Images

MAGE-Vein：基于指静脉图像的多实例年龄和性别估计

Tanaka, Katsuki, Ito, Koichi, Aoki, Takafumi, Fujio, Masakazu, Kaga, Yosuke, Oshima, Kanade, Takahashi, Kenta

Abstract

Age estimation from finger vein images has been widely considered impractical due to severe demographic biases in public datasets and physiological confounding factors like gender. To overcome these limitations, we propose MAGE-Vein, a novel multi-instance, multi-task learning framework. Our approach extracts robust structural aging signs by employing a hybrid feature-level fusion of three fingers, effectively suppressing local imaging noise. Furthermore, simultaneous optimization of gender classification conditions the network to effectively eliminate gender-specific vascular variations. Evaluated on a demographically balanced dataset of 402 subjects, MAGE-Vein achieves a mean absolute error of 6.12 years and a correlation of 0.880. Our results not only overturn the conventional consensus regarding the limitations of the finger vein modality but also demonstrate that previous estimation failures were primarily artifacts of biased public datasets. Our code is available at https://github.com/gsisaoki/MAGE-Vein.

Chinese Translation

从指静脉图像中进行年龄估计因公共数据集中的严重人口统计偏差和性别等生理混杂因素而被广泛认为不切实际。为克服这些限制，我们提出了MAGE-Vein，一种新颖的多实例、多任务学习框架。我们的方法通过采用三根手指的混合特征级融合，提取出稳健的结构性衰老特征，有效抑制局部成像噪声。此外，同时优化性别分类条件使网络能够有效消除性别特异性血管变异。在一个包含402名受试者的具有人口统计平衡的数据集上进行评估，MAGE-Vein实现了6.12年的平均绝对误差和0.880的相关性。我们的结果不仅推翻了关于指静脉模态限制的传统共识，还表明以往的估计失败主要是偏见公共数据集的伪影。我们的代码可在 https://github.com/gsisaoki/MAGE-Vein 获取。

View on arXiv Download PDF AI Translation

cs.CV / 19 / 2607.20900

DINO-VPT: Hierarchical Visual Prompt Tuning for Joint Physical-Digital Face Anti-Spoofing

DINO-VPT：用于联合物理-数字人脸防伪的分层视觉提示调优

Gallin-Martel, Pierre, Feng, Mika, Ito, Koichi, Aoki, Takafumi

Abstract

With the increasing diversity of spoofing attacks, there is a growing demand for unified Face Anti-Spoofing (FAS) models capable of detecting both physical and digital threats. While existing Vision-Language Models (VLMs) demonstrate high generalization in this context, they heavily rely on complex multimodal fusion and external text encoders. In this paper, we propose DINO-VPT, a lightweight, vision-only framework leveraging hierarchical visual prompt tuning. By dynamically injecting prompts conditioned on input features via a Prompt Routing Network (PRN), our method effectively disentangles diverse spoofing artifacts without requiring multimodal fusion. Evaluations on the UniAttackData benchmark demonstrate that DINO-VPT achieves higher accuracy than state-of-the-art VLM-based methods. Our results indicate that a properly structured vision-only architecture can achieve state-of-the-art performance in unified FAS without the need for multimodal supervision.

Chinese Translation

随着欺骗攻击多样性的增加，对能够检测物理和数字威胁的统一人脸防伪（FAS）模型的需求日益增长。尽管现有的视觉-语言模型（VLMs）在这一背景下表现出较高的泛化能力，但它们严重依赖复杂的多模态融合和外部文本编码器。本文提出了DINO-VPT，一种轻量级的仅基于视觉的框架，利用分层视觉提示调优。通过通过提示路由网络（Prompt Routing Network, PRN）动态注入基于输入特征的提示，我们的方法有效地解耦了多样的欺骗伪影，而无需多模态融合。在UniAttackData基准上的评估表明，DINO-VPT的准确性超过了最先进的基于VLM的方法。我们的结果表明，适当结构化的仅基于视觉的架构可以在不需要多模态监督的情况下，在统一FAS中实现最先进的性能。

View on arXiv Download PDF AI Translation

cs.CV / 20 / 2607.20903

Sidewalk Moments: Are Richer Representations Always More Human-Aligned? Evidence from City-Walk Videos

人行道时刻：更丰富的表征是否总是更符合人类认知？来自城市漫步视频的证据

Liu, Liu, Tan, Freya Huying, Duarte, Fábio

Abstract

We examine whether richer visual representations yield more human-aligned measures of urban engagement, using 61 first-person city-walk videos from YouTube segmented into over 50,000 ten-second clips and represented across four modalities: spatiotemporal video features, temporally averaged images (TAIs), audio embeddings, and text-based semantic descriptions. Spearman correlation analysis reveals the expected ordering along the temporal-richness continuum, with video features showing the strongest continuous alignment. However, this ordering breaks down under binary classification of high- versus low-engagement moments (the paradigm most commonly used to train perceptual scoring models), where TAIs consistently match or outperform video across most classifiers and quantile thresholds. An independent two-alternative forced-choice study on Amazon Mechanical Turk confirms that this parity reflects human judgment: participants identified engaging moments with comparable accuracy from TAIs and full video clips, while text performed substantially worse and audio remained near chance. Gap analysis reveals a functional dissociation: video features are advantaged in activity-driven scenes with dynamic content, whereas TAIs better align with human judgments in composition-driven scenes dominated by stable spatial structure. These findings challenge the assumption that richer representations are inherently more human-aligned, and suggest that perceptually grounded temporal compression can be a principled alternative to full video encoding.

Chinese Translation

我们研究了更丰富的视觉表征是否能产生更符合人类认知的城市参与度度量，使用了来自YouTube的61个第一人称城市漫步视频，这些视频被分割成超过50,000个十秒片段，并通过四种模态进行表征：时空视频特征、时间平均图像（TAIs）、音频嵌入和基于文本的语义描述。斯皮尔曼相关分析揭示了沿时间丰富度连续体的预期排序，其中视频特征显示出最强的连续一致性。然而，在高参与度与低参与度时刻的二元分类下（这是训练感知评分模型时最常用的范式），这种排序却崩溃，TAIs在大多数分类器和分位数阈值下始终与视频表现相当或更好。亚马逊机械土耳其的独立二选一强制选择研究证实了这种平衡反映了人类判断：参与者从TAIs和完整视频片段中以相似的准确性识别出引人入胜的时刻，而文本表现明显较差，音频则接近随机。差距分析揭示了功能性解离：在内容动态的活动驱动场景中，视频特征具有优势，而在以稳定空间结构为主的构图驱动场景中，TAIs更符合人类判断。这些发现挑战了更丰富的表征固有地更符合人类认知的假设，并建议感知基础的时间压缩可以作为完整视频编码的原则性替代方案。

View on arXiv Download PDF AI Translation

cs.CV / 21 / 2607.20922

FA-LAM: Focus-Aware Large Avatar Model for One-Shot 4D Animatable Gaussian Head

FA-LAM：聚焦感知的大型头像模型用于一次性4D可动画高斯头部生成

Hu, Yingdong, He, Yisheng, Jiang, Yiming, Lin, Zehong, Hoi, Steven, Zhang, Jun

Abstract

We propose FA-LAM, a Focus-Aware Large Avatar Model for one-shot animatable Gaussian head creation, while simultaneously enabling static 3D and dynamic 4D full-head recovery. The core of our method lies in a thorough analysis of the attention mechanisms and the entangled reconstruction and animation training pipeline adopted by prior state-of-the-art approaches. Our analysis identifies two main factors that compromise the quality of 3D full-head generation: (1) incorrect and noisy attention activations, and (2) conflicts between the tasks of reconstruction and animation. To address the first issue, we introduce a symmetric and semantic attention regularization strategy that leverages the inherent semantics and structural symmetry of human heads. To disentangle the objectives of reconstruction and animation, we develop a novel dual-phase training pipeline that separates the model's capabilities for large-view hallucination and animation into distinct modules. Moreover, we enhance our model to support multi-view and streaming 4D reconstruction in an efficient and memory-friendly manner through a core autoregressive modification with tailored visibility-aware token fusion. Collectively, these innovations enable FA-LAM to reconstruct animatable Gaussian full heads with superior quality, particularly in fine facial regions and large viewing angles.

Chinese Translation

我们提出了FA-LAM，一种聚焦感知的大型头像模型，用于一次性可动画高斯头部的创建，同时支持静态3D和动态4D全头部恢复。我们方法的核心在于对以往最先进方法所采用的注意力机制以及交织的重建与动画训练流程进行深入分析。我们的分析识别出影响3D全头生成质量的两个主要因素：（1）不正确和噪声干扰的注意力激活，以及（2）重建与动画任务之间的冲突。为了解决第一个问题，我们引入了一种对称和语义的注意力正则化策略，利用人头的固有语义和结构对称性。为了理清重建与动画的目标，我们开发了一种新颖的双阶段训练流程，将模型在大视角幻觉和动画方面的能力分离为不同的模块。此外，我们通过核心自回归修改和定制的可见性感知令牌融合，增强了模型以高效且节省内存的方式支持多视角和流式4D重建。综合这些创新，FA-LAM能够以更优质的效果重建可动画的高斯全头，特别是在细致的面部区域和大视角下。

View on arXiv Download PDF AI Translation

cs.CV / 22 / 2607.20924

MagicMakeup: A Region-Controllable Diffusion Transformer for High-Fidelity Makeup-Transfer

MagicMakeup：一种区域可控的扩散变换器用于高保真化妆转移

Wang, Ziyi, Zheng, Siming, Yang, Yang, Xu, Shusong, Zhang, Hao, Li, Bo, Zou, Changqing, Jiang, Peng-Tao

Abstract

Makeup-transfer applies the reference makeup to the source face while preserving the source identity. Despite advances in full-face editing by diffusion-based methods, strong regional controllability, makeup fidelity, and identity preservation remain challenging. The reasons are (i) pixel-to-attention misalignment that causes spillover into non-target areas and weakens regional control; (ii) unclear transfer/preservation concept separation under two-image conditioning, leading to coupling between makeup attributes and identity; and (iii) the lack of a high-resolution dataset that is identity-consistent and region-labeled for fine-grained supervision. In this paper, we propose MagicMakeup, a diffusion transformer-based framework for region-controllable and high-fidelity makeup transfer, built on spatial constraints and concept disentanglement. To enable precise region-specific editing while preserving identity, we propose Token-Aligned Region Gating, which aligns pixel masks with attention and applies region-specific logit gating. To clarify the concepts of transfer and preservation, we further introduce Cross-Modal Perception Guidance, which aligns text and image features to enhance cross-modal concept perception. We also design a pipeline for the generation of 1024 x 1024 data pairs through region-specific makeup removal and establish a unified benchmark in synthetic and real settings. Extensive quantitative and qualitative experiments show that MagicMakeup improves regional controllability, makeup fidelity, and identity preservation, with strong robustness across styles, races, and poses.

Chinese Translation

化妆转移将参考化妆应用于源面孔，同时保持源身份。尽管基于扩散的方法在全脸编辑方面取得了进展，但强大的区域可控性、化妆保真度和身份保留仍然面临挑战。原因包括：(i) 像素与注意力的不对齐导致溢出到非目标区域，削弱了区域控制；(ii) 在双图像条件下，转移/保留概念的分离不清，导致化妆属性与身份之间的耦合；(iii) 缺乏高分辨率的数据集，该数据集在身份一致性和区域标注方面适用于细粒度监督。本文提出了MagicMakeup，一种基于扩散变换器的框架，用于区域可控和高保真的化妆转移，建立在空间约束和概念解耦的基础上。为了实现精确的区域特定编辑同时保持身份，我们提出了Token-Aligned Region Gating，它将像素掩码与注意力对齐，并应用区域特定的逻辑门控。为了明确转移和保留的概念，我们进一步引入了Cross-Modal Perception Guidance，它对齐文本和图像特征以增强跨模态概念感知。我们还设计了一个生成1024 x 1024数据对的管道，通过区域特定的化妆去除，并在合成和真实环境中建立了统一的基准。大量的定量和定性实验表明，MagicMakeup提高了区域可控性、化妆保真度和身份保留，在风格、种族和姿势方面具有强大的鲁棒性。

View on arXiv Download PDF AI Translation

cs.CV / 23 / 2607.20940

Ms. Forcing: Efficient Streaming Video Generation with Multi-Scale Patchification and Attention

Ms. Forcing：基于多尺度补丁化和注意力机制的高效流媒体视频生成

Li, Zekun, Cong, Xiaoyan, Li, Hongyu, Dou, Zhiyang, Guo, Chuan, Mittal, Abhay, An, Sizhe, Sridhar, Srinath

Abstract

Streaming video diffusion models have made substantial progress toward interactive and dynamic world simulation, but the nested autoregressive and denoising loops of conventional next-frame generation hinder real-time deployment. Recent rolling-window methods pipeline denoising across multiple consecutive frames at different noise levels, improving throughput and long-horizon stability. However, they tokenize every state at the same fine spatial granularity, leaving substantial noise-dependent redundancy in the joint denoising window. We propose Ms.Forcing, an efficient streaming video generation paradigm that adapts spatial granularity to each state's noise level. Its Multi-Scale Patchification (MSP) assigns coarser patches to noisier states, reducing the active-window token count by 45%, while Multi-Scale Self-Attention (MSSA) matches the density of visible non-sink keys and values to each query scale to further reduce attention cost. Because both schedules are fixed by window position, Ms.Forcing retains a static, hardware-friendly computation graph. We further introduce Homogeneous-Noise-Level DMD (H-DMD), which assembles each fake video from clean predictions sharing the same source noise level, thereby reducing the mismatch between DMD training sequences and inference-time rollouts. The multi-scale design helps offset the additional training cost of backpropagating through overlapping windows. We include both quantitative and qualitative experiments to show that Ms.Forcing reaches 22.84 FPS on a single H200 GPU, 39.6% faster than Rolling Forcing, while significantly improving VBench scores in both short video and long video generation setting.

Chinese Translation

流媒体视频扩散模型在交互式和动态世界模拟方面取得了显著进展，但传统的下一个帧生成中的嵌套自回归和去噪循环阻碍了实时部署。最近的滚动窗口方法在不同噪声水平下对多个连续帧进行去噪处理，提高了吞吐量和长时间稳定性。然而，它们在相同的细粒度空间上对每个状态进行标记，导致联合去噪窗口中存在大量依赖噪声的冗余。我们提出了Ms.Forcing，一种高效的流媒体视频生成范式，能够根据每个状态的噪声水平调整空间粒度。其多尺度补丁化（Multi-Scale Patchification, MSP）为噪声较大的状态分配更粗的补丁，将活动窗口的标记数量减少了45%；而多尺度自注意力（Multi-Scale Self-Attention, MSSA）则将可见非沉没键值的密度与每个查询尺度相匹配，以进一步降低注意力成本。由于这两种调度方式都是由窗口位置固定的，Ms.Forcing 保持了一个静态、硬件友好的计算图。我们进一步引入了均匀噪声水平的动态模式分解（Homogeneous-Noise-Level DMD, H-DMD），该方法从共享相同源噪声水平的干净预测中组装每个假视频，从而减少了 DMD 训练序列与推理时展开之间的错配。多尺度设计有助于抵消通过重叠窗口进行反向传播的额外训练成本。我们包括了定量和定性实验，表明Ms.Forcing在单个H200 GPU上达到了22.84 FPS，比Rolling Forcing快39.6%，同时在短视频和长视频生成设置中显著提高了VBench评分。

View on arXiv Download PDF AI Translation

cs.CV / 24 / 2607.20947

RECO: Region-Aware Compensation for Extrinsic Perturbations in Roadside 3D Detection

RECO：针对路边3D检测中外部扰动的区域感知补偿

Du, Junsheng, He, Zhaocheng, Lu, Yuhuan

Abstract

In intelligent transportation systems, roadside 3D object detection provides wide-area perception crucial for traffic understanding, cooperative early warning, and safe autonomous driving. However, existing methods suffer from high sensitivity to camera extrinsics; even slight deviations (whether manifesting as transient jitter or persistent drift) can be significantly amplified by projective geometry. This cascade results in severe feature misalignment and degraded localization. To mitigate this limitation, we propose RECO, a region-aware extrinsic compensation framework that corrects extrinsics using piecewise 6-DoF pose offsets. RECO predicts a learnable range boundary to partition the scene into near and far regions, estimating region-specific pose corrections. A differentiable sigmoid gate then smoothly blends the two compensated geometries to preserve continuous BEV sampling and facilitate stable optimization. To supervise the refinement of extrinsics, we introduce an auxiliary reprojection loss that compares 2D bounding boxes projected from 3D ground truth against 2D annotations, optimizing it jointly with the standard detection objective. Extensive experiments on the DAIR-V2X-I and Rope3D benchmarks under extrinsic perturbations demonstrate consistent improvements over state-of-the-art baselines across both yaw and $z$-axis deviations. RECO also generalizes from transient perturbations to persistent shifts, maintaining highly competitive performance under strict calibration uncertainty.

Chinese Translation

在智能交通系统中，路边3D物体检测提供了对交通理解、协作预警和安全自主驾驶至关重要的广域感知。然而，现有方法对相机外部参数高度敏感；即使是轻微的偏差（无论是表现为瞬时抖动还是持续漂移）也会通过投影几何被显著放大。这种级联效应导致特征错位严重和定位精度下降。为了解决这一限制，我们提出了RECO，一个区域感知的外部补偿框架，通过分段的6自由度姿态偏移来校正外部参数。RECO预测一个可学习的范围边界，将场景划分为近区域和远区域，估计区域特定的姿态修正。然后，一个可微分的sigmoid门平滑地融合这两种补偿几何，以保持连续的鸟瞰视图采样并促进稳定的优化。为了监督外部参数的精细调整，我们引入了一个辅助重投影损失，该损失比较从3D真实值投影到2D的边界框与2D注释，联合优化标准检测目标。针对DAIR-V2X-I和Rope3D基准在外部扰动下的广泛实验表明，在偏航和$z$轴偏差方面，相较于最先进的基线方法，RECO始终表现出一致的改进。RECO还能够从瞬时扰动推广到持续偏移，在严格的校准不确定性下保持高度竞争的性能。

View on arXiv Download PDF AI Translation

cs.CV / 25 / 2607.20955

FSB-Net: Frequency-Spatial Boundary Network for Brain Stroke Lesion Segmentation in Non-Contrast CT

FSB-Net：用于非对比CT脑卒中病灶分割的频率-空间边界网络

Fan, Linke, Li, Xianglong, Huang, Huixin, Shu, Kai

Abstract

Accurate segmentation of brain stroke lesions in non-contrast computed tomography (NCCT) scans is critical for rapid clinical decision-making, yet remains difficult due to the low contrast between lesion and normal brain tissue, heterogeneous lesion morphology across ischemic and hemorrhagic subtypes, and ambiguous boundaries caused by partial volume effects. Current deep learning approaches primarily optimize region-level overlap but lack explicit boundary modeling, leading to imprecise delineation that can affect volumetric assessment and treatment planning. We propose FSB-Net, a frequency-spatial boundary network that leverages frequency-domain analysis for boundary-aware stroke lesion segmentation. FSB-Net introduces three components: (i) a Wavelet Boundary Detection Head (WBDH) that applies the discrete wavelet transform to multi-scale encoder features, extracting high-frequency sub-bands as boundary representations; (ii) a Frequency-Spatial Cross-Attention Module (FSCAM) that performs bidirectional attention between wavelet boundary features and spatial decoder features for selective boundary enhancement; and (iii) a Spectral Boundary Loss that penalizes high-frequency discrepancies in the Fourier domain to optimize boundary sharpness. Built on a PVTv2-B2 encoder, FSB-Net is evaluated on a public Brain Stroke CT dataset containing both ischemic and hemorrhagic cases. Experimental results show that FSB-Net outperforms U-Net, UNet++, MANet, and DeepLabV3+ across all metrics, achieving state-of-the-art performance in mean Dice, mean IoU, and HD95.

Chinese Translation

在非对比计算机断层扫描（NCCT）中准确分割脑卒中病灶对于快速临床决策至关重要，但由于病灶与正常脑组织之间的低对比度、缺血性和出血性亚型间异质的病灶形态以及由部分体积效应引起的模糊边界，这一任务仍然十分困难。目前的深度学习方法主要优化区域级重叠，但缺乏明确的边界建模，导致不精确的轮廓划分，从而影响体积评估和治疗计划。我们提出了FSB-Net，这是一种利用频域分析进行边界感知的卒中病灶分割的频率-空间边界网络。FSB-Net引入了三个组件：（i）小波边界检测头（WBDH），该组件对多尺度编码器特征应用离散小波变换，提取高频子带作为边界表示；（ii）频率-空间交叉注意模块（FSCAM），该模块在小波边界特征和空间解码器特征之间执行双向注意，以选择性地增强边界；（iii）光谱边界损失，该损失在傅里叶域中惩罚高频差异，以优化边界的清晰度。基于PVTv2-B2编码器，FSB-Net在一个包含缺血性和出血性病例的公共脑卒中CT数据集上进行了评估。实验结果表明，FSB-Net在所有指标上均优于U-Net、UNet++、MANet和DeepLabV3+，在平均Dice、平均IoU和HD95方面达到了最先进的性能。

View on arXiv Download PDF AI Translation

cs.CV / 26 / 2607.20977

Unsupervised Metal Artifact Reduction in Dental CBCT using Fine-tuned Cycle-Consistent Adversarial Networks

基于微调的循环一致性对抗网络的无监督金属伪影去除在牙科锥形束计算机断层扫描中的应用

Chamika, G. L. T., Dhanapala, S. N. A., Nimalaweera, P. H. S. V., Dissanayake, Maheshi B., Jayasinghe, Ruwan D.

Abstract

Metal artifacts generated by dental implants significantly degrade cone-beam computed tomography (CBCT) volumes, obscuring critical anatomical structures and compromising diagnostic precision. To address this, an unsupervised deep learning framework has been proposed for Metal Artifact Reduction (MAR) utilizing a Cycle-Consistent Adversarial Network (CycleGAN) optimized for high-fidelity restoration. Unlike supervised methods that rely on unattainable voxel-aligned paired datasets, the proposed approach leverages an unpaired dataset of approximately 4,000 images, curated from the public ToothFairy dataset. The architecture integrates U-Net-based generators and PatchGAN discriminators, specifically tuned to mitigate generative hallucinations and preserve morphological integrity. Quantitative benchmarking on a held-out test set demonstrates a 34.6\% improvement in the Blind/Referenceless Image Spatial Quality Evaluator (BRISQUE) score, a substantial reduction in Fr\'echet Inception Distance (FID) from 207.03 to 157.04, and a superior Structural Similarity Index Measure (SSIM) of 0.9105. The framework achieves real-time efficiency with a 3.03 ms inference time per slice, effectively suppressing artifacts while preserving anatomical detail. Expert validation confirms high fidelity; however, to ensure reliability in extreme cases, the architecture is recommended as a clinical decision-support tool under human-in-the-loop oversight. By enhancing diagnostic clarity via a scalable software pipeline, this study provides a robust solution for high-fidelity dental implant imaging.

Chinese Translation

牙科植入物产生的金属伪影显著降低了锥形束计算机断层扫描（CBCT）体积的质量，遮蔽了关键的解剖结构并妨碍了诊断精度。为了解决这一问题，提出了一种无监督深度学习框架用于金属伪影去除（MAR），该框架利用优化的循环一致性对抗网络（CycleGAN）实现高保真度的恢复。与依赖于难以获得的体素对齐配对数据集的监督方法不同，所提方法利用了从公共的ToothFairy数据集中整理出的约4000张图像的无配对数据集。该架构集成了基于U-Net的生成器和PatchGAN判别器，特别调优以减轻生成幻觉并保持形态完整性。在保留的测试集上的定量基准测试显示，Blind/Referenceless Image Spatial Quality Evaluator（BRISQUE）得分提高了34.6\%，Fréchet Inception Distance（FID）显著降低，从207.03降至157.04，结构相似性指数测量（SSIM）达到0.9105。该框架实现了实时效率，每个切片的推理时间为3.03毫秒，有效抑制伪影同时保留解剖细节。专家验证确认了高保真度；然而，为确保在极端情况下的可靠性，建议在人工干预下将该架构作为临床决策支持工具。通过可扩展的软件管道增强诊断清晰度，本研究为高保真度的牙科植入成像提供了一个稳健的解决方案。

View on arXiv Download PDF AI Translation

cs.CV / 27 / 2607.20984

Distribution-Alignment Bridge for Uncertainty-Aware Text-to-Video Retrieval

不确定性感知文本到视频检索的分布对齐桥

Chae, Kyeongmo, Lee, Jihoon, Ahn, Sangtae

Abstract

This paper proposes the Distribution-Alignment Bridge (DAB), a framework that reconceptualizes text-to-video retrieval as a distribution alignment task rather than traditional deterministic point matching. By modeling both text and video embeddings as Gaussian distributions defined by mean and variance, DAB explicitly accounts for modality-specific uncertainty. We employ a deterministic, diffusion-inspired bridge to iteratively refine text distributions toward their target video distributions through a truncated refinement process. This approach unifies probabilistic embedding and distributional transformation into a cohesive, end-to-end trainable system. To optimize cross-modal similarity, we introduce a distribution-aware contrastive loss based on Kullback-Leibler divergence. Extensive evaluations on MSR-VTT, MSVD, and VATEX benchmarks confirm that DAB significantly outperforms existing probabilistic and diffusion-based baselines, while providing calibrated uncertainty-aware ranking through bridge-induced distributional margins.

Chinese Translation

本文提出了分布对齐桥（Distribution-Alignment Bridge, DAB），一个将文本到视频检索重新概念化为分布对齐任务的框架，而非传统的确定性点匹配。通过将文本和视频嵌入建模为由均值和方差定义的高斯分布，DAB 明确考虑了特定模态的不确定性。我们采用了一种确定性、受扩散启发的桥接方法，通过截断的细化过程迭代地将文本分布朝向其目标视频分布进行优化。这种方法将概率嵌入和分布变换统一为一个连贯的、端到端可训练的系统。为了优化跨模态相似性，我们引入了一种基于 Kullback-Leibler 散度的分布感知对比损失。在 MSR-VTT、MSVD 和 VATEX 基准上的广泛评估确认，DAB 显著优于现有的概率和基于扩散的基线，同时通过桥接引起的分布边际提供了经过校准的不确定性感知排名。

View on arXiv Download PDF AI Translation

cs.CV / 28 / 2607.20988

HyWorldVLA: A Vision-Language-Action Model with Hybrid World Modeling for Autonomous Driving

HyWorldVLA：一种具有混合世界建模的视觉-语言-动作模型用于自动驾驶

Yu, Quanfu, Wu, Xian, Xu, Hao, Ma, Liulong

Abstract

Vision-Language-Action (VLA) models augmented with world modeling represent a promising paradigm for end-to-end autonomous driving. While pixel-level future prediction enables fine-grained spatiotemporal reasoning, it compromises robustness in noisy driving scenarios. Conversely, latent-based world models alleviate this sensitivity but often incur limited interpretability and representational degradation due to absent pixel-level grounding. To reconcile this trade-off, we propose HyWorldVLA, a hybrid world-VLA framework that unifies pixel-level supervision and latent representation learning. In the pre-training stage, HyWorldVLA predicts video latents encoded by a pre-trained video VAE, while simultaneously reconstructing video frames to provide precise pixel-level grounding. During the subsequent co-fine-tuning phase, the model exclusively predicts latent features, which are fed into an action expert to generate trajectories. Extensive experiments on NAVSIM v1 and v2 benchmarks demonstrate that HyWorldVLA significantly outperforms both pixel-based and latent-based world model baselines. Notably, we present the first comprehensive qualitative and quantitative analysis of world model noise robustness in autonomous driving, establishing a new benchmark for evaluating future architectures.

Chinese Translation

增强世界建模的视觉-语言-动作（VLA）模型代表了一种有前景的端到端自动驾驶范式。尽管像素级未来预测能够实现细粒度的时空推理，但在嘈杂的驾驶场景中却妥协了鲁棒性。相反，基于潜变量的世界模型减轻了这种敏感性，但由于缺乏像素级的基础，往往导致有限的可解释性和表示退化。为了解决这一权衡问题，我们提出了HyWorldVLA，一种混合世界-VLA框架，统一了像素级监督和潜表示学习。在预训练阶段，HyWorldVLA预测由预训练视频变分自编码器（VAE）编码的视频潜变量，同时重建视频帧以提供精确的像素级基础。在随后的共同微调阶段，该模型专门预测潜特征，这些特征被输入到一个动作专家中以生成轨迹。在NAVSIM v1和v2基准上的广泛实验表明，HyWorldVLA显著优于基于像素和基于潜变量的世界模型基线。值得注意的是，我们首次对自动驾驶中的世界模型噪声鲁棒性进行了全面的定性和定量分析，为评估未来架构建立了新的基准。

View on arXiv Download PDF AI Translation

cs.CV / 29 / 2607.20989

Latent Variable-Mediated Cross-Learning for Few-Shot Acoustic Impedance Imaging

基于潜变量介导的少样本声阻抗成像交叉学习

Peng, Junheng, Li, Yong, Wang, Mingwei, Bao, Yi

Abstract

Acoustic impedance imaging is a fundamental yet severely ill-posed problem in subsurface analysis: the seismic wavelet is unknown, observations are band-limited, and labeled well-log samples are extremely scarce (typically <1% of all traces). Existing semi-supervised deep learning methods mitigate few-shot problem by incorporating forward modeling, yet they either rely on inaccurate prior wavelet assumptions or introduce auxiliary networks, leading to unstable optimization and degraded performance. We propose RD-SCL, a novel framework that integrates regularized deconvolution with semi-supervised cross-learning. At its core lies a differentiable, closed-form first-order Tikhonov deconvolution operator that dynamically estimates the latent wavelet in the frequency domain during training, providing stable physics-guided feedback without explicit auxiliary networks and fixed wavelet priors. Building on this operator, we design a symmetric cross-learning that enforces consistency between predictions on labeled and unlabeled data, thereby effectively exploiting abundant unlabeled traces. Extensive experiments on the SEAM and Marmousi 2 benchmarks demonstrate that RD-SCL consistently outperforms state-of-the-art supervised and semi-supervised methods, achieving substantial gains with lower computational cost. With only 56.5k learnable parameters and competitive runtime, RD-SCL offers a practical, physically consistent, and efficient solution for acoustic impedance imaging.

Chinese Translation

声阻抗成像是地下分析中的一个基础但严重病态的问题：地震波形未知，观测数据带限，标记的井曲线样本极为稀缺（通常少于所有轨迹的1%）。现有的半监督深度学习方法通过结合正向建模来缓解少样本问题，但它们要么依赖于不准确的先验波形假设，要么引入辅助网络，导致优化不稳定和性能下降。我们提出了RD-SCL，这是一种将正则化反卷积与半监督交叉学习相结合的新框架。其核心是一个可微分的、封闭形式的一阶Tikhonov反卷积算子，该算子在训练过程中动态估计频域中的潜在波形，提供稳定的物理引导反馈，而无需显式的辅助网络和固定的波形先验。在此算子的基础上，我们设计了一种对称的交叉学习，强制标记和未标记数据的预测之间保持一致性，从而有效利用丰富的未标记轨迹。在SEAM和Marmousi 2基准上的广泛实验表明，RD-SCL始终优于最先进的监督和半监督方法，在降低计算成本的同时实现了显著的性能提升。RD-SCL仅具有56.5k可学习参数和竞争性的运行时间，为声阻抗成像提供了一种实用、物理一致且高效的解决方案。

View on arXiv Download PDF AI Translation

cs.CV / 30 / 2607.20993

Sparse Concept Channels in Frozen 3D CT Vision Encoders

冻结的3D CT视觉编码器中的稀疏概念通道

Nooralahzadeh, Farhad, Bogensperger, Lea, Bluethgen, Christian, Krauthammer, Michael

Abstract

Large vision-language models are becoming increasingly dominant in 3D medical image interpretation, but we rarely know which internal units encode clinical findings or where that information lives in the representation. We first study this on a 3D chest vision-language model (Pillar-0) by probing its frozen vision embeddings. We show that (i) each radiological finding is encoded by a sparse set of ~10 vision-encoder channels that match full-feature classification performance and far exceed a zero-shot text prompting; (ii) turning off the channels tied to one finding, that finding's score collapses while unrelated labels stay stable; and (iii) the same sparse probe replicates on an architecturally unrelated 3D abdominal VLM (Merlin) suggesting a general property of frozen medical encoders. Our training-free concept channel probe (CCP) method, paired with a corpus-derived report template, outperforms published CT-CHAT on clinical efficacy and NLG metrics (F1 0.549 vs. 0.184; BLEU 0.483 vs. 0.373) at 22x lower latency. Our results provide a clear, reproducible characterization of how frozen medical encoders represent findings, demonstrating direct applicability across models.

Chinese Translation

大型视觉-语言模型在3D医学图像解读中变得越来越主导，但我们很少知道哪些内部单元编码临床发现或这些信息在表示中存在于何处。我们首先在一个3D胸部视觉-语言模型（Pillar-0）上研究这一点，通过探测其冻结的视觉嵌入。我们展示了(i) 每个放射学发现由大约10个视觉编码器通道的稀疏集合编码，这些通道的性能与完整特征分类相匹配，并远超零-shot文本提示；(ii) 关闭与某一发现相关的通道时，该发现的得分崩溃，而无关标签保持稳定；(iii) 同样的稀疏探测在一个架构上无关的3D腹部视觉-语言模型（Merlin）中复制，这表明冻结医学编码器的一个普遍特性。我们的无训练概念通道探测（CCP）方法，结合基于语料库的报告模板，在临床有效性和自然语言生成（NLG）指标上超越了已发布的CT-CHAT（F1 0.549 vs. 0.184；BLEU 0.483 vs. 0.373），且延迟降低了22倍。我们的结果提供了一个清晰、可重复的特征描述，说明冻结医学编码器如何表示发现，展示了在不同模型间的直接适用性。

View on arXiv Download PDF AI Translation

cs.CV / 31 / 2607.21004

AUCH-Net: Action Unit-Based Consistency-Aware Hypergraph Network for Cross-Domain Few-Shot Facial Expression Recognition

AUCH-Net：基于动作单元的一致性感知超图网络用于跨域少样本面部表情识别

Qiu, Xinhan, Yan, Yan, Zhu, Rui, Chen, Si, Wang, Hanzi

Abstract

Recently, cross-domain few-shot facial expression recognition (CF-FER) has received considerable attention. However, the performance of existing CF-FER methods is still unsatisfactory due to inferior transferable feature learning under large domain discrepancy and limited target samples. Fortunately, the action units (AUs), which indicate the movements of different facial muscles, provide consistent conceptual semantics for describing expressions within and across domains. Inspired by this, we propose a novel Action Unit-based Consistency-aware Hypergraph Network (AUCH-Net), which constructs consistency-aware hypergraphs on AUs, for CF-FER. Specifically, AUCH-Net presents a new AU feature learning (AFL) module and a new visual feature learning (VFL) module. The AFL module learns AU features under the guidance of a novel relation consistency loss and an AU regularization loss, while the VFL module learns visual features supervised by a relation consistency loss and a classification loss. By learning consistent AU features, AUCH-Net effectively models the connections between AUs and expression categories. As a result, we can bridge the gap between fine-grained facial variations and high-level expression categories, greatly facilitating the learning of transferable feature representations.Extensive experiments on both in-the-lab and in-the-wild datasets show that our method consistently outperforms several state-of-the-art methods. Our results clearly show that modeling the relationships among AUs holds significant potential for FER under cross-domain few-shot scenarios.

Chinese Translation

近年来，跨域少样本面部表情识别（CF-FER）受到了广泛关注。然而，由于在大域差异和有限目标样本下可迁移特征学习的不足，现有的CF-FER方法性能仍不尽如人意。幸运的是，动作单元（AUs）指示不同面部肌肉的运动，为描述跨域和域内的表情提供了一致的概念语义。受此启发，我们提出了一种新颖的基于动作单元的一致性感知超图网络（AUCH-Net），该网络在AUs上构建了一致性感知超图，以用于CF-FER。具体而言，AUCH-Net提出了一个新的AU特征学习（AFL）模块和一个新的视觉特征学习（VFL）模块。AFL模块在新颖的关系一致性损失和AU正则化损失的指导下学习AU特征，而VFL模块则在关系一致性损失和分类损失的监督下学习视觉特征。通过学习一致的AU特征，AUCH-Net有效地建模了AU与表情类别之间的连接。因此，我们可以弥合细粒度面部变化与高层次表情类别之间的差距，从而极大地促进可迁移特征表示的学习。在实验中，我们在实验室和野外数据集上进行了广泛的测试，结果表明我们的方法在多项先进方法中表现出色。我们的结果清晰地表明，建模AU之间的关系在跨域少样本场景下的面部表情识别中具有重要潜力。

View on arXiv Download PDF AI Translation

cs.CV / 32 / 2607.21007

Explainable Deepfake Detection Challenge

可解释的深伪检测挑战

Narang, Abhijeet, Kuckreja, Kartik, Ghosh, Shreya, Khan, Muhammad Haris, Tariq, Usman, Cai, Jianfei, Dhall, Abhinav

Abstract

Deepfake detection is moving beyond binary classification decisions toward systems that can also explain the visual evidence supporting those decisions. This transition is important for real-world verification settings, where diverse users need to understand not only whether an image is manipulated, but also why it is considered suspicious. The Explainable Deepfake Detection Challenge at ACM Multimedia 2026 is designed to benchmark this joint capability. Built on XPlainVerse, a million-scale benchmark for explainable deepfake detection, the challenge evaluates methods on image classification and grounded natural-language explanation generation. Participants submit a real/fake label together with two explanations for each image: a detailed complex explanation for technical users and a concise simple explanation for general users. The evaluation combines classification metrics with semantic similarity, simplicity, and intent-aware grounding metrics that assess whether explanations identify the relevant manipulated entities and supporting visual evidence. The methodologies developed through the challenge will contribute to the development of next-generation explainable deepfake detectors. Evaluation script, baseline models, and accompanying code are available on https://github.com/Abhijeet8901/XPlainVerse-ACMChallenge.

Chinese Translation

深伪检测正从二元分类决策向能够解释支持这些决策的视觉证据的系统转变。这一转变对于现实世界的验证环境至关重要，因为不同的用户不仅需要了解图像是否被篡改，还需要理解为何它被认为是可疑的。2026年ACM多媒体会议的可解释深伪检测挑战旨在基准测试这一联合能力。该挑战基于XPlainVerse，这是一个用于可解释深伪检测的百万规模基准，评估图像分类和基于自然语言的解释生成方法。参与者需要提交一个真实/伪造标签，并为每个图像提供两个解释：一个针对技术用户的详细复杂解释和一个针对普通用户的简洁简单解释。评估结合了分类指标与语义相似性、简洁性和意图感知的基础指标，以评估解释是否识别出相关的被篡改实体及其支持的视觉证据。通过该挑战开发的方法论将有助于下一代可解释深伪检测器的发展。评估脚本、基线模型及相关代码可在https://github.com/Abhijeet8901/XPlainVerse-ACMChallenge获取。

View on arXiv Download PDF AI Translation

cs.CV / 33 / 2607.21022

ProCap: Prominence-guided Object Rectification for Faithful and Comprehensive Video Captioning

ProCap：基于显著性引导的对象校正用于真实且全面的视频字幕生成

Adhikary, Debjyoti Das, Hazra, Aritra, Chakrabarti, Partha Pratim

Abstract

Improving video captioning quality typically demands retraining large vision-language models, an expensive and often impractical requirement. Existing training-free alternatives instead ground captions in detected objects to curb hallucination, but apply only a single, fixed correction pass without prioritizing which objects matter most, leaving semantically significant content omitted. We propose a prominence-aware, iterative post-hoc rectification framework that overcomes both limitations without modifying the underlying captioning model's parameters: a lightweight scoring mechanism ranks detected objects by spatial saliency, temporal persistence, and relational dynamics, and an iterative, prompt-driven refinement loop uses this ranking to progressively inject missing yet contextually relevant objects into the caption over multiple rounds. We validate the framework on MSVD and MSR-VTT using object-grounded automatic metrics, a 110-participant human study, and qualitative comparison against ChatGPT and Gemini; in human evaluation, the framework raises perceived completeness by up to 48% and reduces hallucination by up to 45% relative to a strong pretrained captioning baseline, all without retraining or reference captions. These results position prominence-guided iterative rectification as a lightweight, scalable, and model-agnostic route to more complete and trustworthy video captioning, with direct relevance to accessibility, retrieval, and other multimedia understanding applications.

Chinese Translation

提高视频字幕质量通常需要重新训练大型视觉-语言模型，这是一项昂贵且往往不切实际的要求。现有的无训练替代方案通过将字幕与检测到的对象结合来减少幻觉现象，但仅应用单一的固定校正过程，而未能优先考虑哪些对象最为重要，从而导致语义上重要的内容被遗漏。我们提出了一种显著性感知的迭代后处理校正框架，克服了这两种限制，而无需修改基础字幕模型的参数：一种轻量级评分机制根据空间显著性、时间持续性和关系动态对检测到的对象进行排名，迭代的、基于提示的细化循环利用这一排名逐步将缺失但在上下文中相关的对象注入到字幕中，经过多轮处理。我们在MSVD和MSR-VTT上验证了该框架，使用对象基础的自动评估指标、110名参与者的人类研究，以及与ChatGPT和Gemini的定性比较；在人类评估中，该框架使感知的完整性提高了多达48%，并相对于强大的预训练字幕基线减少了多达45%的幻觉现象，所有这些均无需重新训练或参考字幕。这些结果将基于显著性引导的迭代校正定位为一种轻量级、可扩展且与模型无关的途径，以实现更完整和可信的视频字幕生成，直接与无障碍访问、检索及其他多媒体理解应用相关。

View on arXiv Download PDF AI Translation

cs.CV / 34 / 2607.21023

WAT3R: Feedforward Underwater 3D Reconstruction

WAT3R：前馈式水下三维重建

Xu, Jiayi, Lu, Jiahao, Zheng, Ziqiang, Tan, Yihao, Zhu, Yaolong, Liu, Yuan, Yeung, Sai-Kit

Abstract

Reliable feedforward underwater 3D reconstruction remains challenging due to severe light attenuation and backscattering, which degrade visual quality and disrupt feature consistency across views, leading to inaccurate multi-view geometry. To address this issue, we propose WAT3R, a feed-forward framework for reconstructing 3D scenes directly from underwater images. By leveraging degradation adaptation as a geometry-constrained process, WAT3R integrates a lightweight neural adaptation module to flexibly account for these underwater imaging effects, thereby improving multi-view reconstruction quality. Implemented in a single forward pass, WAT3R directly and efficiently outputs pixel-aligned 3D point maps and camera poses from underwater videos, allowing a high-quality underwater 3D reconstruction. Experiments conducted on the FLSea, SQUID, and USOD10K datasets show that our method consistently outperforms state-of-the-art approaches on 3D reconstruction tasks, including multi-view/monocular depth estimation and camera pose estimation.

Chinese Translation

由于严重的光衰减和后向散射，可靠的前馈式水下三维重建仍然面临挑战，这些因素降低了视觉质量并破坏了视图间的特征一致性，导致多视图几何的不准确。为了解决这一问题，我们提出了WAT3R，一种直接从水下图像重建三维场景的前馈框架。通过将退化适应作为一种几何约束过程，WAT3R集成了一个轻量级的神经适应模块，以灵活应对这些水下成像效应，从而提高多视图重建质量。WAT3R在单次前向传递中实现，直接高效地从水下视频输出像素对齐的三维点图和相机姿态，实现高质量的水下三维重建。在FLSea、SQUID和USOD10K数据集上进行的实验表明，我们的方法在三维重建任务中，包括多视图/单目深度估计和相机姿态估计，始终优于最先进的方法。

View on arXiv Download PDF AI Translation

cs.CV / 35 / 2607.21027

GroupVideo: Multi-Identity Customized Text-to-Video Generation

GroupVideo：多身份定制的文本到视频生成

Song, Xinyang, Wang, Libin, Sun, Jianxin, Li, Qi, Zheng, Dandan, Chen, JingDong, Sun, Zhenan

Abstract

Current identity customized video generation methodologies are predominantly limited to single-identity scenarios, as the lack of explicit identity separation mechanisms often leads to identity confusion in multi-identity settings. Existing multi-identity approaches, which directly extend single-identity frameworks by concatenating face images as input conditions, frequently result in unnatural facial expressions and motions, manifesting as the "copy-paste" phenomenon. To overcome these limitations, we introduce GroupVideo, a novel framework that leverages multiple individual photographs to generate identitycustomized video. Built upon Video Diffusion Transformers, GroupVideo incorporates multimodal identity alignment: visual alignment jointly encodes multiple face images to provide robust identity references, while semantic alignment introduces a semantic perceiver to enhance the naturalness of motions. An ID localization module with spatial guidance is introduced to address identity blending and enhance identity fidelity, along with bounding box constraints and mask regularization loss, to focus on facial regions and improve training efficiency. In response to the shortage of multi-ID video datasets, we have curated a comprehensive high-quality dataset of 20,000 videos, thereby establishing a crucial resource to advance future research in multi-ID video generation. Extensive experiments demonstrate that GroupVideo outperforms existing methods in generating multi-character videos with consistent identities and natural motions.

Chinese Translation

当前的身份定制视频生成方法主要局限于单一身份场景，因为缺乏明确的身份分离机制，往往导致多身份设置中的身份混淆。现有的多身份方法通过将面部图像串联作为输入条件，直接扩展单一身份框架，常常导致不自然的面部表情和动作，表现为“复制粘贴”现象。为了解决这些局限性，我们提出了GroupVideo，一个新颖的框架，利用多张个人照片生成身份定制视频。GroupVideo基于视频扩散变换器（Video Diffusion Transformers），结合了多模态身份对齐：视觉对齐共同编码多张面部图像，以提供稳健的身份参考，而语义对齐则引入语义感知器以增强动作的自然性。我们引入了一个带有空间引导的身份定位模块，以解决身份混合问题并增强身份保真度，同时结合边界框约束和掩膜正则化损失，专注于面部区域并提高训练效率。针对多身份视频数据集的短缺，我们整理了一个包含20,000个视频的高质量综合数据集，从而建立了一个重要资源，以推动未来多身份视频生成的研究。大量实验表明，GroupVideo在生成具有一致身份和自然动作的多角色视频方面优于现有方法。

View on arXiv Download PDF AI Translation

cs.CV / 36 / 2607.21032

Spectral-Spatial Synergistic Guided Network for Hyperspectral Salient Object Detection

用于高光谱显著目标检测的光谱-空间协同引导网络

Peng, Yanyan, Xu, Tingfa, Xiao, Yao, Liu, Peifu, Bai, Shuyan, Xu, Fengxiang, Li, Jianan

Abstract

Hyperspectral salient object detection aims to identify visually salient regions from hyperspectral images. Existing methods often fail because they fundamentally misunderstand the data, confusing incidental spectral variations caused by external factors such as illumination with essential spectral differences caused by the intrinsic material properties of the object. This leads to fragile representations and noisy predictions. To this end, we propose a lightweight and efficient Spectral-Spatial Synergistic Guided Network (S3GNet), with structure perception as the core, to build a closed-loop information flow around spectrum robust modeling, cross-stream co-perception and multi-scale refinement decoding. S3GNet introduces a parameter-free Spectral Structure-Aware Module that leverages spectral derivatives and regional hierarchical modeling to extract intrinsic features of robustness against illumination variations. Our Stream-Aware Attention Module achieves effective spectral-spatial collaboration through inter-stream global interaction and intra-stream spatial guidance. Furthermore, a Progressive Gated Refinement Decoder ensures precise object boundaries and detail recovery by optimally integrating multi-scale features. Experimental results show that S3GNet achieves superior performance in both computational efficiency and detection accuracy compared to existing methods.

Chinese Translation

高光谱显著目标检测旨在从高光谱图像中识别视觉显著区域。现有方法常常失败，因为它们在根本上误解了数据，将由外部因素（如照明）引起的偶然光谱变化与由物体内在材料特性引起的基本光谱差异混淆。这导致了脆弱的表示和嘈杂的预测。为此，我们提出了一种轻量且高效的光谱-空间协同引导网络（S3GNet），以结构感知为核心，围绕光谱鲁棒建模、跨流协同感知和多尺度精细解码构建闭环信息流。S3GNet引入了一种无参数的光谱结构感知模块，利用光谱导数和区域层次建模提取对照明变化具有鲁棒性的内在特征。我们的流感知注意模块通过跨流全局交互和流内空间引导实现有效的光谱-空间协作。此外，渐进式门控精细解码器通过最佳整合多尺度特征确保精确的目标边界和细节恢复。实验结果表明，与现有方法相比，S3GNet在计算效率和检测精度上均表现出优越的性能。

View on arXiv Download PDF AI Translation

cs.CV / 37 / 2607.21036

GeoThreat: Transferable Targeted Adversarial Attacks on Large Vision-Language Models for Remote Sensing Image Interpretation

GeoThreat：针对大型视觉-语言模型的可转移目标对抗攻击在遥感图像解读中的应用

Fu, Yimin, Bai, Yuefeng, Pan, Baicheng, Liu, Zhunga, Ng, Michael K.

Abstract

Adversarial attacks against large vision-language models (LVLMs) serve as an effective means of assessing their robustness in cross-modal semantic understanding. Existing studies mainly focus on corrupting visual inputs to induce predefined erroneous responses in general vision-language tasks, whereas corresponding investigations in remote sensing fields remain largely underexplored. Compared with natural image understanding, remote sensing image interpretation requires joint reasoning over local discriminative cues and global scene context. This poses additional challenges to achieving transferable semantic manipulation toward specified responses under black-box settings. To tackle these challenges, we propose GeoThreat, a transferable targeted adversarial attack method against LVLMs for remote sensing image interpretation. Specifically, GeoThreat modulates adversarial representations in accordance with the target content at both conceptual and perceptual levels. The class tokens from surrogate image encoders are employed as conceptual representations, while perceptual representations are distilled from patch tokens of the adversarial example through collaborative importance estimation. Beyond merely rolling out attention scores across layers, we incorporate adversarial-target similarity gradients to more faithfully characterize the relevance of local visual cues to the intended semantic manipulation. The perceptual representations are then dynamically aligned with target patch tokens in a cross-attentive manner, facilitating the adaptation of local cues toward designated semantic details. Finally, adversarial perturbations are iteratively updated via ensemble-based joint optimization of conceptual calibration and perceptual adaptation. Extensive experiments across diverse LVLMs demonstrate the superiority of GeoThreat in both transferability and controllability.

Chinese Translation

针对大型视觉-语言模型（LVLMs）的对抗攻击是评估其在跨模态语义理解中鲁棒性的一种有效手段。现有研究主要集中于破坏视觉输入，以诱导在一般视觉-语言任务中产生预定义的错误响应，而在遥感领域的相关研究仍然很少。与自然图像理解相比，遥感图像解读需要对局部判别线索和全局场景上下文进行联合推理。这为在黑箱环境下实现针对特定响应的可转移语义操控带来了额外挑战。为应对这些挑战，我们提出了GeoThreat，一种针对LVLMs的可转移目标对抗攻击方法，用于遥感图像解读。具体而言，GeoThreat根据目标内容在概念和感知层面上调节对抗表示。来自替代图像编码器的类别标记被用作概念表示，而感知表示则通过协作重要性估计从对抗样本的补丁标记中提取。我们不仅在各层之间滚动注意力分数，还结合对抗目标相似性梯度，更真实地表征局部视觉线索与预期语义操控的相关性。然后，感知表示以跨注意的方式与目标补丁标记动态对齐，促进局部线索向指定语义细节的适应。最后，通过概念校准和感知适应的集成联合优化，迭代更新对抗扰动。在多种LVLMs上的广泛实验表明，GeoThreat在可转移性和可控性方面具有优越性。

View on arXiv Download PDF AI Translation

cs.CV / 38 / 2607.21050

HyperImageNet: A Large-Scale High-Spatial Resolution Hyperspectral Imagery Classification Benchmark

HyperImageNet：一个大规模高空间分辨率的高光谱影像分类基准

Zeng, Chuguang, Li, Jingtao, Liu, Yinhe, Zhong, Yanfei

Abstract

We present HyperImageNet, a large-scale benchmark for fine-grained hyperspectral land-cover understanding. The dataset contains 26,084 airborne hyperspectral image patches with 224 spectral bands and 138 fine-grained land-cover categories. Unlike existing datasets, HyperImageNet provides raw imagery, pixel-level semantic labels, and object-level instance masks, supporting both semantic and instance segmentation. Furthermore, we establish an open-environment benchmark with strict spatial separation to evaluate representative methods and the HyperFree foundation model. Experimental results demonstrate the effectiveness of HyperImageNet for fine-grained hyperspectral understanding and open-environment remote sensing research.

Chinese Translation

我们提出了HyperImageNet，这是一个用于细粒度高光谱土地覆盖理解的大规模基准数据集。该数据集包含26,084个航空高光谱影像片段，具有224个光谱波段和138个细粒度土地覆盖类别。与现有数据集不同，HyperImageNet提供原始影像、像素级语义标签和对象级实例掩膜，支持语义分割和实例分割。此外，我们建立了一个具有严格空间分离的开放环境基准，以评估代表性方法和HyperFree基础模型。实验结果表明，HyperImageNet在细粒度高光谱理解和开放环境遥感研究中具有有效性。

View on arXiv Download PDF AI Translation

cs.CV / 39 / 2607.21057

Achieving Text-based Person Retrieval with Any Granularity

实现任意粒度的基于文本的人物检索

Zuo, Jialong, Zhou, Hanyu, Wu, Dongyue, Deng, Yongtai, Tan, Mengdan, Sang, Nong, Gao, Changxin, Bai, Xiang

Abstract

Text-based person retrieval faces a critical but under-explored challenge: the inherent uncertainty of query granularity in real-world scenarios. This paper introduces a new paradigm, Text-based Person Retrieval with Any Granularity, and provides a systematic solution. First, we formalize a five-level granularity spectrum and construct UFine6926-MG, a high-quality multi-grained dataset annotated comprehensively at all granularities via a novel Multi-grained Text Annotation Engine. Second, acknowledging that coarse queries naturally correspond to multiple valid candidates, we propose MG-Eval, a holistic evaluation benchmark with progressively detailed texts and cross-identity labels that reflect real-world semantics, alongside tailored evaluation metrics and protocols. Third, after a comprehensive diagnosis reveals the systemic limitations of existing research, we propose the Cross-modal Multi-grained Aligning and Matching (CMAM) framework. CMAM achieves granularity-aware retrieval through: 1) orthogonal-expert perception to disentangle granularity-specific features; 2) probabilistic alignment to model many-to-many matches under query uncertainty; and 3) granularity-consistent reasoning to steer feature learning via joint cross-modal granularity verification. Experiments demonstrate that CMAM significantly outperforms state-of-the-art methods across all granularity levels. This work establishes a foundational benchmark and a robust baseline, paving the way for more practical person retrieval systems.

Chinese Translation

基于文本的人物检索面临一个关键但未充分探索的挑战：在现实场景中查询粒度的固有不确定性。本文提出了一种新的范式——任意粒度的基于文本的人物检索，并提供了一种系统性的解决方案。首先，我们形式化了一个五级粒度谱，并构建了UFine6926-MG，这是一个高质量的多粒度数据集，通过一种新颖的多粒度文本标注引擎在所有粒度上进行了全面注释。其次，考虑到粗略查询自然对应多个有效候选者，我们提出了MG-Eval，这是一个整体评估基准，具有逐步详细的文本和反映现实语义的跨身份标签，以及量身定制的评估指标和协议。第三，在全面诊断揭示现有研究的系统性局限性后，我们提出了跨模态多粒度对齐与匹配（CMAM）框架。CMAM通过以下方式实现粒度感知检索：1）正交专家感知以解开特定粒度特征；2）概率对齐以建模查询不确定性下的多对多匹配；3）粒度一致推理以通过联合跨模态粒度验证引导特征学习。实验表明，CMAM在所有粒度水平上显著优于最先进的方法。本研究建立了一个基础基准和一个稳健的基线，为更实用的人物检索系统铺平了道路。

View on arXiv Download PDF AI Translation

cs.CV / 40 / 2607.21061

MVEI & EmObserver: Empowering MLLM-Oriented Visual Emotional Intelligence via Emotion Statement Judgement

MVEI与EmObserver：通过情感陈述判断赋能面向多模态大语言模型的视觉情感智能

Wu, Daiqing, Yang, Dongbao, Yao, Jiashu, Zhang, Hongrui, Ma, Can, Zhou, Yu, Zhao, Sicheng

Abstract

Affective Image Content Analysis (AICA) aims to recognize and understand emotions elicited by visual content, representing an indispensable step toward Artificial General Intelligence (AGI). However, despite the rapid progress of Multimodal Large Language Models (MLLMs), systematic evaluation of their visual emotional intelligence remains largely absent from recent model releases. We attribute this gap to a structural mismatch between conventional AICA paradigms and the open-ended, instruction-driven nature of MLLMs, where further analysis reveals four major limitations: omission of plausible responses, limited emotion taxonomies, neglect of contextual factors, and labor-intensive annotation. To overcome these barriers, we introduce Emotion Statement Judgement (ESJ), a statement-verification formulation that preserves the expressiveness of the input space while constraining outputs to discriminative judgements. We further develop INSETS, a labor-efficient pipeline that instantiates ESJ at scale by constructing INSETS-462k and supporting MVEI, a rigorously refined benchmark spanning sentiment polarity, emotion interpretation, scene context, and perception subjectivity. Beyond evaluation, we build EmObserver, an emotion-oriented MLLM optimized on ESJ through an elaborate multi-stage recipe. Extensive evaluation of broad-spectrum MLLMs on MVEI reveals fine-grained insights into current artificial visual emotional intelligence, while experiments on multiple AICA benchmarks demonstrate the accuracy, generalization, and reasoning faithfulness of EmObserver. Collectively, these results establish ESJ as a practical formulation, MVEI as a comprehensive benchmark, and EmObserver as an advanced baseline for advancing MLLM-oriented visual emotional intelligence. Code will be released at: https://github.com/wdqqdw/EmObserver.

Chinese Translation

情感图像内容分析（AICA）旨在识别和理解视觉内容引发的情感，这是实现人工通用智能（AGI）不可或缺的一步。然而，尽管多模态大语言模型（MLLMs）快速发展，近期模型发布中对其视觉情感智能的系统评估仍然基本缺失。我们将这一差距归因于传统AICA范式与MLLMs开放式、指令驱动特性之间的结构不匹配，进一步分析揭示了四个主要局限性：可行响应的遗漏、情感分类的局限、对上下文因素的忽视以及劳动密集型标注。为克服这些障碍，我们引入了情感陈述判断（ESJ），这是一种保留输入空间表现力的陈述验证形式，同时将输出限制为区分性判断。我们进一步开发了INSETS，一个高效的工作流程，通过构建INSETS-462k来大规模实现ESJ，并支持MVEI，这是一个严格精炼的基准，涵盖情感极性、情感解释、场景上下文和感知主观性。除了评估，我们还构建了EmObserver，这是一个基于ESJ优化的情感导向MLLM，采用复杂的多阶段方案进行训练。对广谱MLLM在MVEI上的广泛评估揭示了当前人工视觉情感智能的细致见解，而在多个AICA基准上的实验则展示了EmObserver的准确性、泛化能力和推理可信度。总体而言，这些结果确立了ESJ作为一种实用的形式，MVEI作为一个全面的基准，以及EmObserver作为推动面向MLLM的视觉情感智能的先进基线。代码将发布于：https://github.com/wdqqdw/EmObserver。

View on arXiv Download PDF AI Translation

cs.CV / 41 / 2607.21065

Do Pathology Vision-Language Models Truly See Pathology?

病理视觉语言模型真的能看懂病理吗？

Zhang, Chengyang, Zhang, Wenchuan, Li, Bo, Liu, Xinyu, Yang, Jiaming, Li, Mengran, Deng, Chenxun, Chen, Jie, Zhang, Yang, Ju, Wei, Yi, Yuhao, Bu, Hong, Lv, Jiancheng

Abstract

Pathology vision-language models (VLMs) have recently progressed rapidly and are commonly evaluated by answer accuracy on pathology VQA benchmarks. However, we dig into current evaluations and identify three overlooked issues: 1) Visual evidence is not always necessary. For instance, Gemini-3-Pro achieves 53.5% average accuracy across 5 VQA benchmarks without any visual input. 2) Domain training can improve accuracy without proportional gains in visual binding. Compared with Qwen2.5-VL-7B, Patho-R1-7B exhibits a 5.8-point lower multimodal gain and a 3.7-point lower attention IoU. 3) Entity-level attention is diffuse and weakly query-specific. On PathVG, attention maps remain highly correlated across different entity queries. These issues can lead to substantial misjudgments of pathology VLMs' actual multimodal capabilities. To this end, we present PathBind, a benchmark comprising 2,600 samples: PathBind-VQA with 1,500 questions across six dimensions, PathBind-PTA with 600 questions from a private pathology teaching atlas, and PathBind-Grounding with 500 expert-curated region-level samples. Each component undergoes task-specific automated filtering and expert review to reduce textual shortcuts and improve entity-region correspondence. We evaluate 18 representative VLMs on VQA samples of PathBind and five existing pathology VQA benchmarks, and further evaluate 10 VLMs on PathBind-Grounding and PathVG. Results show that current pathology VLMs still exhibit a substantial gap between answer-side performance and visual-semantic binding.

Chinese Translation

病理视觉语言模型（VLMs）近年来快速发展，通常通过在病理视觉问答（VQA）基准上的回答准确性进行评估。然而，我们深入分析了当前的评估，发现了三个被忽视的问题：1）视觉证据并不总是必要。例如，Gemini-3-Pro在没有任何视觉输入的情况下，在5个VQA基准上实现了53.5%的平均准确率。2）领域训练可以提高准确性，但并未带来视觉绑定的成比例提升。与Qwen2.5-VL-7B相比，Patho-R1-7B的多模态增益低了5.8个百分点，注意力IoU低了3.7个百分点。3）实体级注意力分散且与查询的特异性较弱。在PathVG上，不同实体查询的注意力图高度相关。这些问题可能导致对病理VLMs实际多模态能力的重大误判。为此，我们提出了PathBind，一个包含2600个样本的基准：PathBind-VQA包含跨六个维度的1500个问题，PathBind-PTA包含来自私有病理教学图谱的600个问题，PathBind-Grounding包含500个专家策划的区域级样本。每个组件都经过任务特定的自动过滤和专家审查，以减少文本捷径并改善实体与区域的对应关系。我们在PathBind的VQA样本和五个现有病理VQA基准上评估了18个具有代表性的VLMs，并进一步在PathBind-Grounding和PathVG上评估了10个VLMs。结果显示，当前的病理VLMs在回答性能与视觉语义绑定之间仍存在显著差距。

View on arXiv Download PDF AI Translation

cs.CV / 42 / 2607.21071

TransBiolab: A Real-World Multi-View Dataset of Cluttered Transparent Biomedical Objects

TransBiolab：一个真实世界的多视角杂乱透明生物医学物体数据集

Ma, Ke, Wang, Yifei, Wang, Meng, Xia, Tian

Abstract

Autonomous biomedical laboratories increasingly rely on visual perception to recognize, localize, and manipulate transparent plasticware, yet high-quality real-world datasets for this setting remain limited. The scarcity of domain-relevant data is particularly restrictive in cluttered multi-object scenes, where mutual occlusion and view-dependent appearance changes remain challenging even for contemporary visual foundation models. Existing transparent-object datasets have advanced segmentation, depth, and pose estimation, but they usually do not evaluate the combined setting of multi-object clutter, occlusion, and calibrated multi-view capture that characterizes real laboratory manipulation scenes. To address this gap, we present TrainsBiolab, a real-world RGB-D dataset of cluttered transparent biomedical objects captured as calibrated multi-view sequences. TrainsBiolab contains 161,315 frames from 98 scenes and 1.03M instance annotations over 15 laboratory object types, including 6D poses, full and visible masks, depth, and per-frame camera calibration. The dataset is organized along three axes that reflect operational difficulty: object category, the total number of objects in a frame, and camera viewpoint. We further define dataset-centric benchmarks for segmentation, depth estimation and completion, and 6D pose estimation, and report a system-level robot manipulation evaluation enabled by the released annotations and calibrations. By focusing on repeated transparent instances, clutter, and multi-view laboratory capture, TrainsBiolab provides a resource for segmentation, depth estimation, 6D pose estimation, and multi-view reasoning in autonomous laboratory manipulation. Project page: https://dualtransparency.github.io/TransBiolab/.

Chinese Translation

自主生物医学实验室越来越依赖视觉感知来识别、定位和操作透明塑料器皿，但针对这一环境的高质量真实世界数据集仍然有限。在杂乱的多物体场景中，领域相关数据的稀缺尤其限制了研究进展，因为相互遮挡和视角依赖的外观变化即使对于当代视觉基础模型来说也仍然具有挑战性。现有的透明物体数据集在分割、深度和姿态估计方面取得了一定进展，但通常未能评估多物体杂乱、遮挡和标定多视角捕获的组合设置，这正是现实实验室操作场景的特征。为了解决这一空白，我们提出了TrainsBiolab，这是一个真实世界的RGB-D数据集，包含杂乱的透明生物医学物体，作为标定的多视角序列进行捕获。TrainsBiolab包含来自98个场景的161,315帧和超过1.03M的实例注释，涵盖15种实验室物体类型，包括6D姿态、完整和可见的掩膜、深度以及逐帧相机标定。该数据集沿着反映操作难度的三个轴线进行组织：物体类别、帧中的物体总数和相机视角。我们进一步定义了以数据集为中心的基准，用于分割、深度估计和补全，以及6D姿态估计，并报告了通过发布的注释和标定实现的系统级机器人操作评估。通过关注重复的透明实例、杂乱和多视角实验室捕获，TrainsBiolab为自主实验室操作中的分割、深度估计、6D姿态估计和多视角推理提供了资源。项目页面：https://dualtransparency.github.io/TransBiolab/

View on arXiv Download PDF AI Translation

cs.CV / 43 / 2607.21072

Show, Don't Tell: Evaluating Spatial Cognition in Generative Pixels Rather Than LLM Text

展示，而非叙述：评估生成像素中的空间认知而非LLM文本

Wang, Xu, Yao, Kaixiang, Pan, Miao, Zhou, Xiaohe, Liu, Xuanyu, Zhang, Wenqi, Zhang, Xuhong

Abstract

Spatial intelligence is essential for agents to move from static semantic understanding toward interacting with the physical world. Many spatial tasks are grounded in continuous visual scenes, where locations, regions, and paths are more naturally expressed by pointing, marking, or drawing than by reporting precise coordinates or discrete textual symbols. Yet existing spatial reasoning benchmarks usually require coordinates, options, or text, creating an answer-interface mismatch for image-generation models. This makes it difficult to evaluate image-generation models under the same task semantics as text-output VLMs, despite their ability to externalize spatial judgments directly in pixel space. We propose ProVisE (Protocolized Visual Evaluation), a benchmark-agnostic framework that elicits protocol-constrained visual answers from image-generation models and parses them into structured predictions compatible with original metrics. ProVisE also includes an Agentic builder that constructs and validates task-specific protocols for new benchmarks. We further introduce SpatialGen-Bench, a curated diagnostic benchmark of 470 samples across 14 spatial subtasks, four capability levels, and diverse answer forms. We evaluate representative text-output VLMs and image-generation models in a unified setting and validate Agentic protocol construction on six external spatial benchmarks. Results show that image-generation models are competitive when spatial answers can be externalized directly in pixel space, while text-output VLMs retain a clear advantage in compositional spatial reasoning. These findings reveal complementary strengths of pixel-space expression and text-based reasoning and establish a metric-compatible testbed for studying spatial cognition in image-generation models.

Chinese Translation

空间智能对于代理从静态语义理解转向与物理世界互动至关重要。许多空间任务基于连续的视觉场景，其中位置、区域和路径通过指向、标记或绘制的方式比通过报告精确坐标或离散文本符号更自然。然而，现有的空间推理基准通常需要坐标、选项或文本，这导致图像生成模型与答案接口之间的不匹配。这使得在与文本输出的视觉语言模型（VLMs）相同的任务语义下评估图像生成模型变得困难，尽管它们能够直接在像素空间中外化空间判断。我们提出了ProVisE（协议化视觉评估），这是一个基准无关的框架，能够从图像生成模型中引出协议约束的视觉答案，并将其解析为与原始指标兼容的结构化预测。ProVisE还包括一个代理构建器，用于构建和验证新基准的任务特定协议。我们进一步介绍了SpatialGen-Bench，这是一个经过精心策划的诊断基准，包含470个样本，涵盖14个空间子任务、四个能力水平和多种答案形式。我们在统一设置中评估了代表性的文本输出VLMs和图像生成模型，并在六个外部空间基准上验证了代理协议的构建。结果表明，当空间答案可以直接在像素空间中外化时，图像生成模型具有竞争力，而文本输出的VLMs在组合空间推理方面仍然保持明显优势。这些发现揭示了像素空间表达和基于文本推理的互补优势，并建立了一个与指标兼容的测试平台，用于研究图像生成模型中的空间认知。

View on arXiv Download PDF AI Translation

cs.CV / 44 / 2607.21076

C-PTQ: Fisher-weighted Channel-wise Sensitivity for Post-training Quantization of MLLMs

C-PTQ：用于后训练量化的费舍尔加权通道敏感性在多模态大语言模型中的应用

Li, Jiameng, Zhou, Han, Blaschko, Matthew B.

Abstract

Multimodal large language models (MLLMs) require huge memory and computational costs, which limits their practical deployment. Post-training quantization (PTQ) techniques offer an efficient solution for model compression and inference acceleration. Yet, the quantized model faces performance degradation due to outlier channels, which are highly sensitive to quantization and substantially impair activation fidelity and task accuracy. To protect these salient channels during quantization, existing PTQ methods leverage modality- or token-level metrics to guide channel-wise scaling (CWS) of LLM decoders. However, these orthogonal measurements fail to capture channel-wise impacts on task-specific loss, and the misalignment between importance and scaling factors ultimately leads to suboptimal performance. To address this issue, we propose C-PTQ, a unified channel-wise PTQ method that harmonizes task-specific loss perturbation and quantization error. Motivated by second-order derivatives, we design a Fisher-weighted objective as a tractable Hessian approximation, seamlessly injecting task sensitivity into the scaling process. Notably, we achieve state-of-the-art performance without auxiliary modules like LoRA, thereby maintaining high efficiency. Experiments on Qwen2.5VL, InternVL2 and LLaVA-OV across 8 benchmarks demonstrate our effectiveness in both weight-only and weight-activation settings.

Chinese Translation

多模态大语言模型（MLLMs）需要巨大的内存和计算成本，这限制了它们的实际部署。后训练量化（PTQ）技术为模型压缩和推理加速提供了高效的解决方案。然而，由于异常通道对量化高度敏感，量化模型面临性能下降，这显著影响了激活保真度和任务准确性。为了在量化过程中保护这些显著通道，现有的PTQ方法利用模态或标记级别的指标来指导大语言模型解码器的通道级缩放（CWS）。然而，这些正交测量未能捕捉通道对任务特定损失的影响，重要性与缩放因子之间的不匹配最终导致次优性能。为了解决这个问题，我们提出了C-PTQ，一种统一的通道级PTQ方法，协调任务特定损失扰动和量化误差。受到二阶导数的启发，我们设计了一个费舍尔加权目标作为可处理的海森矩阵近似，顺利地将任务敏感性注入缩放过程。值得注意的是，我们在不使用像LoRA这样的辅助模块的情况下实现了最先进的性能，从而保持了高效率。在Qwen2.5VL、InternVL2和LLaVA-OV上的8个基准实验中证明了我们在仅权重和权重-激活设置下的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 45 / 2607.21078

The RealDefocus Benchmark for Defocus Deblurring

用于失焦去模糊的 RealDefocus 基准

Seizinger, Tim, Zhou, Zhuyun, Timofte, Radu

Abstract

Single-Image Defocus Deblurring (SIDD) aims to recover an all-in-focus image from a single defocused observation, but rigorous and reproducible evaluation remains challenging due to the scarcity of realistic, high-resolution datasets with well-aligned defocused/sharp pairs and standardized protocols. We build on RealDefocus, a benchmark derived from the real-world RealBokeh dataset originally proposed for Bokeh Rendering. RealDefocus provides paired defocused inputs and sharp ground truth images, predefined training/validation/test splits, and a unified evaluation framework for comparing image restoration and neural rendering approaches. We further outline a benchmarking protocol with cross-dataset validation to assess reconstruction quality and generalization. The project page is publicly available at: www.github.com/TimSeizinger/RealDefocus-Benchmark.

Chinese Translation

单幅图像失焦去模糊（Single-Image Defocus Deblurring, SIDD）旨在从单个失焦观察中恢复出一幅全聚焦图像，但由于缺乏真实、高分辨率的数据集，这些数据集具有良对齐的失焦/清晰图像对和标准化的评估协议，因此严格且可重复的评估仍然具有挑战性。我们基于 RealDefocus，这是一项源于真实世界的 RealBokeh 数据集的基准，最初用于虚化渲染。RealDefocus 提供成对的失焦输入和清晰的真实图像，预定义的训练/验证/测试划分，以及一个统一的评估框架，用于比较图像恢复和神经渲染方法。我们进一步概述了一种基准测试协议，通过跨数据集验证来评估重建质量和泛化能力。项目页面可公开访问： www.github.com/TimSeizinger/RealDefocus-Benchmark。

View on arXiv Download PDF AI Translation

cs.CV / 46 / 2607.21085

Geo3R: Mitigating Spatial Reasoning Hallucination in Multimodal Large Language Models

Geo3R：缓解多模态大型语言模型中的空间推理幻觉

Wang, Mingyu, Jin, Weilin, Li, Wenbo, Huang, Haoyang, Jia, Tong, Li, Ying

Abstract

Despite remarkable progress in visual understanding, Multimodal Large Language Models (MLLMs) remain prone to hallucinations when reasoning about spatial relationships, often producing judgments that contradict the true 3D structure of the scene. Though several existing works have proposed to mitigate hallucinations, our analysis indicates that they show limited effectiveness in spatial reasoning, as they fail to bridge the fundamental gap between 2D visual representations and 3D spatial reality. Based on this finding, we define hallucinations arising from insufficient spatial structure modeling as spatial reasoning hallucination, a subcategory of relation hallucination that existing mitigation methods fail to address. We further identify three typical scenarios where such hallucinations frequently occur: perspective effects, object orientation, and viewpoint changes. To this end, we propose Geo3R, a training-free, plug-and-play framework that incorporates geometric evidence and structured 3D reasoning to mitigate spatial reasoning hallucination. Experiments on three benchmarks, covering 18 tasks across all three scenarios, show that Geo3R substantially reduces spatial reasoning hallucination across diverse MLLMs without additional training, outperforming existing models and methods.

Chinese Translation

尽管在视觉理解方面取得了显著进展，多模态大型语言模型（MLLMs）在推理空间关系时仍然容易出现幻觉，常常产生与场景真实三维结构相矛盾的判断。尽管已有若干研究提出了缓解幻觉的方法，但我们的分析表明，它们在空间推理方面的有效性有限，因为它们未能弥合二维视觉表示与三维空间现实之间的根本差距。基于这一发现，我们将由于空间结构建模不足而产生的幻觉定义为空间推理幻觉，这是一种现有缓解方法未能解决的关系幻觉的子类别。我们进一步识别出三种典型场景，在这些场景中此类幻觉经常发生：透视效应、物体方向和视角变化。为此，我们提出了Geo3R，一个无训练、即插即用的框架，结合几何证据和结构化三维推理，以缓解空间推理幻觉。在三个基准测试上的实验，涵盖了所有三种场景的18个任务，显示Geo3R在不进行额外训练的情况下显著减少了多样化MLLMs中的空间推理幻觉，且优于现有模型和方法。

View on arXiv Download PDF AI Translation

cs.CV / 47 / 2607.21089

Loss Landscape Topology Reveals Why Simple Baselines are Competitive at 3D Point Cloud Segmentation Under Class Imbalance

损失景观拓扑揭示了为何简单基线在类不平衡的3D点云分割中具有竞争力

Savva, Antonis, Kyrkou, Christos, Theocharides, Theocharis

Abstract

Semantic segmentation of 3D point clouds faces severe class imbalance, yet the effectiveness of specialized imbalance-aware methods from 2D computer vision remains unclear in 3D contexts. We systematically evaluate 11 imbalance mitigation approaches across datasets with extreme (641:1) and moderate (56:1) imbalance ratios, revealing a surprising finding: standard cross-entropy with uniform weighting achieves competitive performance, typically within 0.8-3.3% mIoU of specialized methods across architectures and datasets. Through multifaceted mechanistic analysis of error patterns, decision boundaries, and the geometry of the optimization landscape, our analyses suggest that imbalance severity shapes the topology, creating narrow solution basins under extreme imbalance and flat plateaus under moderate imbalance. This appears to constrain the effectiveness of loss-level modifications, as all methods must navigate these geometric constraints. Our findings offer practical guidance; standard cross-entropy provides a robust baseline, with specialized methods offering modest improvements (0.8-3.3% mIoU) that vary by architecture and dataset but risk substantial degradation if poorly tuned. This work provides the first mechanistic explanation for why techniques proven effective in 2D do not readily transfer to point-based 3D point cloud segmentation, validated across two representative architectures.

Chinese Translation

3D点云的语义分割面临严重的类不平衡问题，但来自2D计算机视觉的专门不平衡感知方法在3D环境中的有效性仍不明确。我们系统地评估了11种不平衡缓解方法在极端（641:1）和适度（56:1）不平衡比率的数据集上的表现，揭示了一个令人惊讶的发现：标准交叉熵（cross-entropy）结合均匀加权实现了具有竞争力的性能，通常在不同架构和数据集上与专门方法的mIoU相差0.8-3.3%。通过对错误模式、决策边界和优化景观几何的多方面机制分析，我们的研究表明，不平衡的严重程度塑造了拓扑结构，在极端不平衡下形成狭窄的解盆地，而在适度不平衡下则形成平坦的高原。这似乎限制了损失级别修改的有效性，因为所有方法都必须在这些几何约束中进行导航。我们的发现提供了实用的指导；标准交叉熵提供了一个稳健的基线，而专门方法提供的改善（0.8-3.3% mIoU）因架构和数据集而异，但如果调整不当则可能导致显著的性能下降。这项工作首次提供了机制性解释，说明为何在2D中被证明有效的技术未能轻易转移到基于点的3D点云分割，并在两个代表性架构中得到了验证。

View on arXiv Download PDF AI Translation

cs.CV / 48 / 2607.21105

HalluScope: Fine-grained Hallucination Diagnosis for Multimodal Large Language Models

HalluScope：多模态大语言模型的细粒度幻觉诊断

Jin, Weilin, Wang, Mingyu, Li, Wenbo, Huang, Haoyang, Wu, Yifan, Li, Ying, Huang, Gang, Wu, Zhonghai

Abstract

Although Multimodal Large Language Models have achieved strong performance across a wide range of vision-language tasks, they still suffer from hallucinations, where model outputs become inconsistent with the visual content, textual context, or commonsense knowledge. Existing studies primarily address this problem through coarse-grained detection. However, these approaches often provide insufficient diagnostic information for understanding hallucination types and supporting downstream hallucination mitigation. To bridge this gap, we propose fine-grained hallucination diagnosis for MLLMs, a new unified task that jointly performs hallucination detection, classification, and interpretable explanation generation. We develop an automated data generation pipeline and construct HalluScope-30K, a large-scale diagnostic dataset covering eight sources and five task categories. Based on this dataset, we design a multi-granular joint reward function and train two diagnosis models, HalluScope-4B and HalluScope-8B, which achieve state-of-the-art performance on both the MHALO benchmark and our fine-grained hallucination classification benchmark. Notably, detection and classification are mutually beneficial under joint optimization. Furthermore, diagnosis-driven feedback experiments show that the fine-grained diagnostic explanations produced by our model effectively guide target models to correct their hallucinations, with full diagnosis substantially outperforming all baselines on both Qwen3-VL-8B-Instruct and LLaVA-1.5-7B.

Chinese Translation

尽管多模态大语言模型在广泛的视觉-语言任务中取得了强劲的表现，但它们仍然面临幻觉问题，即模型输出与视觉内容、文本上下文或常识知识不一致。现有研究主要通过粗粒度检测来解决这一问题。然而，这些方法往往提供不足的诊断信息，无法有效理解幻觉类型并支持下游的幻觉缓解。为填补这一空白，我们提出了针对多模态大语言模型的细粒度幻觉诊断，这是一项新的统一任务，能够联合执行幻觉检测、分类和可解释的解释生成。我们开发了一个自动化数据生成管道，并构建了HalluScope-30K，这是一个覆盖八个来源和五个任务类别的大规模诊断数据集。基于该数据集，我们设计了一个多粒度联合奖励函数，并训练了两个诊断模型HalluScope-4B和HalluScope-8B，这两个模型在MHALO基准和我们的细粒度幻觉分类基准上均实现了最先进的性能。值得注意的是，在联合优化下，检测与分类是相辅相成的。此外，基于诊断驱动的反馈实验表明，我们模型生成的细粒度诊断解释有效地指导目标模型纠正其幻觉，全面诊断在Qwen3-VL-8B-Instruct和LLaVA-1.5-7B上显著优于所有基线。

View on arXiv Download PDF AI Translation

cs.CV / 49 / 2607.21118

The Second LoViF 2026 Challenge on Real-World All-in-One Image Restoration: Methods and Results

2026年第二届LoViF挑战赛：真实世界一体化图像恢复的方法与结果

Chen, Xiang, Li, Hao, Dong, Jiangxin, Pan, Jinshan, Li, Xin, Ding, Hongbo, Jiang, Junpeng, Qiu, Xingyu, Zhong, Yilian, Chen, Yuxiang, Yin, Shibo, Huang, Zixuan, Fang, Yushun, Zhu, Xilei, Wang, Yahui, Lu, Chen, Zhou, Xiaodong, Cao, Qingyue, Gong, Changwei, Liu, Jingyun, Yi, Xingchen, Shi, Hansen, Liu, Ruiyi, Xie, Jirui, Liu, Tao, Ma, Wenzhuo, Li, Hongzhen, Chen, Yongyong, Zhou, Zheng, Su, Jingyong, Liu, Jie, Zeng, Haijin, Li, Cheng, Zha, Peishuai, Wang, Ziyi, Tang, Jian, Chen, Yan, Bao, Long, Sun, Heng, Zhang, Jiyuan, Liu, Shuai, Ding, Wei, Guo, Chengjun, Huang, Yibin, Wang, Xiaotao, Zou, Dongqing, Lei, Lei, Wang, Xiaofeng, Liu, Xiao, Wu, Yulin, Zhao, Yuhan, Peng, Shurui, Ren, Chao, Wang, Yu-Kai, Shigematsu, Kosuke, Shin, Asuka, Jian, Rong-Lin, Kang, Cheng-Jun, Jiang, Jin-Hui, Zhou, Jialin, Yuan, Kuo, Zhang, Songyu, Benson, E B, Hussain, Ashfaq, AC, Pruthvikanth, Chen, Qirui, Chen, Jinyuan, Zhang, Jun, Zhang, Xu, Cao, Xuhui, Ma, Jiaqi, Chang, Laibin, Miao, Yuchun, Xu, Yichu, Yao, Yuanzhi, Chen, Shi, Cui, Yuning, Zhang, Huan, Zhang, Lefei, Ahmad, Saeed, Lee, Ik Hyun, Park, Jun Young, Yoon, Ji Hwan, Sun, Shangquan, Nobahar-Moghanlou, Behrooz, Edalatjou, Majid, Shahi-Niyar, Karim, Zhang, Ruibo, Hong, Dexiang, Liu, Xinyan, Tang, Shengeng, Chen, Weidong

Abstract

This paper presents a review of the second LoViF Challenge on Real-World All-in-One Image Restoration. The challenge aims to advance unified image restoration under diverse real-world degradation conditions, including blur, low-light, haze, rain, and snow. It provides a common benchmark for evaluating the restoration accuracy, robustness, and generalization capability of models across multiple degradation categories within a unified framework. The competition attracted 158 registered participants, and 20 teams were included in the final ranking after their submitted results were successfully reproduced and verified. This report provides a comprehensive analysis of the submitted solutions and corresponding results, highlighting recent advances in real-world all-in-one image restoration. The summarized methods and empirical findings reveal effective design strategies and establish an updated benchmark for future research in real-world low-level vision.

Chinese Translation

本文回顾了第二届LoViF挑战赛，主题为真实世界一体化图像恢复。该挑战旨在推动在多种真实世界退化条件下的统一图像恢复，包括模糊、低光照、雾霾、雨水和雪等。它为评估模型在多个退化类别下的恢复准确性、鲁棒性和泛化能力提供了一个共同的基准。此次比赛吸引了158名注册参与者，最终有20个团队在其提交的结果成功复现和验证后进入最终排名。本报告对提交的解决方案及其对应结果进行了全面分析，突出了真实世界一体化图像恢复的最新进展。总结的方法和实证发现揭示了有效的设计策略，并为未来在真实世界低级视觉领域的研究建立了更新的基准。

View on arXiv Download PDF AI Translation

cs.CV / 50 / 2607.21125

Causal-AgentIR: Self-Evolving Causal Memory for Adaptive Image Restoration Agents

因果代理图像恢复：自我演化的因果记忆用于自适应图像恢复代理

Gao, Hu, Chen, Yulong, Ma, Lizhuang

Abstract

Image restoration agents have recently emerged as a flexible paradigm for handling diverse and unpredictable degradations in real-world scenarios. Existing agents typically formulate restoration as a tool-using process, where the agent perceives degradations, searches candidate tools, executes restoration operations, and revises the plan through reflection or rollback. However, their knowledge is often stored as static tool descriptions, manually defined degradation priors, or unstructured textual summaries, which limits the accumulation, verification, revision, and forgetting of restoration knowledge over long-term experience. In this paper, we propose Causal-AgentIR, a hierarchical multi-agent framework with self-evolving causal memory for collective image restoration intelligence. Instead of representing restoration experience as isolated textual records, Causal-AgentIR organizes degradation patterns, image regions, restoration tools, actions, quality changes, and user preferences into a structured causal memory graph. This graph supports graph-based retrieval and multi-hop causal reasoning, enabling agents to infer how specific restoration operations or tool sequences affect restoration quality under different degradation conditions. The framework further organizes multiple agents into a collaborative system, including planning, degradation analysis, tool expertise, causal memory reasoning, outcome critique, and memory curation. Through this design, restoration experience can be added, updated, merged, reinforced, ignored, or discarded according to observed quality changes and feedback, allowing the agent to maintain reliable and transferable restoration knowledge. Extensive experiments demonstrate the effectiveness of the proposed framework.

Chinese Translation

图像恢复代理最近作为一种灵活的范式出现，用于处理现实场景中多样且不可预测的退化。现有的代理通常将恢复过程视为一种工具使用过程，其中代理感知退化，搜索候选工具，执行恢复操作，并通过反思或回滚修订计划。然而，它们的知识通常以静态工具描述、手动定义的退化先验或非结构化文本摘要的形式存储，这限制了恢复知识在长期经验中的积累、验证、修订和遗忘。在本文中，我们提出了因果代理图像恢复（Causal-AgentIR），这是一个具有自我演化因果记忆的分层多代理框架，用于集体图像恢复智能。Causal-AgentIR并不是将恢复经验表示为孤立的文本记录，而是将退化模式、图像区域、恢复工具、操作、质量变化和用户偏好组织成一个结构化的因果记忆图。该图支持基于图的检索和多跳因果推理，使代理能够推断特定恢复操作或工具序列在不同退化条件下如何影响恢复质量。该框架进一步将多个代理组织成一个协作系统，包括规划、退化分析、工具专长、因果记忆推理、结果评价和记忆管理。通过这种设计，恢复经验可以根据观察到的质量变化和反馈进行添加、更新、合并、强化、忽略或丢弃，从而使代理能够保持可靠且可转移的恢复知识。大量实验表明了所提框架的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 51 / 2607.21137

Safety-oriented sidewalk and road segmentation for smartphone-based assistive navigation

面向安全的手机辅助导航人行道与道路分割

Calim, Hakan, Dumitrescu, Anamaria, Panambur, Adarsh Bhandary, Asif, Huzaifa, Maier, Andreas

Abstract

Independent sidewalk mobility is essential for blind and visually impaired pedestrians (BVIPs), yet smartphone-based assistive navigation requires perception models that distinguish walkable sidewalks from adjacent unsafe regions. This study presents a safety-oriented semantic segmentation framework for future mobile guidance. We introduce SENSATION-DS, a chest-height pedestrian-view dataset with 2,752 image-mask pairs and nine-class navigation-relevant taxonomy. External urban and sidewalk datasets were harmonized to this label space, and five segmentation architectures were evaluated using staged target-domain adaptation with mask-conditioned synthetic images and Segment Anything Model 2 (SAM2) pseudo-labels. Models were assessed using mean Intersection over Union (mIoU), road- and sidewalk-specific metrics, Road-as-Sidewalk Error Rate as a proxy false-safe measure, and Android Open Neural Network Exchange benchmarking. Synthetic augmentation generally improved segmentation accuracy, whereas SAM2 pseudo-labels more consistently reduced Road-as-Sidewalk errors. UPerNet-MobileNetV3 achieved the highest offline mIoU (0.715 +/- 0.006), while DeepLabV3Plus-MobileNetV3 achieved the lowest Road-as-Sidewalk Error Rate (0.079) and highest Android runtime at 512x384 (7.383 FPS). These results show that assistive sidewalk perception should be evaluated jointly by segmentation accuracy, proxy false-safe behavior, and smartphone deployment feasibility, while real-world benefit requires validation with BVIP users. This evaluation supports selecting models that balance accurate perception, conservative error behavior, and practical runtime.

Chinese Translation

独立的人行道移动对于盲人和视力障碍行人（BVIPs）至关重要，但基于智能手机的辅助导航需要感知模型来区分可行走的人行道与相邻的不安全区域。本研究提出了一种面向安全的语义分割框架，以支持未来的移动导航。我们引入了SENSATION-DS，这是一个胸部高度的行人视角数据集，包含2,752对图像-掩码和九类与导航相关的分类。外部城市和人行道数据集被统一到这一标签空间，并评估了五种分割架构，采用了分阶段的目标领域适应，结合掩码条件的合成图像和Segment Anything Model 2 (SAM2) 的伪标签。模型的评估使用了平均交并比（mIoU）、针对道路和人行道的特定指标、作为代理的道路-人行道错误率以及Android开放神经网络交换基准测试。合成增强通常提高了分割准确性，而SAM2伪标签则更一致地减少了道路-人行道错误。UPerNet-MobileNetV3达到了最高的离线mIoU（0.715 +/- 0.006），而DeepLabV3Plus-MobileNetV3则达到了最低的道路-人行道错误率（0.079）和在512x384分辨率下的最高Android运行时（7.383 FPS）。这些结果表明，辅助人行道感知应通过分割准确性、代理的错误安全行为和智能手机部署的可行性进行联合评估，而实际的现实世界效益需要与BVIP用户进行验证。这一评估支持选择在准确感知、保守错误行为和实际运行时之间取得平衡的模型。

View on arXiv Download PDF AI Translation

cs.CV / 52 / 2607.21138

DTIF: Robust Loop Closure Detection via Delaunay Triangle Topology in Complex Forests

DTIF：基于德劳内三角形拓扑的复杂森林中鲁棒的回环闭合检测

Zhao, Xin, Li, Jianping, Zou, Qin, Liang, Fuxun, Dong, Zhen, Yang, Bisheng

Abstract

Accurate forest inventory and large-scale mapping are essential for ecosystem monitoring and sustainable forest management. Multiple low-cost edge platforms enable efficient large-area data acquisition, but merging independently constructed local maps in GNSS-denied understory environments still requires initialization-free loop closure detection and global registration. This task is challenging because low-cost LiDAR point clouds are sparse and noisy, while repetitive trunk layouts and the lack of distinctive geometric landmarks lead to severe perceptual aliasing and false correspondences. To address these issues, we propose DTIF (Delaunay Triangulation in Forests), a lightweight trunk-topology-based framework for forest loop closure detection and global registration. Tree trunks are first extracted as stable landmarks and encoded using a Delaunay topology for compact scene representation. Candidate submaps are then screened using edge-length and radius statistics, followed by edge--radius consistency verification and strong/weak vertex support aggregation to construct weighted vertex correspondences. Finally, topology-derived reliability weights are incorporated into a decoupled robust pose estimator that separately estimates yaw, horizontal translation, and elevation translation under gravity alignment. Experiments on simulated and real-world forest datasets demonstrate that DTIF achieves accurate registration with low computational overhead, providing a favorable balance among robustness, efficiency, and deployability on resource-constrained edge platforms.

Chinese Translation

准确的森林清查和大规模制图对于生态系统监测和可持续森林管理至关重要。多种低成本边缘平台能够高效地进行大面积数据采集，但在缺乏全球导航卫星系统（GNSS）的下层环境中，合并独立构建的局部地图仍然需要无初始化的回环闭合检测和全局配准。这项任务具有挑战性，因为低成本激光雷达（LiDAR）点云稀疏且噪声较大，而重复的树干布局和缺乏显著几何地标导致严重的感知混淆和错误对应。为了解决这些问题，我们提出了DTIF（森林中的德劳内三角剖分），这是一个基于树干拓扑的轻量级框架，用于森林回环闭合检测和全局配准。首先提取树干作为稳定的地标，并使用德劳内拓扑进行编码，以实现紧凑的场景表示。然后，使用边长和半径统计筛选候选子地图，接着进行边-半径一致性验证和强/弱顶点支持聚合，以构建加权顶点对应关系。最后，将拓扑派生的可靠性权重纳入一个解耦的鲁棒姿态估计器，该估计器在重力对齐下分别估计偏航、水平平移和垂直平移。在模拟和真实世界森林数据集上的实验表明，DTIF以低计算开销实现了准确的配准，在鲁棒性、效率和在资源受限的边缘平台上的可部署性之间提供了良好的平衡。

View on arXiv Download PDF AI Translation

cs.CV / 53 / 2607.21155

CRAG-MM-Diagnostics: Enabling Stage-Wise Analysis of Knowledge-Intensive VQA

CRAG-MM-诊断：实现知识密集型视觉问答的阶段性分析

Oh, Hanseok, BehnamGhader, Parishad, Krojer, Benno, Lee, Hyunji, Liang, Paul, Reddy, Siva, Dankers, Verna

Abstract

Knowledge-Intensive Visual Question Answering (KI-VQA) benchmarks evaluate Vision-Language Models (VLMs) as multimodal knowledge assistants by requiring external information beyond a provided image to answer questions. KI-VQA involves multiple sub-problems -referring expression understanding, visual grounding, object recognition, knowledge retrieval, and reasoning-yet existing benchmarks typically report only end-task accuracy, obscuring where failures arise. To analyze the full KI-VQA pipeline, we introduce CRAG-MM-Diagnostics, a diagnostic benchmark with stage-wise data annotations that isolate 1) language-based visual grounding, 2) object identification, and 3) knowledge retrieval and reasoning. We evaluate fully parametric and retrieval-augmented VLMs, providing fine-grained analyses using newly collected metadata, such as target ROIs, entity names, and visual complexity scores. Our results point to knowledge retrieval and reasoning as the primary bottleneck, but also highlight issues in the other parts of the KI-VQA pipeline, such as the fact that VLMs struggle with target object identification or that image retrievers struggle to integrate textual cues. These findings expose fundamental limitations in current KI-VQA systems and motivate stage-aware evaluation. We, lastly, leverage these findings to propose a grounded bimodal RAG pipeline that integrates a visual grounding module to crop targets before image retrieval, boosting GPT-5 and Qwen's respective accuracies by 13.3 and 8.5 percentage points.

Chinese Translation

知识密集型视觉问答（KI-VQA）基准测试通过要求超出提供图像的外部信息来回答问题，从而评估视觉语言模型（VLMs）作为多模态知识助手的能力。KI-VQA 涉及多个子问题——指称表达理解、视觉定位、物体识别、知识检索和推理——但现有基准通常仅报告最终任务的准确性，掩盖了失败的来源。为了分析整个 KI-VQA 流程，我们引入了 CRAG-MM-诊断，这是一个具有阶段性数据注释的诊断基准，能够孤立出 1）基于语言的视觉定位，2）物体识别，以及 3）知识检索和推理。我们评估了全参数和检索增强的 VLMs，利用新收集的元数据（如目标 ROI、实体名称和视觉复杂度评分）提供细粒度分析。我们的结果表明，知识检索和推理是主要瓶颈，但也突显了 KI-VQA 流程其他部分的问题，例如 VLMs 在目标物体识别方面的困难，或图像检索器在整合文本线索方面的挑战。这些发现揭示了当前 KI-VQA 系统的基本局限性，并推动了阶段感知评估。最后，我们利用这些发现提出了一种基于视觉定位模块的双模态 RAG 流程，以在图像检索之前裁剪目标，从而将 GPT-5 和 Qwen 的准确性分别提高了 13.3 和 8.5 个百分点。

View on arXiv Download PDF AI Translation

cs.CV / 54 / 2607.21174

Decoupling Cross-Modality Manifold Discrepancy: Leveraging Visible Diffusion Priors for Infrared Super-Resolution

解耦跨模态流形差异：利用可见扩散先验进行红外超分辨率

Hua, Yunpeng, Yu, Hongwei, Li, Jiawei, Liu, Qiankun, Ma, Huimin, Chen, Jiansheng

Abstract

Infrared image super-resolution (IISR) mitigates the limitations imposed by low spatial resolution. Existing methods have recognized that IISR should preserve consistency in global distribution and structural information while enhancing image clarity. However, these methods are either insufficient or overly intrusive, a problem that becomes even more pronounced in diffusion-based models. To address these issues, we propose a dual-path diffusion-based framework for IISR, termed Shift-IISR. The proposed method is designed to improve the consistency of IISR results while preserving the generative capacity of diffusion models. Specifically, we develop a Global Representation Modulation (GRM) module to extract modality-specific information from infrared imagery and guide the global distribution of the diffusion model toward the ground truth. In addition, we introduce a Local Structure Refinement (LSR) module to encourage the model to focus on structural information at each step of the iterative denoising process. Extensive experiments demonstrate that the proposed method effectively improves distributional and structural consistency while maintaining competitive super-resolution performance. The source code of the proposed Shift-IISR can be available at https://github.com/Assassink8/Shift-IISR.

Chinese Translation

红外图像超分辨率（IISR）缓解了低空间分辨率带来的限制。现有方法已认识到，IISR 应该在增强图像清晰度的同时，保持全局分布和结构信息的一致性。然而，这些方法要么不足以满足需求，要么过于干扰，这一问题在基于扩散的模型中尤为突出。为了解决这些问题，我们提出了一种双路径基于扩散的 IISR 框架，称为 Shift-IISR。该方法旨在提高 IISR 结果的一致性，同时保持扩散模型的生成能力。具体而言，我们开发了一个全局表示调制（Global Representation Modulation, GRM）模块，以从红外图像中提取特定模态的信息，并引导扩散模型的全局分布朝向真实值。此外，我们引入了一个局部结构细化（Local Structure Refinement, LSR）模块，以鼓励模型在每一步迭代去噪过程中关注结构信息。大量实验表明，所提出的方法有效地改善了分布和结构的一致性，同时保持了竞争力的超分辨率性能。所提出的 Shift-IISR 的源代码可在 https://github.com/Assassink8/Shift-IISR 获取。

View on arXiv Download PDF AI Translation

cs.CV / 55 / 2607.21179

Out of Sight, Still in Mind: Token Compression for Omni-LLMs

不在视野中，但仍在心中：面向全模态大型语言模型的令牌压缩

Yoo, Suho, Jang, Youngjoon, Cho, Hyebin, Chung, Joon Son

Abstract

The goal of this paper is to reduce the input token cost of Omni-modal large language models (Omni-LLMs) at inference time. Omni-LLMs reason jointly over audio, video and text, but the cost of the three streams is highly unbalanced: visual tokens account for the vast majority of the input, and are highly redundant. In this paper, we propose ReMo, a training-free framework that compresses visual tokens by redistributing their information across modalities: a visual token is kept only if its information appears nowhere else. ReMo achieves this in two ways: (i) it aligns audio and video in a common embedding space, and removes visual tokens already explained by the audio or by other visual tokens; and (ii) it replaces object-level visual tokens with compact text proxies, short descriptions of each object and its location, conveying the same content in far fewer tokens. On Qwen2.5-Omni at two model scales, ReMo removes 54% of the input tokens with no loss in accuracy. Indeed, it slightly exceeds the full-token model, reaching 101.2% and 101.3% of its average accuracy over five audio-visual benchmarks.

Chinese Translation

本文的目标是在推理时降低全模态大型语言模型（Omni-LLMs）的输入令牌成本。Omni-LLMs对音频、视频和文本进行联合推理，但这三种流的成本高度不平衡：视觉令牌占输入的绝大多数，并且高度冗余。本文提出了ReMo，一个无训练的框架，通过在不同模态之间重新分配信息来压缩视觉令牌：只有当视觉令牌的信息在其他地方不存在时，才保留该视觉令牌。ReMo通过两种方式实现这一目标：（i）它将音频和视频对齐到一个共同的嵌入空间，并移除已被音频或其他视觉令牌解释的视觉令牌；（ii）它用紧凑的文本代理替换对象级视觉令牌，即每个对象及其位置的简短描述，以更少的令牌传达相同的内容。在两个模型规模的Qwen2.5-Omni上，ReMo在不损失准确度的情况下移除了54%的输入令牌。实际上，它略微超越了全令牌模型，在五个音视频基准测试中达到了其平均准确度的101.2%和101.3%。

View on arXiv Download PDF AI Translation

cs.CV / 56 / 2607.21190

Physics-Informed Deep Learning Model for Cross-Modality Super-Resolution in Fluorescence Microscopy

基于物理知识的深度学习模型用于荧光显微镜中的跨模态超分辨率

Soltaninezhad, Mohammad, Corbetta, Elena, Larios, Francisco Paez, Jordan, Paul M., Werz, Oliver, Eggeling, Christian, Bocklitz, Thomas

Abstract

Cross-modality image translation offers a route to super-resolution fluorescence microscopy from low-resolution images while reducing phototoxicity and instrumentation demands. However, purely data-driven models can produce visually plausible outputs that are inconsistent with optical image formation. Here, we propose a physics-informed generative adversarial network for confocal-to-STED image translation that incorporates microscope-specific point spread function information into the training objective. Simulated and experimentally measured PSFs were evaluated using a limited paired confocal-STED dataset of TOM20-labeled mitochondria in human primary M2 macrophages acquired across different experimental days. Performance was assessed using reference-based and non-reference-based image-quality metrics, together with complementary frequency- and distribution-sensitive analyses. The no-reference metrics probed physics-relevant image properties, including spatial-frequency content, contrast, and signal-to-noise behavior. PSF-guided models improved structural fidelity, reduced local deviations, and achieved closer agreement with STED references than non-PSF baselines, particularly in frequency-domain analyses. These results demonstrate that optical priors can improve the structural fidelity and physical plausibility of generative microscopy models for cross-modality super-resolution imaging.

Chinese Translation

跨模态图像翻译为从低分辨率图像获得超分辨率荧光显微镜提供了一条途径，同时减少了光毒性和仪器需求。然而，纯数据驱动的模型可能会产生在光学图像形成上不一致的视觉上似是而非的输出。在此，我们提出了一种基于物理知识的生成对抗网络，用于共聚焦到STED（激发发射显微镜）图像翻译，该网络将显微镜特定的点扩散函数（PSF）信息纳入训练目标。利用有限的配对共聚焦-STED数据集评估了模拟和实验测得的PSF，该数据集包含在不同实验天中获取的人类初级M2巨噬细胞中标记为TOM20的线粒体。通过参考基础和非参考基础的图像质量指标，以及补充的频率和分布敏感分析评估了性能。无参考指标探测了与物理相关的图像特性，包括空间频率内容、对比度和信噪比行为。PSF引导的模型提高了结构保真度，减少了局部偏差，并在频域分析中与STED参考的吻合度优于非PSF基线。这些结果表明，光学先验可以改善生成显微镜模型在跨模态超分辨率成像中的结构保真度和物理合理性。

View on arXiv Download PDF AI Translation

cs.CV / 57 / 2607.21213

Learning-based Seam Correspondence Reconstruction in Sewing Patterns

基于学习的缝合线对应重建在缝纫图案中的应用

Wang, Zhendong, Wang, Jintong, Liu, Chen, Jin, Yao, Liu, Ligang, Wang, Huamin

Abstract

Digital sewing patterns typically consist of disjoint 2D panels without explicit stitch annotations, making downstream 3D modeling reliant on labor-intensive expert specification. In this paper, we present a graph-based learning framework that reconstructs two-level stitching information, coarse panel connectivity and fine-grained seam correspondence, from 2D panel geometry alone. At the coarse level, panel connectivity is inferred by predicting panel semantics associated with anatomical body regions, enforcing consistency with body structure and garment design conventions. Based on the reconstructed panel graph, fine-grained seam correspondences between panel pairs are inferred by learning latent edge representations that jointly encode local seam geometry and global garment context through graph message passing. The resulting edge embeddings are subsequently decoded into detailed seam correspondences. Our method supports complex sewing-pattern topologies, including many-to-one correspondences, intra-panel seams, and curved seams. Experiments demonstrate high stitching accuracy and strong generalization across garment styles.

Chinese Translation

数字缝纫图案通常由不相连的二维面板组成，且没有明确的缝合标注，这使得后续的三维建模依赖于劳动密集型的专家规范。在本文中，我们提出了一种基于图的学习框架，该框架仅通过二维面板几何形状重建两级缝合信息，包括粗略的面板连接性和细粒度的缝合线对应关系。在粗略层面上，通过预测与解剖体区域相关的面板语义来推断面板连接性，从而确保与身体结构和服装设计规范的一致性。基于重建的面板图，通过学习潜在边表示来推断面板对之间的细粒度缝合线对应关系，这些边表示共同编码局部缝合几何和全局服装上下文，采用图消息传递的方法。最终，得到的边嵌入被解码为详细的缝合线对应关系。我们的方法支持复杂的缝纫图案拓扑，包括多对一对应、面板内缝合线和曲线缝合线。实验结果表明，该方法在缝合精度和对不同服装风格的强泛化能力方面表现出色。

View on arXiv Download PDF AI Translation

cs.CV / 58 / 2607.21219

DART: A Degradation-Aware Recurrent Transformer for Archival Film Restoration

DART：一种考虑退化的递归变换器用于档案电影修复

Jastrzębski, Mikołaj, Kozłowski, Wojciech, Adamczewski, Kamil

Abstract

Archival film restoration is a challenging problem because historical footage contains compound degradations such as scratches, dust, blur, noise, flicker, and photometric aging, while clean reference videos are unavailable. Existing video restoration methods largely treat these degradations implicitly, reconstructing frames without explicit knowledge of where damage occurs or how severe it is. We propose DART, a degradation-aware recurrent transformer for archival film restoration. DART predicts and propagates a soft defect mask through time, using it to guide temporal fusion and condition the restoration network on both damage location and severity. This makes the restoration process explicitly aware of film artifacts rather than relying only on reconstruction losses. Experiments on real archival benchmarks show that DART improves no-reference perceptual quality over prior restoration architectures while remaining compact and efficient, producing cleaner and more temporally consistent restorations of structured film damage.

Chinese Translation

档案电影修复是一个具有挑战性的问题，因为历史影像包含复合退化现象，如划痕、灰尘、模糊、噪声、闪烁和光度老化，同时缺乏干净的参考视频。现有的视频修复方法在很大程度上隐含地处理这些退化现象，重建帧时并未明确了解损坏发生的位置或其严重程度。我们提出了DART，一种考虑退化的递归变换器，用于档案电影修复。DART通过时间预测并传播软缺陷掩膜，利用该掩膜指导时间融合，并使修复网络在损坏位置和严重程度上进行条件化。这使得修复过程明确地关注电影伪影，而不仅仅依赖于重建损失。在真实档案基准上的实验表明，DART在无参考感知质量上优于先前的修复架构，同时保持紧凑和高效，产生更清晰且时间一致性更好的结构化电影损坏修复。

View on arXiv Download PDF AI Translation

cs.CV / 59 / 2607.21228

T-STAR: A Large-Scale Benchmark for Spatio-Temporal Panoptic Scene Graph Generation in Satellite Video

T-STAR：卫星视频中时空全景场景图生成的大规模基准

Wang, Linlin, Yang, Xue, Zhou, Zhihuang, Zhong, Zhenyu, Zhang, Ruiyuan, Li, Yansheng

Abstract

Structured understanding of satellite video is essential for advancing dynamic geospatial scene analysis from low-level perception to high-level cognition. To move beyond object-centric perception, this paper introduces spatio-temporal panoptic scene graph generation (TPSG) in satellite video as a new benchmark task. TPSG aims to generate a structured graph composed of a set of triplets with explicit temporal spans, thereby describing dynamic geospatial scenes by jointly modeling identity-consistent instance masks and spatio-temporal relationships among panoptic scene elements. However, there is still no dedicated dataset for TPSG in satellite video. Moreover, TPSG in satellite video is intrinsically challenging, as objects are often small and weakly textured, cross-frame association is easily disrupted by occlusion and background clutter, and relationship semantics are highly coupled with spatial structure and temporal evolution. Consequently, TPSG models developed for natural videos are not directly applicable to satellite video. This paper presents T-STAR, a large-scale benchmark dataset for TPSG in satellite video, comprising over 1.1 million instance masks and over 3.8 million spatio-temporal triplets across 39 fine-grained object categories and 70 fine-grained relationship categories. To enable TPSG in satellite video, we propose a unified framework to enhance cross-frame instance consistency and spatio-temporal relationship prediction. Extensive experiments demonstrate the significance of T-STAR and the effectiveness of the proposed framework, establishing a strong benchmark for future research on structured satellite video understanding. The dataset and code are available at https://github.com/linlin-dev/T-STAR.

Chinese Translation

对卫星视频的结构化理解对于将动态地理场景分析从低级感知提升到高级认知至关重要。为了超越以对象为中心的感知，本文提出了卫星视频中的时空全景场景图生成（TPSG）作为一项新的基准任务。TPSG旨在生成一个由一组三元组组成的结构化图，具有明确的时间跨度，从而通过联合建模身份一致的实例掩膜和全景场景元素之间的时空关系来描述动态地理场景。然而，目前尚无专门针对卫星视频的TPSG数据集。此外，卫星视频中的TPSG本质上具有挑战性，因为对象通常较小且纹理较弱，跨帧关联容易受到遮挡和背景杂乱的干扰，关系语义与空间结构和时间演变高度耦合。因此，针对自然视频开发的TPSG模型并不直接适用于卫星视频。本文提出了T-STAR，一个用于卫星视频中TPSG的大规模基准数据集，包含超过110万个实例掩膜和超过380万个时空三元组，涵盖39个细粒度对象类别和70个细粒度关系类别。为了实现卫星视频中的TPSG，我们提出了一个统一框架，以增强跨帧实例一致性和时空关系预测。大量实验表明了T-STAR的重要性和所提框架的有效性，为未来的结构化卫星视频理解研究建立了强有力的基准。数据集和代码可在 https://github.com/linlin-dev/T-STAR 获取。

View on arXiv Download PDF AI Translation

cs.CV / 60 / 2607.21239

Stokes-Informed Diffusion for Robust Linear Polarization Estimation

基于斯托克斯信息的鲁棒线性偏振估计扩散方法

Luo, Yidong, Li, Chenggong, Feng, Yuchao, Shi, Boxin, Zhang, Junchao, Yuan, Xin

Abstract

Polarization cues benefit applications such as material detection and de-reflection, yet acquiring them typically requires dedicated hardware. This motivates us to estimate the linear polarization from a single RGB image. However, the task is inherently ill-posed, with the Angle of Polarization (AoP) becoming particularly unstable in weak polarization regions, where the polarimetric signal is overwhelmed by noise, leading to erratic angle estimates. To address these limitations, we propose GenPolar, a Stokes-informed diffusion framework grounded in the Mueller formalism from an intensity observation. Specifically, GenPolar predicts channel-wise linear Stokes components (S1,S2) from intensity S0, from which degree of linear polarization (DoLP) and AoP are analytically derived; AoP is further supervised with an observability-aware loss. In addition, to enable efficient and high-fidelity inference, we adopt a two-stage training strategy. Firstly, a multi-step conditional diffusion model is trained with a physics-based loss. Subsequently, we distill it into a one-step generator, which further supports stable Low-Rank Adaptation (LoRA) of the VAE encoder to mitigate domain-specific autoencoding bias. Extensive experiments across rotating-polarizer, division-of-focal-plane, and hybrid datasets demonstrate that GenPolar achieves state-of-the-art performance in both DoLP fidelity and AoP stability. Crucially, these improvements translate to significant and consistent gains in downstream applications, including material detection and de-reflection.

Chinese Translation

偏振线索有助于材料检测和去反射等应用，但获取这些线索通常需要专用硬件。这促使我们从单张RGB图像中估计线性偏振。然而，这一任务本质上是病态的，偏振角（Angle of Polarization, AoP）在弱偏振区域变得特别不稳定，在这些区域，偏振信号被噪声淹没，导致角度估计不稳定。为了解决这些局限性，我们提出了GenPolar，一个基于斯托克斯信息的扩散框架，基于从强度观测中得到的穆勒形式（Mueller formalism）。具体而言，GenPolar从强度S0预测通道-wise线性斯托克斯分量（S1, S2），从中可以解析导出线性偏振度（Degree of Linear Polarization, DoLP）和AoP；AoP进一步通过考虑可观测性损失进行监督。此外，为了实现高效且高保真的推理，我们采用了两阶段训练策略。首先，使用基于物理的损失训练多步条件扩散模型。随后，我们将其提炼为一步生成器，进一步支持变分自编码器（VAE）编码器的稳定低秩适应（Low-Rank Adaptation, LoRA），以减轻领域特定的自编码偏差。在旋转偏振器、焦平面分离和混合数据集上的大量实验表明，GenPolar在DoLP保真度和AoP稳定性方面达到了最先进的性能。重要的是，这些改进在下游应用中转化为显著且一致的收益，包括材料检测和去反射。

View on arXiv Download PDF AI Translation

cs.CV / 61 / 2607.21243

Detectors Learn the Wrong Thing: Shortcut-Resistant Adversarial Training Against Physically Realizable Attacks

检测器学习错误的内容：针对可物理实现攻击的抗捷径对抗训练

Huang, Yuanhao, Ren, Yilong, Wang, Jinlei, Bai, Xuesong, Zhang, Zheng, Yu, Haiyang

Abstract

AI-enabled visual perception systems are increasingly deployed in intelligent transportation infrastructure and autonomous vehicle related applications. However, physically realizable adversarial appearances pose a significant reliability challenge for these safety-critical systems. Adversarial training is effective, but repeated co-occurrence between adversarial texture and positive person instances can cause detectors to treat the texture itself as evidence of object presence, forming a patch texture shortcut. The detector may then treat texture as evidence for the target, causing false detections on texture-only inputs and weakening cross attack generalisation. We propose InsCAT, an instance-level contrastive adversarial training framework that prevents detectors from using adversarial texture as an independent decision cue. SICA aligns adversarial person features with matched clean features and separates them from texture-only negatives, while ROPO and Guard maintain online attack pressure and coordinate training. We evaluate eight independently generated attack textures on rendered nuScenes, INRIAPerson, printed garments, and three detector families. InsCAT achieves an average attack AP of 82.3% on rendered nuScenes, exceeding the strongest baseline by 11.1 points.Relative to AT-Mix, texture FPR decreases from 46.9% to 7.3%. Physical tests yield an F1 score of 96.6% and an FPR of 1.8%. Consistent gains across separately trained detectors demonstrate applicability across architectures with direct inference. The findings show that robust physical detection depends on preserving target related evidence while preventing adversarial texture from becoming an independent decision cu

Chinese Translation

基于人工智能的视觉感知系统越来越多地应用于智能交通基础设施和自动驾驶相关应用。然而，可物理实现的对抗性外观对这些安全关键系统构成了显著的可靠性挑战。对抗训练是有效的，但对抗纹理与正样本（如行人实例）之间的重复共现可能导致检测器将纹理本身视为物体存在的证据，从而形成一个补丁纹理捷径。检测器可能因此将纹理视为目标的证据，导致在仅有纹理的输入上出现误检，并削弱跨攻击的泛化能力。我们提出了InsCAT，一个实例级对比对抗训练框架，旨在防止检测器将对抗纹理作为独立的决策线索。SICA将对抗行人特征与匹配的干净特征对齐，并将其与仅有纹理的负样本分离，而ROPO和Guard则维持在线攻击压力并协调训练。我们在渲染的nuScenes、INRIAPerson、印刷服装和三种检测器家族上评估了八种独立生成的攻击纹理。InsCAT在渲染的nuScenes上实现了82.3%的平均攻击AP，超出最强基线11.1个百分点。相较于AT-Mix，纹理的假阳性率（FPR）从46.9%降至7.3%。物理测试的F1分数为96.6%，FPR为1.8%。在分别训练的检测器上持续获得的增益表明该方法在具有直接推理能力的架构中具有适用性。研究结果表明，稳健的物理检测依赖于保留与目标相关的证据，同时防止对抗纹理成为独立的决策线索。

View on arXiv Download PDF AI Translation

cs.CV / 62 / 2607.21271

Flash EQ-Linear: Accelerating Equivariant Linear Layers via Group-wise Discrete Fourier Transform

Flash EQ-Linear：通过组-wise 离散傅里叶变换加速等变线性层

Zhao, Zhongchen, Wang, Jixin, Xie, Qi, Lin, Hui, Zhang, Lei, Meng, Deyu, Xu, Zongben

Abstract

Equivariant networks embed geometric symmetries as structural priors through weight sharing, achieving remarkable parameter efficiency across vision tasks. However, this parameter efficiency does not translate into compute efficiency: existing implementations unroll the structured weights into dense matrices and dispatch them to generic dense kernels, so the FLOPs of an equivariant layer are no smaller than those of a non-equivariant counterpart. In this paper, we observe that the equivariant linear (EQ-Linear) layer---the most fundamental and frequently used module in modern equivariant architectures---is essentially a circular convolution along the group dimension composed with a linear transform along the channel dimension. Building on this observation, we propose Flash EQ-Linear, an exact acceleration algorithm that reduces the complexity from $\mathcal{O}(NDC)$ to $\mathcal{O}(NDC/T)$ by combining the Fourier convolution theorem along the group dimension with the conjugate symmetry of the real DFT. We further provide dedicated CUDA kernels for Flash EQ-Linear, covering both forward and backward passes and both FP32 and FP16 precision. At the operator level, Flash EQ-Linear achieves up to ${2\times}$ forward speedup over PyTorch's F.linear; at the network level, Flash EQ-ViT and Flash EQ-Swin achieve up to ${1.7\times}$ end-to-end speedup over both equivariant and non-equivariant baselines. To our knowledge, this is the first time equivariant networks strictly dominate their non-equivariant counterparts along all three axes simultaneously: accuracy, parameter efficiency, and inference speed.Code is available at https://github.com/zhongchenzhao/FlashEQLinear.

Chinese Translation

等变网络通过权重共享将几何对称性嵌入结构先验，在视觉任务中实现了显著的参数效率。然而，这种参数效率并未转化为计算效率：现有实现将结构化权重展开为稠密矩阵，并将其分派给通用稠密内核，因此等变层的浮点运算次数（FLOPs）并不小于其非等变对应物。在本文中，我们观察到等变线性（EQ-Linear）层——现代等变架构中最基本且最常用的模块——本质上是在组维度上的循环卷积与通道维度上的线性变换的组合。基于这一观察，我们提出了Flash EQ-Linear，这是一种精确的加速算法，通过将组维度上的傅里叶卷积定理与实数离散傅里叶变换的共轭对称性相结合，将复杂度从 $ ext{O}(NDC)$ 降低到 $ ext{O}(NDC/T)$。我们进一步为Flash EQ-Linear提供专用的CUDA内核，涵盖前向和反向传播，以及FP32和FP16精度。在操作级别，Flash EQ-Linear在前向传播上实现了比PyTorch的F.linear快${2 imes}$；在网络级别，Flash EQ-ViT和Flash EQ-Swin在等变和非等变基准上实现了高达${1.7 imes}$的端到端加速。据我们所知，这是等变网络首次在准确性、参数效率和推理速度三个维度上严格优于其非等变对应物。代码可在 https://github.com/zhongchenzhao/FlashEQLinear 获取。

View on arXiv Download PDF AI Translation

cs.CV / 63 / 2607.21281

HGeo-TopoMap: Boosting Topological Mapping with Hierarchical Geometric Priors

HGeo-TopoMap：利用层次几何先验增强拓扑映射

Li, Siyu, Peng, Kunyu, Wen, Di, Hou, Beiping, Li, Zhiyong, Yang, Kailun

Abstract

Topological maps are key outputs of autonomous driving perception systems, delivering essential road information for path planning. They identify instances such as centerlines and traffic signs, along with their connectivity relationships. Due to the lack of explicit markings for centerlines in real-world environments, the detection of centerline instances remains a significant challenge. To tackle this problem, we propose HGeo-TopoMap, which leverages an explicit prior map and implicit spatial relations to hierarchically boost topological mapping. First, a geometric adaptive learning module is designed for the road structure map obtained via inverse perspective mapping. This module discretely encodes semantic and spatial features from the map, followed by a prior-mask attention mechanism that selectively focuses on informative regions. Then, a geometric consistency learning module is devised, which leverages the geometric properties and spatial relationships of centerlines. Built on the geometry-aware decoder, it enforces spatial consistency by aligning features of centerline instances with identical geometric orientations. The proposed method is evaluated on the OpenLane-V2 dataset across the centerline, lane segment, and robustness benchmarks. Beyond substantial improvements in topological mapping accuracy, the proposed method offers the benefit of enhanced robustness, consistently outperforming baselines under both standard and challenging conditions. The source code and model weights will be made publicly available at https://github.com/lynn-yu/HGeo-TopoMap.

Chinese Translation

拓扑地图是自主驾驶感知系统的关键输出，提供路径规划所需的基本道路信息。它们识别中心线和交通标志等实例及其连接关系。由于现实环境中缺乏对中心线的明确标记，中心线实例的检测仍然是一个重大挑战。为了解决这个问题，我们提出了HGeo-TopoMap，该方法利用显式的先验地图和隐式的空间关系，以层次方式增强拓扑映射。首先，设计了一个几何自适应学习模块，用于通过逆透视映射获得的道路结构图。该模块离散编码地图中的语义和空间特征，随后采用先验掩码注意机制，选择性地关注信息丰富的区域。然后，设计了一个几何一致性学习模块，利用中心线的几何特性和空间关系。该模块基于几何感知解码器，通过对齐具有相同几何方向的中心线实例特征，强制执行空间一致性。所提出的方法在OpenLane-V2数据集上进行了评估，涵盖中心线、车道段和鲁棒性基准测试。除了在拓扑映射精度上显著提升外，所提出的方法还具有增强的鲁棒性，在标准和挑战性条件下均持续优于基线。源代码和模型权重将公开发布在https://github.com/lynn-yu/HGeo-TopoMap。

View on arXiv Download PDF AI Translation

cs.CV / 64 / 2607.21300

Unlearning Under Imbalance: Benchmarking Fairness in Multimodal LLM Unlearning

在不平衡条件下的遗忘：多模态大语言模型遗忘公平性的基准测试

Orsingher, Lorenzo, De Min, Thomas, Mancini, Massimiliano, Talon, Davide, Ricci, Elisa

Abstract

Machine unlearning has emerged as a tool for removing personal data from trained models to comply with recent AI regulations. To evaluate unlearning effectiveness in multimodal large language models (MLLMs), prior works fine-tune models on fictitious identities, simulating unlearning requests on subsets of these IDs, which are typically uniformly distributed. However, in realistic scenarios, people from different demographic groups may request to be unlearned at different frequencies, potentially altering the model's internal beliefs for these groups and leading to biased behaviors. To fill this gap, we propose FAIRGET, the first Visual Question Answering benchmark that evaluates unlearning under unbalanced, realistic, forget requests. These requests are designed to simulate multiple realistic scenarios, ranging from simple to challenging settings, that lead to biased unlearned models if fairness is not accounted for. Additionally, we propose FAUN, the first unlearning algorithm for MLLMs that forgets unlearning data while preserving model fairness. FAUN exploits a bias-aware activation steering mechanism to unlearn identities while accounting for the unbalanced nature of the forget data. Experiments on FAIRGET and the established FIUBench demonstrate our method's superiority both in unlearning quality and fairness.

Chinese Translation

机器遗忘作为一种工具，已成为从训练模型中移除个人数据以遵守近期人工智能法规的手段。为了评估多模态大语言模型（MLLMs）中的遗忘效果，先前的研究在虚构身份上微调模型，模拟对这些身份子集的遗忘请求，而这些身份通常是均匀分布的。然而，在现实场景中，不同人口群体可能以不同频率请求遗忘，这可能改变模型对这些群体的内部信念，并导致偏见行为。为填补这一空白，我们提出了FAIRGET，这是第一个评估不平衡、现实遗忘请求下遗忘效果的视觉问答基准。这些请求旨在模拟多个现实场景，从简单到具有挑战性的设置，如果不考虑公平性，将导致偏见的遗忘模型。此外，我们提出了FAUN，这是第一个针对MLLMs的遗忘算法，能够在保留模型公平性的同时遗忘遗忘数据。FAUN利用一种偏见感知的激活引导机制，在考虑遗忘数据的不平衡特性的同时遗忘身份。我们在FAIRGET和已建立的FIUBench上的实验表明，我们的方法在遗忘质量和公平性方面均表现出优越性。

View on arXiv Download PDF AI Translation

cs.CV / 65 / 2607.21318

PC-Edit: Prompt-Contrastive Region Discovery and Region-Guided Editing

PC-Edit：提示对比区域发现与区域引导编辑

Zhang, Jian, Zhang, Zhijun

Abstract

Replacing an object with one that differs in category or shape requires complete source removal, natural target formation unconstrained by the source silhouette, and preservation of unrelated content. Existing training-free editors either localize edits from terminal predictions under source and target prompts or preserve unrelated content through spatially unselective source-feature reuse without explicit region discovery. Before reaching the terminal predictions, prompt-induced semantic differences undergo additional network transformations that may obscure their spatial localization, reducing localization precision. Spatially unselective feature reuse forces a trade-off between edit completeness and background preservation. Therefore, we propose PC-Edit, a prompt-contrastive framework for training-free MM-DiT editing. PC-Edit contrasts the image-token attention outputs under the source and target prompts, capturing prompt-induced semantic differences directly where text-conditioned information is delivered to image tokens. The same contrast identifies a source-erasure region during inversion and a target-emergence region during denoising. Their union suppresses source remnants while allowing the target object to form naturally. PC-Edit further couples region discovery and background preservation within each sampling step by estimating the current edit region from preceding attention blocks and immediately injecting cached source K/V features outside it in subsequent blocks, thereby protecting unrelated content before the latent update. Experiments on PIE-Bench and our EditRegion-Bench, with human-verified edit-region annotations for single- and multi-object addition and replacement, show that PC-Edit achieves the best editing quality and background preservation among methods without user-specified edit regions.

Chinese Translation

用类别或形状不同的对象替换一个对象需要完全去除源对象、在不受源轮廓限制的情况下自然形成目标对象，并保留无关内容。现有的无训练编辑器要么在源和目标提示下从终端预测中定位编辑，要么通过空间上不具选择性的源特征重用来保留无关内容，而没有明确的区域发现。在达到终端预测之前，提示引发的语义差异经历额外的网络变换，这可能会模糊其空间定位，降低定位精度。空间上不具选择性的特征重用迫使编辑完整性与背景保留之间进行权衡。因此，我们提出了PC-Edit，一种用于无训练MM-DiT编辑的提示对比框架。PC-Edit对比源和目标提示下的图像标记注意力输出，直接捕捉文本条件信息传递给图像标记时引发的语义差异。同样的对比在反转过程中识别源去除区域，在去噪过程中识别目标出现区域。它们的并集抑制源残余，同时允许目标对象自然形成。PC-Edit进一步在每个采样步骤中将区域发现与背景保留结合，通过从前面的注意力块估计当前编辑区域，并立即在后续块中注入缓存的源K/V特征，从而在潜在更新之前保护无关内容。在PIE-Bench和我们的EditRegion-Bench上的实验，结合了经过人工验证的单对象和多对象添加及替换的编辑区域注释，显示PC-Edit在没有用户指定编辑区域的方法中实现了最佳的编辑质量和背景保留。

View on arXiv Download PDF AI Translation

cs.CV / 66 / 2607.21326

SlerpFlow: Spherical Trajectory Correction for Rectified Flow Inversion

SlerpFlow：用于校正流反演的球面轨迹

Duan, Wenbin, Shu, Yan, Fu, Zhuoyuan, Zhao, Fangmin, Li, Yan, Zhao, Yaru, Li, Binyang

Abstract

Rectified-flow-based diffusion transformers, particularly FLUX, have demonstrated outstanding performance in high-quality image generation. However, achieving fast and accurate inversion--transforming images back to latent noise for faithful reconstruction and editing--remains a challenging bottleneck due to the discretization errors of linear solvers. This paper introduces SlerpFlow, a straightforward yet highly effective zero-shot approach that unlocks the full potential of FLUX for high-fidelity inversion and editing. Unlike existing approaches (e.g., RF-Solver) that rely on complex numerical approximations such as high-order Taylor expansions to correct trajectory errors, we present a geometric view based on the Manifold Hypothesis: the empirically observed trajectory curvature is not a numerical artifact, but rather serves as a necessary "centripetal force" that constrains the flow to remain on the data manifold. Guided by this insight, SlerpFlow integrates Spherical Linear Interpolation (Slerp) to rectify flow velocity directions on the hypersphere, strictly adhering to the intrinsic curvature of the latent space. Crucially, by caching the corrected velocity for subsequent steps, SlerpFlow achieves high-precision inversion while maintaining the computational efficiency of a first-order Euler solver. Extensive experiments on FLUX-based reconstruction and editing tasks demonstrate that SlerpFlow improves reconstruction fidelity and achieves stronger semantic alignment in editing without requiring additional training. Code is available at https://github.com/0answer0/SlerpFlow.

Chinese Translation

基于校正流的扩散变换器，特别是 FLUX，在高质量图像生成方面表现出色。然而，实现快速且准确的反演——将图像转换回潜在噪声以进行真实的重建和编辑——仍然是一个由于线性求解器的离散化误差而面临的挑战性瓶颈。本文介绍了 SlerpFlow，一种简单而高效的零-shot 方法，充分发挥 FLUX 在高保真反演和编辑中的潜力。与现有方法（例如 RF-Solver）依赖于复杂的数值近似（如高阶泰勒展开）来修正轨迹误差不同，我们提出了一种基于流形假设的几何视角：经验观察到的轨迹曲率并不是数值伪影，而是作为一种必要的“向心力”，约束流保持在数据流形上。基于这一见解，SlerpFlow 集成了球面线性插值（Slerp），在超球面上校正流速方向，严格遵循潜在空间的内在曲率。重要的是，通过缓存校正后的速度以用于后续步骤，SlerpFlow 实现了高精度反演，同时保持了一阶欧拉求解器的计算效率。在基于 FLUX 的重建和编辑任务上的大量实验表明，SlerpFlow 提高了重建保真度，并在编辑中实现了更强的语义对齐，而无需额外的训练。代码可在 https://github.com/0answer0/SlerpFlow 获取。

View on arXiv Download PDF AI Translation

cs.CV / 67 / 2607.21347

Quality-Aware Multimodal Fusion Reveals Implicit Identity in Valence-Arousal Features

质量感知的多模态融合揭示了情绪-唤醒特征中的隐含身份

Kim, Jisu, Riggan, Benjamin S.

Abstract

Conventional face recognition relies on static appearance cues and degrades in unconstrained settings with expression variation, occlusion, and poor lighting. We hypothesize that audiovisual expression dynamics carry identity-discriminative information complementary to static appearance, and that extracting this signal requires multimodal representations robust to the variable input quality of in-the-wild video. To learn such representations, we cast multimodal valence-arousal (VA) estimation as a pretext task and propose Quality-Aware Adaptive Fusion (QAAF), which estimates per-sample, per-modality reliability and adapts each modality's contribution through learned soft gating and a quality-dependent dropout. For the problem of VA estimation, QAAF achieves an average Concordance Correlation Coefficient (CCC) of 0.472 via late fusion ensembling on Aff-wild2, improving over a baseline ensemble under the same setting (0.415) as well as a single-backbone baseline (0.288). Furthermore, the proposed QAAF demonstrates greater resilience to unavailable modalities, with only a 7.5-34.4% relative decrease in CCC when one modality is missing. We then probe whether these VA-trained features encode identity without identity-specific training. On AFEW-VA (67 actors) and YTF (1,595 subjects), VA-trained backbone features rank first among evaluated soft biometric methods, and score-level fusion with ArcFace lowers EER on both datasets (0.022 to 0.021 on AFEW-VA, 0.106 to 0.104 on YTF), correcting 68.2% of ArcFace's false accepts on AFEW-VA. These findings establish multimodal VA estimation as a soft biometric modality complementary to conventional face recognition.

Chinese Translation

传统的人脸识别依赖于静态外观线索，并在表情变化、遮挡和光照不足的非约束环境中表现不佳。我们假设视听表达动态携带与静态外观互补的身份区分信息，并且提取这一信号需要对野外视频输入质量变化具有鲁棒性的多模态表示。为了学习这种表示，我们将多模态情绪-唤醒（VA）估计视为一个前置任务，并提出了质量感知自适应融合（Quality-Aware Adaptive Fusion, QAAF），该方法通过学习的软门控和质量依赖的丢弃率来估计每个样本、每种模态的可靠性，并调整每种模态的贡献。在VA估计问题上，QAAF通过在Aff-wild2上进行后期融合集成，达到了平均一致性相关系数（Concordance Correlation Coefficient, CCC）0.472，优于相同设置下的基线集成（0.415）以及单一骨干网络基线（0.288）。此外，所提出的QAAF对缺失模态表现出更大的韧性，当缺失一种模态时，CCC仅相对下降7.5-34.4%。随后，我们探讨这些经过VA训练的特征是否在没有特定身份训练的情况下编码身份。在AFEW-VA（67名演员）和YTF（1,595名受试者）上，VA训练的骨干特征在评估的软生物特征方法中排名第一，且与ArcFace的得分级融合降低了两个数据集上的错误接受率（EER）（AFEW-VA从0.022降至0.021，YTF从0.106降至0.104），纠正了ArcFace在AFEW-VA上的68.2%的错误接受。这些发现确立了多模态VA估计作为一种互补于传统人脸识别的软生物特征模态。

View on arXiv Download PDF AI Translation

cs.CV / 68 / 2607.21368

Incremental Optimal Assignment for Real-Time Crowd Tracking

实时人群追踪的增量最优分配

Toroslu, Ismail H.

Abstract

Multi-object tracking in dense crowds requires solving a bipartite assignment problem between detections and trajectories at every video frame. The classical Hungarian algorithm solves this in $O(N^3)$ time, which becomes a bottleneck for large scenes with hundreds of people. We propose an \emph{incremental} assignment algorithm that exploits the block-sparse structure of crowd tracking cost matrices --- dense within each crowd cluster, near-zero between clusters. We compute the exact same optimal $N \times N$ assignment as the Hungarian algorithm, but via an incremental strategy: we add one person at a time, exploiting the fact that after step $n-1$ the dual potentials are \emph{exactly optimal} for the $(n-1)\times(n-1)$ subproblem --- a strictly stronger condition than the intermediate feasibility maintained by the Hungarian algorithm during its $N$ outer iterations. Each new step therefore requires only a single augmenting path search from a certified optimal starting point. This avoids repeated full-matrix scans while guaranteeing an identical globally optimal result. A diagonal-reordering invariant keeps the data structure compact and cache-friendly. On realistic crowd benchmarks with $N \in [200, 5000]$ people organised into dense clusters, our algorithm achieves \textbf{3.7--6.5$\times$ speedup} over the Hungarian baseline while producing provably optimal matchings identical to those of Hungarian. The speedup grows with $N$ and remains stable beyond $N=3000$, making the method especially attractive for large-scale crowd scenes such as stadium exits and mass public events.

Chinese Translation

在密集人群中进行多目标追踪需要在每个视频帧之间解决检测与轨迹之间的二分配问题。经典的匈牙利算法以 $O(N^3)$ 的时间复杂度解决此问题，这在有数百人的大场景中成为瓶颈。我们提出了一种 extit{增量} 分配算法，利用人群追踪成本矩阵的块稀疏结构——在每个聚集的人群内是密集的，而在聚集之间接近零。我们计算与匈牙利算法完全相同的最优 $N imes N$ 分配，但通过增量策略实现：我们一次添加一个人，利用在第 $n-1$ 步之后，双重潜力对于 $(n-1) imes(n-1)$ 子问题是 extit{完全最优} 的这一事实——这是比匈牙利算法在其 $N$ 次外部迭代过程中保持的中间可行性更严格的条件。因此，每个新步骤只需从一个经过认证的最优起点进行单次增广路径搜索。这避免了重复的全矩阵扫描，同时保证了相同的全局最优结果。对角重排不变性保持数据结构紧凑且适合缓存。在包含 $N ext{ in } [200, 5000]$ 的现实人群基准测试中，我们的算法在生成与匈牙利算法相同的可证明最优匹配的同时，实现了 extbf{3.7--6.5$ imes$ 的加速}。加速随着 $N$ 的增加而增长，并在 $N=3000$ 以上保持稳定，使该方法在大型人群场景（如体育场出口和大型公共活动）中特别具有吸引力。

View on arXiv Download PDF AI Translation

cs.CV / 69 / 2607.21370

ASTRA-Net: Anatomy-Specific Transfer and Representation Alignment for Drug-Induced Sleep Endoscopy Segmentation

ASTRA-Net：针对药物诱导睡眠内窥镜分割的解剖特异性转移与表示对齐

Sun, Suhua, Wang, Yuqiao, Liu, Sheng, Fan, Rui, Wang, Jiajun, Xu, Ruoyan, Chen, Yixin, Li, Tao, Yan, Yan

Abstract

Quantitative drug-induced sleep endoscopy (DISE) requires reliable airway boundaries at specific anatomical levels. Pixel-level DISE annotations are scarce, and manual contouring limits the scalability of quantitative assessment. To address this limitation, we developed ASTRA-Net for known-plane DISE segmentation with limited real annotations. Stage 1 aligned intermediate ConvNeXt-Base representations from 14,250 unlabeled virtual endoscopy frames derived from computed tomography and real DISE frames. Virtual images were used only for feature alignment. Stage 2 fine-tuned four independent UNet++ decoders on 401 real annotated frames. Structured zero-mask supervision constrained incompatible plane outputs and invalid frames. Six alignment configurations used maximum mean discrepancy, domain adversarial learning, or both objectives. On a hold-out evaluation set of 100 frames, the five-model MMD-only segmentation ensemble achieved a mean Dice of 0.8927, with a 95% image-level bootstrap interval of 0.8631 to 0.9160. The mean intersection over union was 0.8239. A classification- enabled variant of the same alignment configuration reached a restricted four-plane top-1 accuracy of 0.92 on the same hold-out frames. These results indicate that ASTRA-Net can support frame-level, plane-specific DISE boundary delineation when real annotations are limited.

Chinese Translation

定量药物诱导睡眠内窥镜（DISE）需要在特定解剖层面上可靠的气道边界。像素级DISE标注稀缺，手动轮廓描绘限制了定量评估的可扩展性。为了解决这一限制，我们开发了ASTRA-Net，用于已知平面的DISE分割，且仅依赖有限的真实标注。第一阶段对来自计算机断层扫描和真实DISE帧的14,250个未标记虚拟内窥镜帧的中间ConvNeXt-Base表示进行了对齐。虚拟图像仅用于特征对齐。第二阶段在401个真实标注帧上微调了四个独立的UNet++解码器。结构化的零掩码监督限制了不兼容平面输出和无效帧。六种对齐配置使用最大均值差异、领域对抗学习或两者的目标。在100帧的保留评估集中，五模型的仅MMD分割集成达到了0.8927的平均Dice，图像级95%自助区间为0.8631至0.9160。平均交并比为0.8239。相同对齐配置的分类启用变体在相同的保留帧上达到了限制四平面顶级准确率0.92。这些结果表明，ASTRA-Net能够支持在真实标注有限的情况下进行帧级、平面特异性的DISE边界划定。

View on arXiv Download PDF AI Translation

cs.CV / 70 / 2607.21371

DINOde: Continuous Vision-Text Alignment for Open-Vocabulary Semantic Segmentation

DINOde：用于开放词汇语义分割的连续视觉-文本对齐

Yoon, Sung-Hoon, Kwon, Hoyong, Oh, Changgyoon, Yoon, Kuk-Jin

Abstract

Open-vocabulary semantic segmentation (OVSS) leverages textual semantics to segment objects beyond predefined categories. While the self-supervised model DINOv3 provides strong structured visual representations, its lack of native textual alignment hinders its direct application to OVSS. To bridge this gap, we propose DINOde, an ODE-based framework that continuously aligns CLIP text embeddings with the DINO visual manifold. Our approach employs two complementary components: (i) Semantic Text Flow (STF), which evolves text embeddings toward the DINO manifold through a continuous ODE trajectory, and (ii) Global Context Flow (GCF), which progressively refines the holistic image representation carried by DINO's CLS token. To preserve the hyperspherical geometry of the feature space during this evolution, we further introduce Velocity Tangent Projection, which constrains the learned velocity field to the tangent space. By modeling alignment as a continuous trajectory, DINOde avoids the manifold entanglement inherent in discrete MLP projections and yields more robust cross-modal alignment. Extensive experiments demonstrate that DINOde consistently outperforms existing methods and achieves state-of-the-art performance across multiple OVSS benchmarks. The code is available at https://github.com/yoon307/DINOde.

Chinese Translation

开放词汇语义分割（OVSS）利用文本语义对超出预定义类别的对象进行分割。尽管自监督模型DINOv3提供了强大的结构化视觉表示，但其缺乏原生的文本对齐限制了其在OVSS中的直接应用。为了解决这一问题，我们提出了DINOde，一个基于常微分方程（ODE）的框架，能够持续地将CLIP文本嵌入与DINO视觉流形对齐。我们的方法采用了两个互补的组件：（i）语义文本流（Semantic Text Flow，STF），通过连续的ODE轨迹将文本嵌入演变至DINO流形；（ii）全局上下文流（Global Context Flow，GCF），逐步细化DINO的CLS标记所携带的整体图像表示。为了在这一演变过程中保持特征空间的超球几何，我们进一步引入了速度切向投影（Velocity Tangent Projection），该方法将学习到的速度场限制在切空间内。通过将对齐建模为连续轨迹，DINOde避免了离散多层感知器（MLP）投影中固有的流形缠结，并产生了更为稳健的跨模态对齐。大量实验表明，DINOde在多个OVSS基准测试中始终优于现有方法，并实现了最先进的性能。代码可在https://github.com/yoon307/DINOde获取。

View on arXiv Download PDF AI Translation

cs.CV / 71 / 2607.21401

When Are Reasoning-Based Guardrails Not Efficient? ResponseGuard: A Fast Vision-Language Guard for Real-Time Moderation

推理基础的安全防护何时效率不高？ResponseGuard：一种用于实时审查的快速视觉-语言防护

Na, Dongbin

Abstract

A vision-language AI assistant returns its answer as a stream of generated tokens. Therefore, a safety guard that watches that answer has to keep up with the stream and stop a harmful reply before a user reads it. Recent vision-language guardrails instead generate a chain of thought before they issue a verdict. They believe that step-by-step reasoning yields a safer guard. This design makes the guard heavy and slow, since the model must decode many tokens for harmfulness detection. We pose the question of whether a vision-language guard really needs to reason in order to screen a response. We answer with a guard that has no chain. ResponseGuard reads a harmful verdict from a single pooled representation of the request, the response, and the image in one forward pass. Across a standard multimodal guardrail benchmark, our 2B ResponseGuard outperforms a recent 3B reasoning-based vision-language guard on response harmfulness detection, without any reasoning and at about 150 times lower time cost. On request harmfulness the reasoning guard retains an overall lead, and the remaining gap on both tracks sits on the image-only cells. We observe that the gap may stem from the frozen vision encoders that both designs use rather than from the missing chain. We have also found the reasoning guard directs almost none of its verdict attention to the image. Based on a single-pass detection, ResponseGuard can screen an answer sentence by sentence as it streams and stop a harmful answer before it finishes. For guarding the response of a vision-language model, a calibrated single-pass label may provide a sufficient safety signal. We fully release all source code, trained models, and datasets at https://github.com/ndb796/ResponseGuard.

Chinese Translation

视觉-语言 AI 助手将其答案以生成的标记流的形式返回。因此，监控该答案的安全防护必须跟上这一流，并在用户阅读之前阻止有害的回复。最近的视觉-语言防护则在给出裁决之前生成一系列思考。他们认为逐步推理会产生更安全的防护。这种设计使得防护变得笨重且缓慢，因为模型必须解码许多标记以进行有害性检测。我们提出一个问题：视觉-语言防护是否真的需要推理来筛选回复。我们提出了一种没有推理链的防护。ResponseGuard 通过一次前向传递，从请求、回复和图像的单一汇聚表示中读取有害裁决。在标准的多模态防护基准测试中，我们的 2B ResponseGuard 在回复有害性检测上超越了一种最近的 3B 推理基础的视觉-语言防护，且没有任何推理，时间成本约低 150 倍。在请求有害性方面，推理防护仍保持总体领先，而在两个轨道上的剩余差距则集中在仅图像的单元上。我们观察到，这一差距可能源于两种设计所使用的冻结视觉编码器，而非缺失的推理链。我们还发现，推理防护几乎没有将其裁决注意力指向图像。基于单次检测，ResponseGuard 可以在流式传输时逐句筛选答案，并在其完成之前停止有害答案。对于视觉-语言模型的回复防护，经过校准的单次传递标签可能提供足够的安全信号。我们已在 https://github.com/ndb796/ResponseGuard 上完全发布所有源代码、训练模型和数据集。

View on arXiv Download PDF AI Translation

cs.CV / 72 / 2607.21417

Towards Privacy-Preserving Federated Prompt Tuning under Data Heterogeneity: A Subspace-Decomposed Expert Approach

面向隐私保护的联邦提示调优在数据异质性下的研究：一种子空间分解专家方法

Wang, Yuhua, Li, Xiaodong, Guo, Yihao, Jia, Yuxiang, Zhang, Qinnan, Sun, Yifan, Zhang, Hainan, Tong, Yongxin, Zheng, Zhiming

Abstract

Federated prompt tuning (FPT) enables collaborative adaptation of vision--language models (VLMs) using lightweight prompts. Existing methods often address heterogeneity and privacy through a split-prompt design under local differential privacy (DP), combining a shared prompt for global transfer with private prompts for local adaptation. However, a single shared prompt may over-smooth diverse transferable knowledge, weakening the balance between personalization and generalization. Multi-expert prompts (MEPs) can better capture this diversity, but enlarge the communicated space, increasing DP noise and communication cost while making robust expert composition more difficult. We propose FedSEPT, a privacy-preserving Fed}erated Subspace-decomposed Expert Prompt Tuning. Specifically, we employ Subspace-decomposed Expert Modeling (SEM) to parameterize multiple prompt experts with shared low-rank factors, a fixed public basis, and private residuals, thereby confining communication and DP perturbation to a compact factor space while enabling direct server aggregation in a common coordinate system. We further design Instance-aware Expert Fusion (IEF), which adaptively combines semantically complementary experts via on-device routing and performs efficient logit-level fusion using cached expert-specific text features. Extensive experiments on 11 heterogeneous benchmarks show that, under the same privacy constraints, FedSEPT achieves a better trade-off between local adaptation and global generalization than strong baselines.

Chinese Translation

联邦提示调优（Federated Prompt Tuning, FPT）通过轻量级提示实现视觉-语言模型（Vision-Language Models, VLMs）的协同适应。现有方法通常通过在局部差分隐私（Local Differential Privacy, DP）下的分裂提示设计来解决异质性和隐私问题，结合用于全局迁移的共享提示与用于本地适应的私有提示。然而，单一的共享提示可能会过度平滑多样化的可迁移知识，从而削弱个性化与泛化之间的平衡。多专家提示（Multi-expert Prompts, MEPs）能够更好地捕捉这种多样性，但会扩大通信空间，增加DP噪声和通信成本，同时使得稳健的专家组合变得更加困难。我们提出了FedSEPT，一种隐私保护的联邦子空间分解专家提示调优方法。具体而言，我们采用子空间分解专家建模（Subspace-decomposed Expert Modeling, SEM）来参数化多个提示专家，使用共享的低秩因子、固定的公共基底和私有残差，从而将通信和DP扰动限制在一个紧凑的因子空间内，同时在共同坐标系中实现直接的服务器聚合。我们进一步设计了实例感知专家融合（Instance-aware Expert Fusion, IEF），通过设备上的路由自适应地组合语义互补的专家，并利用缓存的专家特定文本特征进行高效的logit级融合。在11个异质基准上的大量实验表明，在相同的隐私约束下，FedSEPT在本地适应与全局泛化之间实现了比强基线更好的权衡。

View on arXiv Download PDF AI Translation

cs.CV / 73 / 2607.21434

Adaptive Identity Anchoring: Closed-Loop Keyframe Placement for Synthetic Paired Supervision in Video Face Swapping

自适应身份锚定：视频换脸中的闭环关键帧放置用于合成配对监督

Robbins, Logan

Abstract

Video face swapping has no natural paired supervision: no real footage exists of one person's face performing another person's video. The strongest current answer, DreamID-V's SyncID-Pipe, mints pairs by replacing the identity in exactly two frames of a real clip -- the first and the last -- and regenerating the rest from a pose sequence alone. Pose carries no appearance evidence of the swapped-in identity, so over long clips, occlusions, and extreme pose excursions the synthesized identity has a long unanchored span on which to drift; no published ablation examines anchor count or placement. We propose Adaptive Identity Anchoring (AIA): (i) generalize the synthesizer to arbitrary anchor sets, architecturally natural for diffusion-forcing-style transformers where conditioning on a frame is clamping its tokens to zero noise; (ii) place anchors by a closed feedback loop that scores every generated frame against the real reference identity and inserts an image-face-swapped anchor at the worst-scoring frame until the pair passes a threshold or exhausts a budget; (iii) reuse the loop's verdict as an automatic data filter. A second pathology, the beauty-filter look of over-smoothed skin, has the same root cause: micro-texture, like identity, is priced by none of the pipeline's objectives. We therefore pair AIA with Reality-Referenced Texture Restoration: matched re-graining from each real frame's non-face regions, band-split transfer of sub-identity micro-texture from the real footage, and a second, spectral acceptance channel refereed by the footage's own spectrum. Identity-anchor density, we argue, is a controllable quality dial, and we specify falsifiable experiments -- drift-versus-gap curves, uniform-versus-adaptive placement at matched budgets, student training on AIA-minted data, and texture ablations with a human beauty-filter study -- that would validate or refute the proposal.

Chinese Translation

视频换脸没有自然的配对监督：不存在一个人的面孔在另一个人的视频中表演的真实镜头。目前最强的解决方案是DreamID-V的SyncID-Pipe，通过在真实片段的两个帧（第一个和最后一个）中替换身份来生成配对，并仅从姿态序列中再生其余部分。姿态并不携带被替换身份的外观证据，因此在较长的片段、遮挡和极端姿态变化中，合成的身份在没有锚定的情况下会漂移很长一段时间；目前没有发表的消融研究考察锚点的数量或放置。我们提出自适应身份锚定（Adaptive Identity Anchoring, AIA）：(i) 将合成器推广到任意锚点集，这在扩散强制风格的变换器中是架构上自然的，因为对帧的条件相当于将其标记固定为零噪声；(ii) 通过闭环反馈机制放置锚点，该机制对每个生成的帧与真实参考身份进行评分，并在评分最低的帧插入一个图像换脸锚点，直到配对通过阈值或耗尽预算；(iii) 将循环的判决作为自动数据过滤器进行重用。第二种病态现象，即过度平滑皮肤的美容滤镜效果，具有相同的根本原因：微纹理与身份一样，未被管道的任何目标定价。因此，我们将AIA与现实参考纹理恢复相结合：从每个真实帧的非面部区域进行匹配的再颗粒化，从真实视频中进行子身份微纹理的带分离传输，以及由视频自身光谱审核的第二个光谱接受通道。我们认为，身份锚定密度是一个可控的质量调节器，并且我们指定了可证伪的实验——漂移与间隙曲线、在匹配预算下的均匀与自适应放置、基于AIA生成数据的学生训练，以及与人类美容滤镜研究相关的纹理消融——这些实验将验证或反驳该提案。

View on arXiv Download PDF AI Translation

cs.CV / 74 / 2607.21438

DAPM: UAV Monocular Depth Estimation from Any Height, Pitch, Roll and FOV

DAPM：来自任意高度、俯仰角、滚转角和视场的无人机单目深度估计

Ling, Tong, Diao, Wenhui, Feng, Yingchao, Bi, Hanbo, Hou, Zhongyan, Sun, Xian

Abstract

Monocular depth estimation is a fundamental prerequisite for 3D reconstruction and autonomous navigation in Unmanned Aerial Vehicles (UAVs). In practical deployments, UAVs operate under highly dynamic camera poses characterized by continuous variations in height, pitch, roll, and field of view (FOV). Existing monocular depth estimation methods frequently fail to generalize across such diverse perspectives and the expansive scale of depth distributions inherent in aerial scenes. To address these challenges, we establish a quantitative representation of UAV viewing angles through rigorous theoretical analysis, deriving the geometric correspondence between viewing angles and view distances using the ground plane as a reference for observation. Building upon this, we propose Depth Estimation for Any Perspectives Model (DAPM), representing the first monocular framework specifically designed for UAV aerial imagery to jointly estimate camera pose and depth under continuously varying viewpoints. Specifically, we introduce an Ideal Ground Depth (IGD) module that leverages the derived geometric relationships between UAV perspectives and view distances to implement dense camera-pose supervision and enhance depth features. And we further develop a coarse-to-fine Progressive Quantization Bins (PQB) module. By incorporating progressive supervision and hierarchical quantization bins, the PQB module enables robust estimation in complex UAV aerial imagery. To evaluate the proposed framework, we present the UAV Any Perspectives Depth (UAPD) dataset, featuring comprehensive and continuous distributions of pose parameters. Experimental results on UAPD demonstrate that DAPM achieves state-of-the-art performance across both depth and camera-pose estimation metrics. The source code and datasets are available at: https://github.com/ThisIsLT/DAPM.

Chinese Translation

单目深度估计是无人机（UAV）进行三维重建和自主导航的基本前提。在实际应用中，无人机在高度、俯仰角、滚转角和视场（FOV）等高度动态的相机姿态下操作。这些现有的单目深度估计方法常常无法在如此多样的视角和空中场景固有的深度分布广泛范围内进行有效泛化。为了解决这些挑战，我们通过严格的理论分析建立了无人机视角的定量表示，推导出以地面平面作为观察参考的视角与视距之间的几何对应关系。在此基础上，我们提出了任意视角深度估计模型（Depth Estimation for Any Perspectives Model，DAPM），这是第一个专门为无人机空中图像设计的单目框架，旨在在不断变化的视点下联合估计相机姿态和深度。具体而言，我们引入了理想地面深度（Ideal Ground Depth，IGD）模块，利用推导出的无人机视角与视距之间的几何关系，实现密集的相机姿态监督并增强深度特征。此外，我们进一步开发了粗到细的渐进量化箱（Progressive Quantization Bins，PQB）模块。通过结合渐进监督和分层量化箱，PQB模块能够在复杂的无人机空中图像中实现稳健的估计。为了评估所提出的框架，我们呈现了无人机任意视角深度（UAV Any Perspectives Depth，UAPD）数据集，包含全面和连续的姿态参数分布。在UAPD上的实验结果表明，DAPM在深度和相机姿态估计指标上均达到了最先进的性能。源代码和数据集可在以下网址获取：https://github.com/ThisIsLT/DAPM。

View on arXiv Download PDF AI Translation

cs.CV / 75 / 2607.21448

GrainGS: Gradient-Decoupled Gaussian Splatting for Efficient Dynamic Novel View Synthesis

GrainGS：用于高效动态新视图合成的梯度解耦高斯点云

He, Jiahao, Shao, Yihua, Zhao, Zhengkai, Gao, Pan, Ma, Fei, Guo, Jingcai, Tang, Hao, Sebe, Nicu, Tian, Qi

Abstract

Dynamic scene reconstruction with 3D Gaussian Splatting requires a balance between fine-grained motion modeling, structural stability, and compact representation. Existing per-primitive methods provide flexible local deformation but often suffer from redundant primitive growth, while anchor-based methods improve spatial regularity at the cost of suppressing locally varying motion. To address these issues, we present GrainGS, a dynamic Gaussian framework that combines a hierarchical anchor scaffold with per-Gaussian deformation. A static warm-up stage first establishes a time-invariant canonical representation from observations across all timestamps. During joint training, a stop-gradient operation blocks the deformation-mediated gradient pathway to the canonical positions while preserving their direct refinement through the reconstruction objective. Each Gaussian then predicts independent temporal offsets for position, rotation, and scale, enabling detailed local motion within a structurally constrained scaffold. A canonical-residual appearance decomposition further models frame-dependent photometric changes without forcing them into geometric deformation. Experiments on synthetic monocular and real-world multiview benchmarks show that GrainGS achieves high reconstruction quality, real-time novel view synthesis, and compact storage. Under the synthetic benchmark setting, it reaches an average peak signal-to-noise ratio of 36.98 decibels, renders at 435.6 frames per second, and requires 4.67 megabytes of storage.

Chinese Translation

使用3D高斯点云进行动态场景重建需要在细粒度运动建模、结构稳定性和紧凑表示之间取得平衡。现有的每个原始方法提供灵活的局部变形，但往往面临冗余原始增长的问题，而基于锚点的方法则在提高空间规则性的同时抑制了局部变化的运动。为了解决这些问题，我们提出了GrainGS，一个动态高斯框架，结合了分层锚点支架和每个高斯的变形。静态预热阶段首先从所有时间戳的观测中建立一个时间不变的典型表示。在联合训练过程中，停止梯度操作阻断了通过变形介导的梯度路径到达典型位置，同时通过重建目标保留其直接细化。每个高斯随后预测位置、旋转和缩放的独立时间偏移，从而在结构约束的支架内实现详细的局部运动。典型-残差外观分解进一步建模帧依赖的光度变化，而不强制将其纳入几何变形。在合成单目和真实世界多视图基准测试中的实验表明，GrainGS实现了高重建质量、实时新视图合成和紧凑存储。在合成基准测试设置下，其平均峰值信噪比达到36.98分贝，以435.6帧每秒的速度渲染，并需要4.67兆字节的存储空间。

View on arXiv Download PDF AI Translation

cs.CV / 76 / 2607.21456

SPDCN: Strip-based Deformable Convolutional Network for Steel Surface Defect Segmentation

SPDCN：基于条带的可变形卷积网络用于钢材表面缺陷分割

Liu, Zhongming, Jiang, Bingbing, Wan, Guangxin, Zou, Xiang

Abstract

Steel surface defect segmentation is critical for industrial quality inspection, yet existing methods struggle with elongated, anisotropic defects such as cracks and scratches due to the isotropic receptive fields of standard convolutions and rigid sampling grids that cannot adapt to irregular defect boundaries. To address these limitations, we propose Strip-based Predictor for Deformable Convolutional Networks (SPDCN) with two key innovations. The \textbf{Fuzzy-enhanced Multi-scale Context Module (FMCM)} employs group-wise multi-branch convolutions with an intuitionistic fuzzy channel attention mechanism to adaptively capture multi-scale contextual information across varying defect sizes. The \textbf{Adaptive Direction-Aware Deformable Convolution (ADADC)} replaces the conventional offset predictor with decoupled horizontal and vertical strip convolutions, enabling the deformable sampling grid to anisotropically align with the principal orientation of elongated defects. Extensive experiments on public steel surface defect benchmarks demonstrate that SPDCN consistently outperforms state-of-the-art methods, achieving 89.60\% mIoU on NEU-Seg with only 3.54M parameters. The source code is publicly available at https://github.com/DWlzm .

Chinese Translation

钢材表面缺陷分割对于工业质量检测至关重要，但现有方法在处理如裂纹和划痕等细长、各向异性缺陷时表现不佳，这主要是由于标准卷积的各向同性感受野和无法适应不规则缺陷边界的刚性采样网格。为了解决这些局限性，我们提出了基于条带的可变形卷积网络预测器（SPDCN），其具有两个关键创新。 extbf{模糊增强多尺度上下文模块（FMCM）}采用了基于组的多分支卷积，并结合直观模糊通道注意机制，以自适应捕捉不同缺陷尺寸下的多尺度上下文信息。 extbf{自适应方向感知可变形卷积（ADADC）}用解耦的水平和垂直条带卷积替代了传统的偏移预测器，使得可变形采样网格能够各向异性地与细长缺陷的主方向对齐。在公共钢材表面缺陷基准测试上的大量实验表明，SPDCN在性能上始终优于最先进的方法，在NEU-Seg上实现了89.60\%的mIoU，仅使用3.54M参数。源代码已公开，地址为 https://github.com/DWlzm 。

View on arXiv Download PDF AI Translation

cs.CV / 77 / 2607.21467

CLUIE: Clustering-Aware Recurrent Propagation with Local Structural Compensation for Underwater Image Enhancement

CLUIE：具有局部结构补偿的聚类感知递归传播用于水下图像增强

Jiang, Kui, Feng, Zefan, Chang, Laibin, Luo, Yan, Jiang, Junjun, Fan, Xiaopeng

Abstract

Underwater image enhancement remains challenging due to wavelength-dependent light absorption, scattering, and backscattering, which jointly cause color distortion, contrast degradation, and detail loss. Since these degradations vary with scene depth and imaging conditions, different regions within the same image often exhibit heterogeneous degradation patterns and thus require region-adaptive restoration. Although visual RWKV models offer an efficient linear-complexity solution for long-range dependency modeling, their predefined scanning orders are content-agnostic and therefore fail to adapt recurrent state propagation to spatially non-uniform restoration demands. To address this limitation, we propose a Clustering-aware RWKV framework, termed CRWKV, which reformulates the fixed recurrent propagation path of conventional RWKV into a content-adaptive token trajectory. Specifically, we introduce Clustering-aware Semantic Dynamic Reordering (CSDR), which groups tokens according to semantic feature similarity and derives a dynamic traversal order from inter-cluster contextual relations. This design enables WKV states to be accumulated along semantically correlated regions rather than fixed spatial or spectral orders. Since dynamic reordering may disrupt the local continuity of original spatial neighborhoods, we further propose Dark-response Modulated Local Propagation (DMLP), which extracts local structural responses via depth-wise convolution and adaptively modulates their propagation strength using a neighborhood-aware pseudo-dark response map. In this way, local structural cues are compensated before recurrent aggregation while preserving content-adaptive long-range modeling. Extensive experiments on multiple underwater image enhancement benchmarks demonstrate that CRWKV achieves state-of-the-art quantitative performance and superior visual quality.

Chinese Translation

水下图像增强仍然面临挑战，因为光的吸收、散射和反向散射依赖于波长，这共同导致颜色失真、对比度下降和细节丢失。由于这些退化随着场景深度和成像条件而变化，同一图像中的不同区域往往表现出异质的退化模式，因此需要区域自适应的恢复。尽管视觉RWKV模型为长距离依赖建模提供了一种高效的线性复杂度解决方案，但其预定义的扫描顺序与内容无关，因此无法将递归状态传播适应于空间上不均匀的恢复需求。为了解决这一限制，我们提出了一种聚类感知的RWKV框架，称为CRWKV，它将传统RWKV的固定递归传播路径重新构造成内容自适应的标记轨迹。具体而言，我们引入了聚类感知语义动态重排序（CSDR），根据语义特征相似性对标记进行分组，并从簇间上下文关系中推导出动态遍历顺序。该设计使得WKV状态能够沿着语义相关区域进行累积，而不是固定的空间或光谱顺序。由于动态重排序可能会破坏原始空间邻域的局部连续性，我们进一步提出了暗响应调制局部传播（DMLP），通过深度卷积提取局部结构响应，并使用邻域感知的伪暗响应图自适应调制其传播强度。通过这种方式，在递归聚合之前补偿局部结构线索，同时保持内容自适应的长距离建模。在多个水下图像增强基准上的大量实验表明，CRWKV实现了最先进的定量性能和优越的视觉质量。

View on arXiv Download PDF AI Translation

cs.CV / 78 / 2607.21471

Future Rendering $\neq$ Future Surface: A Benchmark and Dataset for Dynamic Surface Reconstruction Beyond the Observed Window

未来渲染 $ eq$ 未来表面：超越观察窗口的动态表面重建基准与数据集

Shi, Yukun, Gong, Minglun

Abstract

Dynamic-scene reconstruction is almost always evaluated inside the observed time window, yet deployment settings such as AR overlays, robot interaction, and anticipatory planning need the future surface: the geometry at times beyond those captured. No standard benchmark measures this. We introduce FutureSurf, a controlled diagnostic benchmark and dataset for future-time surface reconstruction that trades scene diversity for exact future ground truth and falsification controls. A method trains on the observed first 75% of a sequence; we score its extracted per-frame surface on the held-out future by Chamfer distance, reporting absolute future CD as the primary score and the future/observed gap as a diagnostic. The dataset contains eight analytically defined controlled motions, including three falsification controls, with exact per-frame ground-truth meshes. We also provide a ground-truth-side recoverability oracle. The release includes split files, scoring code, a benchmark card, and Croissant metadata. On the controlled motions, the DG-Mesh backbone leaves a 2.7-4.1$\times$ gap even for futures predictable in principle (four of five recoverable from observed motion by a fixed rule), while the falsification controls behave as designed (the surface-invariant motion shows no gap). Beyond the contributed dataset, the gap persists across six animated DG-Mesh asset scenes and a second backbone, Deformable-3DGS (2.0-6.6$\times$; both share a deformation-MLP temporal model). The benchmark also shows that future rendering quality and future-surface accuracy are statistically decoupled, so the novel-view-synthesis metrics the field reports do not track future geometry. The future error is structured, concentrating where the surface moves. The dataset, evaluation toolkit, and scoring code are available on Hugging Face and GitHub (https://github.com/Ricky-S/futuresurf).

Chinese Translation

动态场景重建几乎总是在观察时间窗口内进行评估，然而，诸如增强现实叠加、机器人交互和预期规划等部署场景需要未来表面：即超出捕获时间的几何形状。目前没有标准基准来衡量这一点。我们引入了FutureSurf，这是一个受控的诊断基准和数据集，用于未来时间的表面重建，它在场景多样性与准确的未来真实值和伪造控制之间进行了权衡。该方法在序列的前75%观察数据上进行训练；我们通过Chamfer距离对提取的每帧表面在保留的未来数据上进行评分，报告绝对未来CD作为主要分数，并将未来/观察差距作为诊断指标。该数据集包含八种解析定义的受控运动，包括三个伪造控制，提供了准确的每帧真实网格。我们还提供了一个真实值侧的可恢复性预言者。发布内容包括分割文件、评分代码、基准卡和Croissant元数据。在受控运动中，DG-Mesh主干即使对于原则上可预测的未来（五个中有四个可以通过固定规则从观察运动中恢复）也留下了2.7-4.1$ imes$的差距，而伪造控制则按预期表现（表面不变运动没有差距）。除了贡献的数据集外，这一差距在六个动画DG-Mesh资产场景和第二个主干Deformable-3DGS中仍然存在（2.0-6.6$ imes$；两者共享一个变形-MLP时间模型）。基准还表明，未来渲染质量与未来表面准确性在统计上是解耦的，因此该领域报告的新视图合成指标并未跟踪未来几何形状。未来误差是有结构的，集中在表面移动的地方。数据集、评估工具包和评分代码可在Hugging Face和GitHub上获取（https://github.com/Ricky-S/futuresurf）。

View on arXiv Download PDF AI Translation

cs.CV / 79 / 2607.21485

Recurrent Sinusoidal INRs for Efficient High-Fidelity Representation

高效高保真表示的递归正弦隐式神经表示

Cho, Hyunmin, Yoo, Jaejun, Jin, Kyong Hwan

Abstract

We study sinusoidal recurrence as an iterative mechanism for harmonic spectral enrichment in implicit neural representations (INRs). Our analysis reveals that sinusoidal activations induce a harmonic line spectrum, providing a spectral account of how recurrent unrolling enriches the effective spectral support. We realize this principle with a shared sinusoidal block that iteratively refines the latent representation. We empirically validate the resulting spectral behavior against feed-forward INRs, non-sinusoidal recurrent variants, and equilibrium-style sinusoidal models. Complementing this analysis, we evaluate the proposed architecture across image and 3D representation tasks. On RGB image benchmarks, our method achieves higher fidelity than feed-forward baselines with fewer parameters and fewer optimization steps, and it further transfers favorably to super-resolution, NeRF, and SDF tasks.

Chinese Translation

我们研究了正弦递归作为隐式神经表示（INRs）中谐波谱增强的迭代机制。我们的分析揭示了正弦激活引入了谐波线谱，提供了递归展开如何丰富有效谱支持的谱学解释。我们通过一个共享的正弦块实现了这一原理，该块迭代地细化潜在表示。我们通过与前馈INRs、非正弦递归变体和均衡风格的正弦模型进行比较，实证验证了所得到的谱行为。补充这一分析，我们在图像和3D表示任务中评估了所提架构。在RGB图像基准测试中，我们的方法在参数更少和优化步骤更少的情况下，实现了比前馈基线更高的保真度，并且在超分辨率、NeRF和SDF任务中也表现出良好的迁移能力。

View on arXiv Download PDF AI Translation

cs.CV / 80 / 2607.21504

Texture++: Elevating 3D Asset Texture Resolution with a Region-Aware Diffusion Model

Texture++：通过区域感知扩散模型提升3D资产纹理分辨率

Wang, Shuaiwei, Li, Shi, Xu, Jieting, Huo, Yuchi, Wang, Qi, Zheng, Wenting, Xie, Rengan

Abstract

Numerous 3D assets are discarded due to low texture resolution, while current super-resolution models ignore texture maps and focus on natural images. An efficient and generalizable texture super-resolution model can revitalize a large corpus of aging yet valuable assets across industries such as film and video games. We present Texture++, a novel framework for texture super-resolution, which enhances the low-resolution textures of assets to produce high-resolution, high-quality results. Specifically, we reformulate the task of super-resolution in UV space into performing it across multiple rendered views and merging the outputs. Firstly, to achieve more complete and continuous textures in the view space, we propose an adaptive view selection strategy to integrate textures dispersed across UV texture patches. Furthermore, we introduce a quadtree-based texture region organization method for combining super-resolved textures from different viewpoints, providing masks to distinguish regions that require improvement. Finally, we design a diffusion-based super-resolution model that enhances the texture resolution for specified masked regions, seamlessly integrating with surrounding regions. Through comprehensive evaluations, we demonstrate that our approach yields textures with substantially improved detail and coherence over existing methods.

Chinese Translation

由于纹理分辨率低，许多3D资产被舍弃，而当前的超分辨率模型忽视了纹理图，专注于自然图像。一个高效且可推广的纹理超分辨率模型能够为电影和视频游戏等行业中的大量老旧但有价值的资产注入新的生命。我们提出了Texture++，一个用于纹理超分辨率的新框架，它增强了资产的低分辨率纹理，生成高分辨率、高质量的结果。具体而言，我们将UV空间中的超分辨率任务重新定义为在多个渲染视图中执行并合并输出。首先，为了在视图空间中实现更完整和连续的纹理，我们提出了一种自适应视图选择策略，以整合分散在UV纹理贴片中的纹理。此外，我们引入了一种基于四叉树的纹理区域组织方法，用于结合来自不同视点的超分辨率纹理，并提供掩模以区分需要改进的区域。最后，我们设计了一种基于扩散的超分辨率模型，增强指定掩模区域的纹理分辨率，与周围区域无缝融合。通过全面的评估，我们证明了我们的方法在细节和一致性上显著优于现有方法。

View on arXiv Download PDF AI Translation

cs.CV / 81 / 2607.21526

Boosting Robustness for All-Weather Self-Supervised Depth Estimation in Autonomous Driving

增强自主驾驶中全天气自监督深度估计的鲁棒性

Qi, Mengshi, Bi, Xiaoyang, Zhang, Xianlin, Ma, Huadong

Abstract

Self-supervised depth estimation is challenging for safe autonomous driving under various adverse weather conditions due to sensor perception degradation. These challenges arise from two main aspects. Firstly, adverse conditions can distort pixel correspondences and violate the assumptions embedded in the self-supervised loss function, leading to erroneous depth predictions. Secondly, while radar is a widely adopted sensor in adverse weather conditions, the sparse distribution of radar points in the Point of View (POV) poses challenges for self-supervised fusion. To address these issues, we introduce a novel self-training pipeline using unpaired real all-weather data through multi-teacher distillation and robust radar fusion. We propose the Uncertainty-Aware Multi-Teacher Distillation method to generate diverse teacher models with different adverse condition inputs, and then employ uncertainty modeling to weigh the knowledge distillation loss. Additionally, we design the POV-BEV Radar Fusion approach, which leverages camera-pixel ray constraints to establish connections between the camera's Point of View (POV) and the radar's Bird's-Eye View (BEV). This approach enables the utilization of denser radar points, effectively capturing the complementary perspectives of both POV and BEV. Extensive quantitative and qualitative experiments demonstrate the robustness of our proposed method on all-weather datasets, achieving state-of-the-art performance. Our code and models are available at https://github.com/MICLAB-BUPT/RobustDepth.

Chinese Translation

自监督深度估计在各种恶劣天气条件下进行安全自主驾驶面临挑战，因为传感器感知能力下降。这些挑战主要源于两个方面。首先，恶劣条件可能会扭曲像素对应关系，并违反自监督损失函数中嵌入的假设，从而导致错误的深度预测。其次，尽管雷达是恶劣天气条件下广泛采用的传感器，但在视角（Point of View, POV）中雷达点的稀疏分布对自监督融合构成了挑战。为了解决这些问题，我们引入了一种新的自我训练流程，利用未配对的真实全天气数据，通过多教师蒸馏和鲁棒雷达融合进行训练。我们提出了不确定性感知多教师蒸馏方法，以生成具有不同恶劣条件输入的多样化教师模型，然后采用不确定性建模来加权知识蒸馏损失。此外，我们设计了POV-BEV雷达融合方法，利用相机像素光线约束建立相机视角（POV）与雷达鸟瞰视图（BEV）之间的联系。该方法能够利用更密集的雷达点，有效捕捉POV和BEV的互补视角。大量定量和定性实验表明，我们提出的方法在全天气数据集上具有鲁棒性，达到了最先进的性能。我们的代码和模型可在 https://github.com/MICLAB-BUPT/RobustDepth 获取。

View on arXiv Download PDF AI Translation

cs.CV / 82 / 2607.21529

ElasticTTT: Prior-Preserving Test-Time Tuning for Video Editing

ElasticTTT：保留先验的测试时调优用于视频编辑

Liu, Yueyi, Zhang, Chi, Cui, Sen, Liu, Miao

Abstract

Test-Time Tuning (TTT) on pretrained diffusion models has emerged as a powerful paradigm for video editing. However, there exists a foundational mismatch between the distribution-mapping nature of generative models and the single-point optimization of standard TTT. In this paper, we demonstrate that this mismatch triggers \textit{Prior Collapse}, a degenerate state where the model discards the text conditions and spatial latents, collapsing generations to the source video, or entangling the features of distinct regions. To resolve this, we propose \textbf{ElasticTTT}, a novel framework that preserves the prior generative distribution and rescues generative elasticity. Specifically, we propose \textit{Target Distribution Regularization} to prevent sharp memorization minima, \textit{Contrastive CFG} to guide inference away from source biases, and \textit{Asynchronous Noise Schedule} to preserve unedited regions. Extensive evaluations, supported by theoretical analysis, demonstrate that ElasticTTT successfully preserves the generative prior of the base model, achieving state-of-the-art performance on one-shot video editing.

Chinese Translation

在预训练扩散模型上进行的测试时调优（TTT）已成为视频编辑的强大范式。然而，生成模型的分布映射特性与标准TTT的单点优化之间存在根本性不匹配。本文展示了这种不匹配会引发 extit{先验崩溃}，即模型丢弃文本条件和空间潜变量，导致生成结果退化为源视频，或将不同区域的特征纠缠在一起。为了解决这一问题，我们提出了 extbf{ElasticTTT}，一个新颖的框架，旨在保留先验生成分布并恢复生成的弹性。具体而言，我们提出了 extit{目标分布正则化}以防止尖锐的记忆最小值， extit{对比CFG}以引导推理远离源偏差，以及 extit{异步噪声调度}以保留未编辑区域。通过理论分析支持的广泛评估表明，ElasticTTT成功保留了基础模型的生成先验，在单次视频编辑中实现了最先进的性能。

View on arXiv Download PDF AI Translation

cs.CV / 83 / 2607.21545

Towards Robust Iris Recognition Through Occlusion Identification and Conditional Diffusion-Based Reconstruction

通过遮挡识别和基于条件扩散的重建实现鲁棒虹膜识别

Hasan, Kamrul, Farias, Mylene C. Q., Komogortsev, Oleg V.

Abstract

Iris recognition is a reliable biometric approach that identifies individuals using the distinctive and stable texture of the iris. However, recognition performance can degrade when discriminative iris texture is partially occluded by eyelids, eyelashes, specular reflections, or other acquisition artifacts. Existing approaches often perform recognition directly on degraded samples or rely only on the remaining visible iris region, which may be inadequate when substantial texture is corrupted. To address this limitation, we propose an occlusion-aware iris recognition framework with three sequential modules: occlusion-type identification, diffusion-based reconstruction, and deep-learning-based recognition. First, a residual 2D CNN-based network determines whether an iris image is non-occluded or belongs to one of the controlled occlusion categories. Second, the occluded image, binary mask, and predicted occlusion type condition a denoising diffusion probabilistic model to reconstruct the corrupted region. Finally, VGG19-HPMNet, a modified VGG19 model with horizontal pyramid mapping, extracts discriminative global and part-wise local iris features for recognition. Experiments on the CASIA-Iris-Thousand dataset under a controlled synthetic-occlusion protocol show that the proposed framework improves iris recognition performance by identifying the occlusion type, reconstructing masked regions, and re-evaluating the restored iris samples.

Chinese Translation

虹膜识别是一种可靠的生物特征识别方法，通过虹膜独特且稳定的纹理来识别个体。然而，当虹膜纹理部分被眼睑、睫毛、镜面反射或其他采集伪影遮挡时，识别性能可能会下降。现有方法通常直接对降级样本进行识别，或仅依赖于剩余可见的虹膜区域，这在大量纹理受损时可能不足以满足需求。为了解决这一限制，我们提出了一种考虑遮挡的虹膜识别框架，包含三个顺序模块：遮挡类型识别、基于扩散的重建和基于深度学习的识别。首先，基于残差的二维卷积神经网络（CNN）确定虹膜图像是否未被遮挡或属于某一受控遮挡类别。其次，遮挡图像、二进制掩膜和预测的遮挡类型共同条件化去噪扩散概率模型，以重建受损区域。最后，VGG19-HPMNet（修改版VGG19模型，具有水平金字塔映射）提取可用于识别的区分性全局和局部虹膜特征。在受控合成遮挡协议下对CASIA-Iris-Thousand数据集的实验表明，所提框架通过识别遮挡类型、重建被遮挡区域和重新评估恢复的虹膜样本，提高了虹膜识别性能。

View on arXiv Download PDF AI Translation

cs.CV / 84 / 2607.21546

UnDA: Unpaired Domain Alignment for Cross-Modal Knowledge Transfer in Medical Imaging

UnDA：用于医学影像中跨模态知识转移的无配对领域对齐

Jany, Rafsan, Ahmad, Shadab Tanjeed, Bulbul, Ahsan, Islam, Tahsinul, Hossain, Md Azam, Kamal, Abu Raihan Mostofa

Abstract

Multimodal based approaches often outperform single modality approaches in downstream tasks as the different modalities provide complementary information, yet acquiring paired clinical data remains a significant challenge in real world scenarios. While cross-modal knowledge distillation addresses this, existing methods often struggle with large modality gaps and the propagation of noise from uncertain source-domain predictions. To overcome these challenges, we propose UnDA, an anchor-guided framework for unpaired cross-modal distillation. Our approach introduces a backbone-agnostic Alignment Module that extracts semantically structured class tokens via an attention based pooling mechanism. To ensure robust knowledge transfer, we propose Uncertainty-Weighted Optimal Transport (UCT-OT), which dynamically weights feature-level alignment based on prediction confidence, effectively suppressing noisy supervision. Furthermore, a per-class ProtoNCE objective maintains stable prototype memories to enforce global discriminability across unpaired batches. Evaluations on representative segmentation tasks under strictly unpaired settings show consistent improvements in accuracy and boundary precision in the target modality, demonstrating that meaningful structural knowledge can be transferred across heterogeneous data sources without paired datasets.

Chinese Translation

基于多模态的方法在下游任务中通常优于单一模态的方法，因为不同模态提供了互补的信息，然而在现实场景中获取配对的临床数据仍然是一个重大挑战。尽管跨模态知识蒸馏对此有所解决，但现有方法往往在大模态差距和不确定源域预测的噪声传播方面面临困难。为克服这些挑战，我们提出了UnDA，一种基于锚点引导的无配对跨模态蒸馏框架。我们的方法引入了一个与主干网络无关的对齐模块，通过基于注意力的池化机制提取语义结构化的类别标记。为了确保稳健的知识转移，我们提出了不确定性加权最优传输（UCT-OT），该方法根据预测置信度动态加权特征级对齐，有效抑制噪声监督。此外，逐类的ProtoNCE目标维护稳定的原型记忆，以增强无配对批次间的全局可区分性。在严格的无配对设置下，对代表性分割任务的评估显示目标模态的准确性和边界精度持续提高，证明了有意义的结构知识可以在异构数据源之间转移，而无需配对数据集。

View on arXiv Download PDF AI Translation

cs.CV / 85 / 2607.21553

SANA-Video 2.0: Hybrid Linear Attention with Attention Residuals for Efficient Video Generation

SANA-Video 2.0：具有注意力残差的混合线性注意力用于高效视频生成

Chen, Junsong, Yu, Jincheng, Li, Yitong, Xue, Shuchen, Liu, Haozhe, Xin, Jingyu, Zhao, Yuyang, Ye, Tian, Wu, Zhangjie, Wang, Zian, Zhou, Daquan, Luo, Ping, Han, Song, Xie, Enze

Abstract

We introduce SANA-Video 2.0, a hybrid video diffusion transformer instantiated at 5B and 14B scales under a unified architecture. Designed to generate high-quality video up to 720p on a single GPU, SANA-Video 2.0 matches full-softmax video DiTs in quality while retaining the favorable long-sequence scaling of linear attention. To avoid quadratic attention throughout, Hybrid Linear-Softmax Attention combines gated linear attention for O(N)-dominated mixing with periodic gated-softmax anchors at a 3:1 ratio, restoring the full-rank token interactions that pure linear attention lacks. To propagate these refreshed representations across depth, Block Attention Residuals (AttnRes) route completed block summaries into later linear layers, enabling anchor-feature reuse and boosting deep-layer effective rank by ~12%. Through from-scratch training, SANA-Video 2.0 learns the complete hybrid directly rather than linearizing pretrained models, with reduced-resolution proxy studies establishing 25% softmax as the optimal quality-efficiency trade-off. With 40-step sampling, SANA-Video 2.0 achieves a VBench score of 84.30 in 13.2s at 480p on a single H100, remaining competitive with far larger softmax video DiTs at a fraction of the latency. Its compiled DiT forward pass is 3.2x faster than a matched full-softmax baseline at 720p/60s, a gap that expands with video duration. Furthermore, full-stack Sol-Engine optimization (kernel fusion, caching, and sparse attention) accelerates this hardware-friendly backbone by a further 3.58x, bringing the 5B pipeline to 13.06s at 720p/5s and making it 120x faster than Wan 2.2-A14B on one H100. Overall, our hybrid design recovers softmax-level expressiveness at substantially reduced cost, unlocking scalable long, high resolution video generation.

Chinese Translation

我们介绍了 SANA-Video 2.0，这是一种在统一架构下以 5B 和 14B 规模实例化的混合视频扩散变换器。SANA-Video 2.0 旨在在单个 GPU 上生成高达 720p 的高质量视频，其质量与全软最大视频 DiTs 相匹配，同时保留了线性注意力在长序列扩展方面的优势。为了避免整个过程中的二次注意力，混合线性-软最大注意力结合了门控线性注意力以实现 O(N) 主导的混合，并以 3:1 的比例结合周期性门控软最大锚点，恢复了纯线性注意力所缺乏的全秩令牌交互。为了在深度上传播这些更新的表示，块注意力残差（Block Attention Residuals，AttnRes）将完成的块摘要路由到后续的线性层中，从而实现锚点特征的重用，并将深层有效秩提升约 12%。通过从头训练，SANA-Video 2.0 直接学习完整的混合模型，而不是线性化预训练模型，降低分辨率的代理研究确定 25% 软最大作为最佳质量-效率权衡。在 40 步采样下，SANA-Video 2.0 在单个 H100 上以 480p 在 13.2 秒内达到了 84.30 的 VBench 分数，仍然在延迟的极小部分内与更大规模的软最大视频 DiTs 竞争。其编译的 DiT 前向传播在 720p/60s 下比匹配的全软最大基线快 3.2 倍，这一差距随着视频时长的增加而扩大。此外，全面的 Sol-Engine 优化（内核融合、缓存和稀疏注意力）进一步加速了这一硬件友好的骨干网络，提升了 3.58 倍，使 5B 管道在 720p/5s 下达到 13.06 秒，比 Wan 2.2-A14B 在一台 H100 上快 120 倍。总体而言，我们的混合设计在显著降低成本的同时恢复了软最大级别的表现力，解锁了可扩展的长时间高分辨率视频生成。

View on arXiv Download PDF AI Translation

cs.CV / 86 / 2607.21556

Visual Contrastive Self-Distillation

视觉对比自蒸馏

Liang, Yijun, Tian, Yunjie, Li, Yijiang, Jia, Yuqi, Huang, Furong, Zhou, Tianyi, Fu, Di

Abstract

On-policy self-distillation (OPSD) is promising as it removes the external teacher required by on-policy distillation (OPD), yet it still needs asymmetric information between teacher and student to ensure that the self-teacher provides a stronger learning signal than the student. Existing methods create this asymmetry either through privileged answers or visual evidence. We ask whether both can be removed, yielding a simpler form of OPSD driven purely by input conditioning. For this purpose, we propose Visual Contrastive Self-Distillation, namely VCSD, which converts image-content removal into an on-policy self-distillation signal. At each student-generated response prefix, the EMA teacher produces two next-token distributions under the same prompt and prefix -- one conditioned on the original image and the other on a content-erased control. Their token-wise log-probability difference highlights candidates whose likelihood is specifically increased by the instance-level visual content. We use this contrast to sharpen the teacher's original-image distribution within its plausible support, and distill the resulting full-distribution target into the student. Using ViRL39K dataset, VCSD consistently outperforms matched OPSD across Qwen3-VL and Qwen3.5 models. For example, on Qwen3-VL, it improves the seven-benchmark aggregate from $62.27\% \rightarrow 67.04\%$ at 2B, $71.30\% \rightarrow 73.16\%$ at 4B, and $72.51\% \rightarrow 76.26\%$ at 8B. Furthermore, VCSD requires no external teacher, privileged answers, visual evidence signals, reasoning traces, or additional inference-time cost.

Chinese Translation

在线自蒸馏（On-policy Self-Distillation, OPSD）具有良好的前景，因为它消除了在线蒸馏（On-policy Distillation, OPD）所需的外部教师，但仍然需要教师与学生之间的不对称信息，以确保自教师提供的学习信号强于学生。现有方法通过特权答案或视觉证据来创建这种不对称性。我们探讨是否可以去除这两者，从而产生一种仅由输入条件驱动的更简单的OPSD形式。为此，我们提出了视觉对比自蒸馏（Visual Contrastive Self-Distillation, VCSD），该方法将图像内容的去除转化为在线自蒸馏信号。在每个学生生成的响应前缀下，EMA教师在相同的提示和前缀下生成两个下一个标记分布——一个基于原始图像，另一个基于内容去除的控制。它们的逐标记对数概率差异突出了那些其可能性因实例级视觉内容而特别增加的候选项。我们利用这种对比来增强教师的原始图像分布在其合理支持范围内，并将得到的全分布目标蒸馏到学生中。使用ViRL39K数据集，VCSD在Qwen3-VL和Qwen3.5模型上始终优于匹配的OPSD。例如，在Qwen3-VL上，它将七项基准的总和从$62.27\% ightarrow 67.04\%$（2B），$71.30\ ightarrow 73.16\ ext{%}$（4B），以及$72.51\ ightarrow 76.26\ ext{%}$（8B）提升。此外，VCSD不需要外部教师、特权答案、视觉证据信号、推理痕迹或额外的推理时间成本。

View on arXiv Download PDF AI Translation

cs.CV / 87 / 2607.21562

Scene Parameter Saliency via Differentiable Light Transport

可微光传输下的场景参数显著性

Beresna, Linas, Fiume, Eugene

Abstract

Gradient-based saliency methods reveal which input features most influence a neural network's output, and are a standard tool for model interpretability. We observe that differentiable renderers, which are conventionally used for parameter optimisation, produce an analogous form of saliency: given any scalar metric evaluated on a rendered image, a single reverse-mode differentiation pass yields per-parameter gradients that identify which scene elements most influence the metric. We call these gradient fields metric saliency maps. Unlike neural saliency, which propagates attribution through learned weights, metric saliency propagates through the image formation process itself, including multi-bounce light transport, capturing parameter dependencies that are semi-opaque to manual inspection. We compute metric saliency maps for qualitatively different objectives: psychovisual glare indices, mean scene luminance, and neural perceptual scores. The saliency rankings differ substantially across metrics for the same scene, with parameters that dominate one objective being negligible for another. The saliency map is specific to the metric, not an intrinsic property of the scene. Our results suggest that differentiable renderers produce derivative images that are as informative for scene understanding as the primal images they were designed to generate.

Chinese Translation

基于梯度的显著性方法揭示了哪些输入特征对神经网络的输出影响最大，是模型可解释性的标准工具。我们观察到，可微渲染器通常用于参数优化，产生了一种类似的显著性形式：对于在渲染图像上评估的任何标量度量，单次反向模式微分传递产生的每个参数梯度能够识别出哪些场景元素对该度量影响最大。我们称这些梯度场为度量显著性图。与通过学习权重传播归因的神经显著性不同，度量显著性通过图像形成过程本身传播，包括多次反射光传输，捕捉到手动检查时半透明的参数依赖性。我们计算了针对不同目标的度量显著性图：心理视觉眩光指数、场景平均亮度和神经感知评分。对于同一场景，不同度量的显著性排名差异显著，主导一个目标的参数在另一个目标中可能微不足道。显著性图是特定于度量的，而不是场景的内在属性。我们的结果表明，可微渲染器生成的导数图像在场景理解方面与其设计生成的原始图像同样具有信息量。

View on arXiv Download PDF AI Translation

cs.CV / 88 / 2607.21576

Self-Supervised Learning of Structured Dynamics from Videos

从视频中自监督学习结构化动态

Knobel, Lukas, Zisserman, Andrew, Asano, Yuki M.

Abstract

Understanding motion in video is a fundamental challenge for visual learning, as frame-to-frame change entangles two sources of dynamics: camera motion and object motion. This decomposition has remained underexplored in representation learning, partly because these factors are tightly coupled in natural videos and difficult to supervise separately. Yet recovering it is important for learning robust motion representations that separate meaningful object dynamics from camera-induced variation. We study whether such structured motion representations can be recovered from frozen features of a pretrained image vision transformer. We propose the Structured Dynamics Model (SDM), which explicitly separates the dominant source of temporal change from residual dynamics through future-feature prediction, rather than representing video change with a single entangled latent or with unstructured, spatially dense transition tokens. Training combines self-supervised learning on real video with weak supervision of scene dynamics on synthetic Kubric data. We evaluate SDM on ProbeMotion, a new evaluation suite spanning synthetic and real videos with camera motion, object motion, and combined dynamics. SDM outperforms backbone baselines using global CLS or average-pooled features, and compares favorably to strongly supervised representations such as VGGT on several probes, despite using substantially weaker supervision. These results suggest that pretrained image models can be readily repurposed into structured video-dynamics representations, providing a useful inductive bias for learning and analyzing latent video dynamics.

Chinese Translation

理解视频中的运动是视觉学习的一个基本挑战，因为帧与帧之间的变化纠缠了两种动态来源：相机运动和物体运动。这种分解在表征学习中仍未得到充分探索，部分原因是这些因素在自然视频中紧密耦合且难以单独监督。然而，恢复这种分解对于学习稳健的运动表征至关重要，这样可以将有意义的物体动态与相机引起的变化分离开来。我们研究是否可以从预训练的图像视觉变换器的冻结特征中恢复这种结构化运动表征。我们提出了结构化动态模型（Structured Dynamics Model, SDM），该模型通过未来特征预测明确地将主导的时间变化来源与残余动态分离，而不是通过单一纠缠的潜变量或非结构化的空间密集过渡标记来表示视频变化。训练结合了对真实视频的自监督学习与对合成Kubric数据的场景动态的弱监督。我们在ProbeMotion上评估SDM，这是一个新的评估套件，涵盖了具有相机运动、物体运动和组合动态的合成和真实视频。尽管使用的监督明显较弱，SDM在使用全局CLS或平均池化特征的基础线模型上表现优越，并在多个探测中与强监督表征（如VGGT）相比表现良好。这些结果表明，预训练的图像模型可以轻松地重新用于结构化视频动态表征，为学习和分析潜在视频动态提供了有用的归纳偏置。

View on arXiv Download PDF AI Translation

cs.CV / 89 / 2607.21577

Synthetic data generation framework for quality control automation in gravure printing

用于凹版印刷质量控制自动化的合成数据生成框架

Coulibaly, Korota Arsène, Hamlich, Mohamed, Hmali, Khalid, Trombin, Andrea

Abstract

Quality control in printing, particularly in rotogravure printing, still depends on slow, costly, and subjective manual inspection. Automated surface defect detection is critical for maintaining high-quality standards in rotogravure printing. Deep learning models give prospects for automation. However, training robust deep learning models, such as YOLO or Vision Transformers, is heavily hindered by the extreme scarcity of real-world industrial defects images. To overcome this limitation, this paper introduces a novel synthetic data generation framework tailored for rotogravure printing quality control. The proposed pipeline automatically generates high-fidelity images of specific printing defects (creases, streaks, misregistration, etc.) and outputs corresponding bounding boxes and annotations. To validate the framework, a synthetic dataset of 7533 images was generated and used to train the state-of-the-art object-detection model RFDETR. Experimental results demonstrate that the model trained on our synthetic data achieves a Mean Average Precision (mAP) of 80.9\% on real industrial testing samples. This framework provides a zero-cost, rapid-deployment solution for automating defect inspection in printing lines without requiring massive manual data collection.

Chinese Translation

印刷中的质量控制，特别是在凹版印刷中，仍然依赖于缓慢、昂贵且主观的人工检查。自动化表面缺陷检测对于维持凹版印刷的高质量标准至关重要。深度学习模型为自动化提供了前景。然而，训练稳健的深度学习模型，如 YOLO 或 Vision Transformers，受到现实工业缺陷图像极度稀缺的严重阻碍。为克服这一限制，本文提出了一种新颖的合成数据生成框架，专门针对凹版印刷质量控制。所提出的流程自动生成特定印刷缺陷（如折痕、条纹、错位等）的高保真图像，并输出相应的边界框和注释。为了验证该框架，生成了一个包含 7533 张图像的合成数据集，并用于训练最先进的目标检测模型 RFDETR。实验结果表明，在我们的合成数据上训练的模型在真实工业测试样本上达到了 80.9\% 的平均精度均值（mAP）。该框架为在印刷生产线中自动化缺陷检查提供了一种零成本、快速部署的解决方案，而无需大量的人工数据收集。

View on arXiv Download PDF AI Translation

cs.CV / 90 / 2607.21580

GraphVid: Interactive Graph-Controllable Video Generation

GraphVid：交互式图控视频生成

Shah, Vedant, Susladkar, Onkar, Prakash, Tushar, Nguyen, Kiet, Yu, Tianjio, Juvekar, Adheesh, Waheed, Muntasir, Lourentzou, Ismini

Abstract

Controllable video generation remains challenging due to the difficulty of specifying precise multi-object interactions using text prompts or motion-control inputs that primarily constrain pixel movement. In practice, trajectory-based control often requires users to draw accurate tracks for multiple objects, which scales poorly with scene complexity and becomes ambiguous under occlusion or overlap. To enable flexible yet precise multi-subject control, we introduce $\textbf{GraphVid}$, a graph-conditioned image-to-video generation model that enables interactive control through structured interaction graphs. We further curate $\textbf{GraphVid-Bench}$, a large-scale interaction-centric video dataset with structured relational annotations to enable training of interaction-aware video generation models. Despite using substantially less training data and fewer trainable parameters than prior motion-control methods, GraphVid delivers strong controllability and video quality. Compared with Motion-I2V, GraphVid reduces FID by up to 39.9% and FVD by 37.6%, while improving PSNR (9.87=>15.98) and SSIM (0.38=>0.61). Our results highlight the potential of structured semantic interfaces as a powerful paradigm for controllable video generation.

Chinese Translation

可控视频生成仍然面临挑战，因为使用文本提示或主要限制像素运动的运动控制输入来指定精确的多对象交互非常困难。在实践中，基于轨迹的控制通常要求用户为多个对象绘制准确的轨迹，这在场景复杂性增加时扩展性较差，并且在遮挡或重叠情况下变得模糊。为了实现灵活而精确的多主体控制，我们提出了$ extbf{GraphVid}$，一种图条件的图像到视频生成模型，通过结构化交互图实现交互式控制。我们进一步整理了$ extbf{GraphVid-Bench}$，一个大规模以交互为中心的视频数据集，具有结构化的关系注释，以支持交互感知视频生成模型的训练。尽管使用的训练数据和可训练参数显著少于先前的运动控制方法，GraphVid仍然提供了强大的可控性和视频质量。与Motion-I2V相比，GraphVid将FID降低了多达39.9%，将FVD降低了37.6%，同时提高了PSNR（9.87=>15.98）和SSIM（0.38=>0.61）。我们的结果突显了结构化语义接口作为可控视频生成强大范式的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 91 / 2607.21591

Inference-Time Scaling of Diffusion Models via Progressive Seed Pruning

通过渐进式种子修剪实现扩散模型的推理时间扩展

Guimaraes, Rogerio, Perona, Pietro

Abstract

Diffusion and flow-matching models dominate conditional image generation, yet inference-time scaling for these models is far less developed than for autoregressive language models. Because final quality is highly sensitive to the initial noise seed, many approaches spend extra compute on seed search or resampling under a black-box reward, but typically maintaining a constant memory footprint throughout inference. We show that relaxing this constraint enables an underexplored inference-time scaling axis: by front-loading exploration, evaluating many seeds early, and pruning aggressively, we can use a fixed compute budget more effectively. \emph{Progressive Seed Pruning} (\PSP) scores intermediate denoised estimates and progressively narrows the candidate set so that only promising trajectories are fully denoised, while keeping the total number of model evaluations fixed. Across diffusion and flow-matching backbones, \PSP \ consistently improves reward-guided selection and achieves higher GenEval scores (automated) and better human evaluation on prompt-alignment than best-of-$N$, importance-sampling, and tree-search baselines at matched compute. Project page: https://www.vision.caltech.edu/psp. Code: https://github.com/rogerioagjr/psp.

Chinese Translation

扩散和流匹配模型在条件图像生成中占据主导地位，但这些模型的推理时间扩展远不如自回归语言模型发达。由于最终质量对初始噪声种子高度敏感，许多方法在种子搜索或在黑箱奖励下重新采样时花费额外的计算资源，但通常在推理过程中保持恒定的内存占用。我们展示了放宽这一限制可以启用一个未被充分探索的推理时间扩展轴：通过前期探索、早期评估多个种子并进行积极修剪，我们可以更有效地利用固定的计算预算。渐进式种子修剪（Progressive Seed Pruning, extit{PSP}）对中间去噪估计进行评分，并逐步缩小候选集，使得只有有前景的轨迹被完全去噪，同时保持模型评估的总次数不变。在扩散和流匹配基础模型中， extit{PSP} 一致地改善了基于奖励的选择，并在匹配计算下实现了比最佳的 $N$、重要性采样和树搜索基线更高的自动化 GenEval 分数和更好的人工评估结果。项目页面：https://www.vision.caltech.edu/psp。代码：https://github.com/rogerioagjr/psp。

View on arXiv Download PDF AI Translation

cs.CV / 92 / 2607.21592

Unified Video Dense Prediction from Disjoint Data

来自不相交数据的统一视频密集预测

Sun, Yihong, Oh, Seoung Wug, Huang, Jiahui, Hariharan, Bharath, Lee, Joon-Young

Abstract

Scene understanding requires simultaneous prediction about geometry, appearance, and semantics. However, existing task-specific annotations are fragmented across incompatible, domain-specific datasets. Current unified systems circumvent this by restricting training to fully co-annotated data, or by incurring the large computational cost of pseudo-labeling. To mitigate this, we introduce UniD, a unified video model that jointly predicts eight dense scene properties-depth, surface normals, semantic segmentation, boundaries, human parts, albedo, shading, and materials-all learned from disjoint, domain-specific datasets. We propose a simple yet effective distillation step in which per-task experts supervise a unified backbone through lightweight task projectors, eliminating the need for annotation overlap or pseudo-labeling. Our key insight is that the strong visual priors of a pretrained diffusion model are sufficient to bridge the domain gaps introduced by disjoint training sources, enabling robust generalization to scene-task combinations never seen during training. UniD achieves competitive performance against per-task specialists and multi-task baselines, with strong generalization to out-of-distribution scenarios and enhanced temporal and cross-task consistency. Code and video results are available at https://unid-video.github.io/.

Chinese Translation

场景理解需要同时对几何、外观和语义进行预测。然而，现有的任务特定注释在不兼容的领域特定数据集中是碎片化的。目前的统一系统通过限制训练在完全共同注释的数据上，或通过承担伪标注的高计算成本来规避这一问题。为了解决这个问题，我们提出了UniD，一个统一的视频模型，它共同预测八种密集场景属性——深度、表面法线、语义分割、边界、人类部位、反射率、阴影和材料——这些属性均来自不相交的领域特定数据集。我们提出了一种简单而有效的蒸馏步骤，其中每个任务的专家通过轻量级任务投影器监督一个统一的主干，消除了对注释重叠或伪标注的需求。我们的关键见解是，预训练扩散模型的强视觉先验足以弥合由不相交训练源引入的领域差距，从而实现对训练期间从未见过的场景任务组合的强健泛化。UniD在与每个任务专家和多任务基线的比较中表现出竞争力，并在分布外场景中展现出强大的泛化能力，同时增强了时间一致性和跨任务一致性。代码和视频结果可在 https://unid-video.github.io/ 获取。

View on arXiv Download PDF AI Translation

cs.CV / 93 / 2607.21594

Streaming Multi-Agent Autoregressive Diffusion Model with World State Registers

具有世界状态寄存器的流式多智能体自回归扩散模型

Mo, Sicheng, Li, Yuheng, Leng, Ziyang, Singh, Krishna Kumar, Zhou, Bolei

Abstract

Multi-agent interactive world models should not only generate consistent observations, but also maintain world states that persist across agents and evolve across views. Existing autoregressive video diffusion pipelines carry forward observation history as conditioning context, which makes shared state difficult to maintain in multi-agent and multi-view settings. We present WorldWeaver (W^2), a streaming multi-agent video diffusion model that augments rollout with cross-agent world state registers: learnable tokens that store shared world information, track individual agent status, and are dynamically updated after each generated chunk. We ground these registers with supervision signals spanning individual agent status, global state views including bird's-eye views, and scene text. We further improve the architecture with a Mixture-of-Transformers design that uses separate weights for world state modeling and visual frame modeling. Extensive experiments in two-agent Minecraft video generation show that explicit world-state modeling improves logical consistency and generation quality.

Chinese Translation

多智能体交互世界模型不仅应生成一致的观察结果，还应维护跨智能体持续存在的世界状态，并在不同视角间演变。现有的自回归视频扩散管道将观察历史作为条件上下文进行传递，这使得在多智能体和多视角环境中维护共享状态变得困难。我们提出了WorldWeaver (W^2)，这是一种流式多智能体视频扩散模型，通过跨智能体世界状态寄存器增强了生成过程：这些可学习的标记存储共享的世界信息，跟踪个体智能体状态，并在每次生成的片段后动态更新。我们通过覆盖个体智能体状态、包括鸟瞰图的全局状态视图以及场景文本的监督信号来为这些寄存器提供基础。我们进一步通过混合变换器设计改进了架构，该设计为世界状态建模和视觉帧建模使用了不同的权重。在两个智能体的Minecraft视频生成中的大量实验表明，显式的世界状态建模提高了逻辑一致性和生成质量。

View on arXiv Download PDF AI Translation

cs.CV / 94 / 2607.21595

3D-Aware VLMs with Implicit and Explicit Geometries

具有隐式和显式几何信息的3D感知视觉语言模型

Li, Wenhao, Jiang, Xueying, Qian, Quanhao, Zhao, Deli, Xu, Ran, Lu, Shijian, Zhang, Gongjie

Abstract

Despite rapid progress, most existing vision-language models (VLMs) built from 2D visual inputs often struggle when handling various 3D tasks that require fine-grained spatial understanding and reasoning. To bridge this gap, we present VLM-IE3D, a unified framework that enhances the 3D spatial awareness of VLMs by equipping them with both implicit and explicit 3D geometries learned from RGB videos. Our VLM-IE3D introduces Implicit Geometry Tokens (IGTs) that capture high-level geometric priors from input videos, as well as complementary Explicit Geometry Tokens (EGTs) that encode detailed geometric structures from reconstructed 3D attributes. On top of that, VLM-IE3D comes with a 3D-aware adapter that effectively fuses the two types of geometric representations with 2D visual cues. This RGB-only design injects strong 3D inductive biases for fine-grained spatial understanding and reasoning without requiring any additional 3D inputs. Extensive experiments show that VLM-IE3D achieves superior performance consistently across various 3D tasks including 3D video detection, 3D visual grounding, 3D dense captioning, and spatial reasoning. Code and models are available at https://github.com/Vegetebird/VLM-IE3D.

Chinese Translation

尽管取得了快速进展，但大多数现有的基于2D视觉输入构建的视觉语言模型（VLMs）在处理需要细粒度空间理解和推理的各种3D任务时仍然面临困难。为了解决这一问题，我们提出了VLM-IE3D，这是一个统一框架，通过为VLMs配备从RGB视频中学习的隐式和显式3D几何信息，增强其3D空间感知能力。我们的VLM-IE3D引入了隐式几何标记（Implicit Geometry Tokens, IGTs），用于捕捉输入视频中的高级几何先验，以及互补的显式几何标记（Explicit Geometry Tokens, EGTs），用于编码从重建的3D属性中提取的详细几何结构。此外，VLM-IE3D配备了一个3D感知适配器，有效地将这两种几何表示与2D视觉线索融合。该RGB-only设计为细粒度空间理解和推理注入了强大的3D归纳偏置，而无需任何额外的3D输入。大量实验表明，VLM-IE3D在包括3D视频检测、3D视觉定位、3D密集字幕生成和空间推理等各种3D任务中始终实现了卓越的性能。代码和模型可在https://github.com/Vegetebird/VLM-IE3D获取。

View on arXiv Download PDF AI Translation

人工智能 (Artificial Intelligence)

123

cs.AI / 1 / 2607.20452

AINTMA: Agentic AI Architecture for Autonomous Test Management with Generative Intelligence, Secure Cloud Communication and Adaptive Quality Analytics

AINTMA：具有生成智能、安全云通信和自适应质量分析的自主测试管理代理智能AI架构

Pasupuleti, Vinil, Allala, Shyalendar Reddy, Bayyavarapu, Siva Rama Krishna Varma, Tyagi, Shrey, Songa, Srinivasateja

Abstract

Modern software quality assurance demands intelligent, autonomous systems capable of adaptive decision-making across distributed cloud environments. This paper presents AINTMA (Agentic Intelligent Test Management Architecture), a multi-agent agentic AI system that transforms traditional test management into an autonomous quality intelligence ecosystem. AINTMA deploys six specialized AI agents (Test Discovery, Risk Assessment, Reinforcement Learning Prioritization, Execution Orchestration, Generative Quality Intelligence, and Cloud Security Monitor) coordinated through a secure multi-agent communication framework over a cloud-native microservices infrastructure. The Generative Quality Intelligence agent employs large language models to produce plain language quality narratives, defect risk summaries, and data-augmented test recommendations. The RL Prioritization agent models test selection as a Markov Decision Process, learning contextual policies from large-scale historical test execution data (47 features, rolling 36-month window). Secure cloud communication is enforced through a zero-trust API gateway with OAuth2/JWT authentication, encrypted inter-agent messaging, and multi-tenant isolation. Evaluation across 12 heterogeneous software projects over 18 months demonstrates: 88.4% test prioritization accuracy (APFD, vs. 51.2% random, 82.1% best commercial baseline); 43% test cycle time reduction; defect escape rate reduced from 8.3% to 2.1%; 340% ROI at 9-month payback. The agentic architecture scales to 50,000+ test cases with sub-400ms response time, and the generative intelligence module achieves 4.3/5.0 developer usefulness rating. AINTMA demonstrates that agentic AI, combining autonomous multi-agent coordination, generative intelligence and secure smart connectivity, can fundamentally advance software quality management in cloud-scale enterprise environments.

Chinese Translation

现代软件质量保证要求智能、自主的系统，能够在分布式云环境中进行自适应决策。本文提出了AINTMA（代理智能测试管理架构），这是一个多代理的代理智能AI系统，将传统的测试管理转变为一个自主的质量智能生态系统。AINTMA部署了六个专门的AI代理（测试发现、风险评估、强化学习优先级、执行编排、生成质量智能和云安全监控），通过一个安全的多代理通信框架在云原生微服务基础设施上进行协调。生成质量智能代理利用大型语言模型生成通俗易懂的质量叙述、缺陷风险摘要和数据增强的测试建议。强化学习优先级代理将测试选择建模为马尔可夫决策过程，从大规模历史测试执行数据（47个特征，滚动36个月窗口）中学习上下文策略。通过零信任API网关实施安全云通信，采用OAuth2/JWT身份验证、加密的代理间消息传递和多租户隔离。在18个月内对12个异构软件项目的评估表明：测试优先级准确率为88.4%（APFD，相较于51.2%的随机选择和82.1%的最佳商业基线）；测试周期时间减少43%；缺陷逃逸率从8.3%降低到2.1%；9个月回报期内实现340%的投资回报率。该代理架构可扩展至50,000多个测试用例，响应时间低于400毫秒，生成智能模块获得4.3/5.0的开发者实用性评分。AINTMA展示了代理智能AI结合自主多代理协调、生成智能和安全智能连接，能够在云规模企业环境中根本性地推动软件质量管理的进步。

View on arXiv Download PDF AI Translation

cs.AI / 2 / 2607.20462

Marking the Wrong Symptoms: Evaluating LLM Watermarks in Medical Texts

标记错误症状：评估医疗文本中的大型语言模型水印

Rieff, Melanie, Staab, Robin, Gloaguen, Thibaud, Hegselmann, Stefan, Vechev, Martin

Abstract

Large language models (LLMs) are increasingly integrated into clinical workflows, stressing the need for reliable traceability of model-generated output with watermarking. Yet, most watermarks are evaluated on general-purpose benchmarks, leaving domains like medicine, where small token-level perturbations can result in significant semantic changes, underexplored. In this work, we present the first rigorous study of how LLM watermarks affect medical performance, benchmarking 5 watermarking schemes across 11 LLMs and 7 VLMs on various tasks spanning unimodal and multimodal clinical reasoning. Importantly, we complement existing evaluations by introducing a human-expert-validated pipeline for systematically auditing medical reasoning quality, terminological precision, and induced hallucinations. Our results reveal that watermarking can induce substantial degradation across multiple failure modes, including lexical corruption, hallucinated terminology, and amplified misattribution or omission of image findings. Notably, we find that the absence of domain-specific analyses, combined with aggregate metrics that miss failures inherent to clinical text, can systematically obscure practical watermark-induced degradations. Our findings establish domain-specific evaluation as a prerequisite for the safe deployment of watermarked models in medicine, where current benchmarks can otherwise mask clinically consequential failures.

Chinese Translation

大型语言模型（LLMs）正日益融入临床工作流程中，这凸显了对模型生成输出进行可靠追踪的需求，水印技术应运而生。然而，大多数水印的评估是在通用基准上进行的，导致像医学这样的小标记级扰动可能引起显著语义变化的领域未得到充分探索。在本研究中，我们首次对LLM水印如何影响医疗性能进行了严格研究，基准测试了5种水印方案在11个LLM和7个VLM上的表现，涵盖了单模态和多模态临床推理的各种任务。重要的是，我们通过引入一个经过人类专家验证的流程，系统性地审计医疗推理质量、术语精确性和诱发的幻觉，来补充现有的评估。我们的结果表明，水印可以在多个失效模式下导致显著降级，包括词汇损坏、幻觉术语以及图像发现的误归因或遗漏的放大。值得注意的是，我们发现缺乏特定领域的分析，加上忽视临床文本固有失效的汇总指标，可能系统性地掩盖水印引起的实际降级。我们的研究结果确立了特定领域评估作为在医学中安全部署水印模型的前提条件，而当前的基准可能掩盖临床上重要的失效。

View on arXiv Download PDF AI Translation

cs.AI / 3 / 2607.20463

ClickGuard: Detecting and Spoiling Clickbait News with Informativeness Measures and Large Language Models

ClickGuard：利用信息量测量和大型语言模型检测和揭穿标题党新闻

Michaluk, Wojciech, Urban, Tymoteusz, Kubita, Mateusz, Kuntur, Soveatin, Wróblewska, Anna

Abstract

This paper presents an AI-driven browser extension that identifies clickbait to help users avoid misleading Internet articles. Moving beyond traditional detection, the application employs a hybrid machine learning architecture that combines transformer-based embeddings with linguistically motivated features and a custom "baitness" score. After evaluating various natural language processing techniques -- from classic vectorizers to large language model (LLM) embeddings -- an XGBoost-based model was developed that achieves an F1-score of 91% on the open combined dataset. Most importantly, the tool can warn users before and after they access a clickbait article. After opening an article, the user receives a percentage score indicating the likelihood that it is clickbait. The prediction is explained based on the analyzed metrics, including those specifically developed within the proposed system. The browser extension also provides a clickbait spoiler -- a one- to two-sentence summary of the entire article. Demo video:https://www.youtube.com/watch?v=IJ1gkQV82C4}{https://www.youtube.com/watch?v=IJ1gkQV82C4

Chinese Translation

本文提出了一种基于人工智能的浏览器扩展，旨在识别标题党，以帮助用户避免误导性的互联网文章。该应用超越了传统的检测方法，采用了一种混合机器学习架构，将基于变换器的嵌入与语言学驱动的特征以及自定义的“诱饵性”评分相结合。在评估了多种自然语言处理技术后——从经典的向量化方法到大型语言模型（LLM）嵌入——开发了一种基于XGBoost的模型，该模型在开放的综合数据集上达到了91%的F1分数。最重要的是，该工具能够在用户访问标题党文章之前和之后发出警告。在打开文章后，用户会收到一个百分比评分，指示该文章为标题党的可能性。该预测基于分析的指标进行解释，包括在所提议的系统中专门开发的指标。该浏览器扩展还提供了标题党的剧透——对整篇文章的一到两句总结。演示视频：https://www.youtube.com/watch?v=IJ1gkQV82C4

View on arXiv Download PDF AI Translation

cs.AI / 4 / 2607.20464

Stochastic Sampling is Epistemically Shallow: The Dimensionality Gap Between Temperature Variation and Model Diversity in LLMs

随机采样在认知上是浅薄的：温度变化与大型语言模型中的模型多样性之间的维度差距

Ali, Izhar

Abstract

When a language model gives different answers on repeated runs, does that variation reveal what it does not know? Self-consistency turns the variation into a per-question uncertainty estimate via majority voting. But does the same variation reveal cross-question structure -- related questions flipping together, the way a diverse ensemble does? We compare two regimes on the same questions: one model run $100$ times at $\tau=1$ versus an ensemble of $24$ LLMs run once each at $\tau=0$. A Marchenko--Pastur random-matrix test separates signal from sampling noise on both sides. Within any single model, at most one dimension rises above noise across five families and three benchmarks (MMLU, HellaSwag, GSM8K). Across the ensemble, four eigenvalues clear the noise edge, while a matched-difficulty Bernoulli null produces at most one in $500$ Monte Carlo draws. Self-consistency gives accurate per-question uncertainty but no detectable cross-question structure; only a diverse ensemble surfaces what a model does not know.

Chinese Translation

当一个语言模型在重复运行中给出不同的答案时，这种变化是否揭示了它所不知道的内容？自一致性通过多数投票将这种变化转化为每个问题的不确定性估计。但这种变化是否揭示了跨问题的结构——相关问题一起翻转，就像一个多样化的集成模型一样？我们在相同的问题上比较了两种模式：一个模型在 $ au=1$ 下运行 $100$ 次，与一个在 $ au=0$ 下各运行一次的 $24$ 个 LLM 的集成。Marchenko–Pastur 随机矩阵测试在两侧分离信号与采样噪声。在任何单一模型中，最多只有一个维度在五个家族和三个基准（MMLU、HellaSwag、GSM8K）中高于噪声。在集成中，四个特征值超出了噪声边界，而一个匹配难度的伯努利零假设在最多 $500$ 次蒙特卡洛抽样中产生的特征值最多只有一个。自一致性提供了准确的每个问题的不确定性，但没有可检测的跨问题结构；只有多样化的集成模型能够揭示模型所不知道的内容。

View on arXiv Download PDF AI Translation

cs.AI / 5 / 2607.20466

JAXBench: Benchmarking Autonomous TPU Kernel Optimization

JAXBench：自主TPU内核优化基准测试

Tschand, Arya, Hong, Charles, Walker, Julian, Cai, Nina, Wang, Shangkun, Subramanian, Suvinay, Dev, Sundar, Reddi, Vijay Janapa, Yazdanbakhsh, Amir, Sankaran, Sethu

Abstract

Rigorous benchmarks have driven progress in autonomous GPU kernel performance optimization by establishing a shared target to hillclimb on, but no equivalent exists for TPUs. We present JAXBench, a TPU-native benchmark suite for AI-generated kernel optimization on Google Cloud TPUs. JAXBench comprises 50 JAX workloads that are both relevant and provide headroom for optimization. We extract 17 production ML operators from architectures in the public MaxText library such as Llama-3.1, DeepSeek-V3, Mixtral, Mamba-2, and AlphaFold2, and translate 33 operators from KernelBench that are validated for correctness and set with new problem sizes that achieve high TPU v6e MXU utilization. Eight of the 17 production operators ship with hand-optimized Pallas kernels from the public Tokamax library and block-size tuned to establish an expert upper-bound baseline. We evaluate four feedback-driven methods on generating candidate Pallas kernels for JAXBench. Across the full suite with Gemini 3 Flash, we find that target-specific context matters more than model scale on a sparsely-documented DSL like Pallas. Conditioning on curated TPU documentation raises per-sample correctness from 5.8% to 37.3% and solves 48 of 50 benchmarks at a 1.28x geomean speedup. Search structure yields significant gains once correctness is achieved, with Autocomp's beam-search pipeline reaching a 1.36x geomean speedup over XLA. On the 8 hand-tuned kernels, Autocomp reaches 1.60x geomean over XLA, recovering most of the 2.08x Tokamax upper bound but trailing on the specialized paged and ragged attention operators. High-quality TPU kernel optimization remains a challenging task, and we release the JAXBench benchmark, evaluation harness, and baseline results to support open source contributions.

Chinese Translation

严格的基准测试推动了自主GPU内核性能优化的进展，通过建立一个共享目标进行爬坡，但TPU尚无等效的基准。我们提出了JAXBench，这是一个针对Google Cloud TPU的TPU原生基准套件，用于AI生成的内核优化。JAXBench包含50个相关的JAX工作负载，并提供了优化的余地。我们从公共MaxText库中的架构（如Llama-3.1、DeepSeek-V3、Mixtral、Mamba-2和AlphaFold2）中提取了17个生产级机器学习操作符，并从KernelBench中翻译了33个经过验证的正确性操作符，并设置了新的问题规模，以实现高TPU v6e MXU利用率。17个生产操作符中的8个配备了来自公共Tokamax库的手动优化Pallas内核，并经过块大小调优，以建立专家级的上限基线。我们评估了四种基于反馈的方法，以生成JAXBench的候选Pallas内核。在与Gemini 3 Flash的完整套件中，我们发现特定目标的上下文在像Pallas这样文档稀疏的领域特定语言中比模型规模更为重要。基于策划的TPU文档的条件下，每个样本的正确性从5.8%提高到37.3%，并在1.28倍几何平均加速下解决了50个基准中的48个。一旦达成正确性，搜索结构带来了显著的收益，Autocomp的束搜索管道在XLA上达到了1.36倍的几何平均加速。在8个手动调优的内核上，Autocomp达到了1.60倍的几何平均加速，恢复了大部分2.08倍的Tokamax上限，但在专门的分页和不规则注意力操作符上表现不佳。高质量的TPU内核优化仍然是一项具有挑战性的任务，我们发布了JAXBench基准测试、评估工具和基线结果，以支持开源贡献。

View on arXiv Download PDF AI Translation

cs.AI / 6 / 2607.20467

DC-Leap: Training-Free Acceleration of dLLMs via Draft-Guided Contiguous Leaping Decoding

DC-Leap：通过草稿引导的连续跳跃解码实现无训练加速的 dLLMs

Jiao, Yanhua, Wu, Tianyi, Sun, Xiaoxi, Li, Yulin, Zhen, HuiLing, Qin, Libo, Hu, Baotian, Tian, Zhuotao, Zhang, Min

Abstract

While parallel decoding is central to the efficiency of Diffusion Large Language Models (dLLMs), current strategies are often hindered by overly conservative confidence thresholds. These thresholds, necessitated by the Joint Probability Dependence Error (JPDE), result in redundant denoising iterations and suboptimal inference speeds. To overcome this, we propose DC-Leap, a training-free framework that enables reliable acceleration of dLLMs in the moderate-confidence regime. DC-Leap introduces a Dynamic Contiguous Verification strategy that integrates strictly-ordered causal constraints into the parallel decoding process. By progressively validating token dependencies, this mechanism effectively neutralizes the JPDE, enabling reliable acceleration with comparable performance. Furthermore, DC-Leap incorporates the draft-guided decoding mechanism, where the draft helps extend the context by leaping forward across multiple tokens, providing look-ahead context and retaining the structural benefits of bidirectional attention during inference. Extensive experiments on standard benchmarks demonstrate that DC-Leap achieves substantial speedups, up to 53.19x on MBPP for long-sequence generation, and up to 105.02x when combined with KV-Cache with comparable generation quality. Code is available at https://github.com/ffh-wyls/DC-Leap .

Chinese Translation

尽管并行解码对扩散大语言模型（dLLMs）的效率至关重要，但当前的策略常常受到过于保守的置信阈值的限制。这些阈值是由联合概率依赖误差（JPDE）所决定，导致冗余的去噪迭代和次优的推理速度。为了解决这一问题，我们提出了 DC-Leap，一个无训练的框架，能够在中等置信度范围内可靠地加速 dLLMs。DC-Leap 引入了一种动态连续验证策略，将严格有序的因果约束整合到并行解码过程中。通过逐步验证标记依赖关系，该机制有效中和了 JPDE，从而实现了可靠的加速，并保持了相当的性能。此外，DC-Leap 还结合了草稿引导解码机制，草稿通过跨越多个标记向前跳跃来扩展上下文，提供前瞻性上下文，并在推理过程中保留双向注意力的结构优势。在标准基准测试上的大量实验表明，DC-Leap 实现了显著的加速，在 MBPP 上的长序列生成中速度提升高达 53.19 倍，并且与 KV-Cache 结合时速度提升高达 105.02 倍，同时生成质量相当。代码可在 https://github.com/ffh-wyls/DC-Leap 获取。

View on arXiv Download PDF AI Translation

cs.AI / 7 / 2607.20468

InferenceBench: A Benchmark for Open-Ended LLM Inference Optimization by AI Agents

InferenceBench：由AI代理进行开放式LLM推理优化的基准测试

Yeon, Jehyeok, Rank, Ben, Andriushchenko, Maksym

Abstract

AI agents are increasingly used to automate research and development tasks, yet existing benchmarks typically evaluate them on prescribed workflows or narrow action spaces. Even nominally open-ended tasks can often be solved by retrieving a well-known recipe and tuning a few hyperparameters, making it unclear whether strong results reflect genuine optimization or memorized solutions. We introduce InferenceBench, where an agent must deploy an OpenAI-compatible inference server and optimize the speed of LLM inference. Each agent receives a target LLM, one H100 GPU, an optimization scenario, and a wall-clock time budget of two hours. Three optimization scenarios isolate distinct bottlenecks of inference (prefill latency, decode latency, and concurrent request throughput) and a fourth balances all three at the same time. Across 15 frontier agent configurations, agents reliably improve over a naive PyTorch baseline (up to $8.08\times$) and often match or exceed serving engines with default settings ($4.05\times$ for vLLM), but still fall below a simple hyperparameter search under the same time budget (up to $11.53\times$). Qualitative analysis of agent trajectories shows that although agents enumerate many relevant optimization techniques, they overwhelmingly converge on a single inference framework. They test only a few distinct configurations and spend the remaining budget re-measuring, repairing, or optimizing hyperparameters rather than exploring substantially different strategies. This suggests the bottleneck is not domain knowledge, but the ability to propose diverse configurations, evaluate them systematically, and submit the best identified solution. Overall, InferenceBench reflects the ability of agents to operate in an open-ended AI engineering setting, where memorized solutions lead to limited improvements.

Chinese Translation

AI代理越来越多地被用于自动化研究和开发任务，但现有的基准测试通常在规定的工作流程或狭窄的行动空间中评估它们。即使是名义上开放式的任务，通常也可以通过检索一个众所周知的方案并调整几个超参数来解决，这使得强劲的结果是否反映真正的优化或是记忆化的解决方案变得不明确。我们引入了InferenceBench，在该基准中，代理必须部署一个兼容OpenAI的推理服务器并优化LLM推理的速度。每个代理接收一个目标LLM、一块H100 GPU、一个优化场景以及两个小时的实际时间预算。三个优化场景隔离了推理的不同瓶颈（预填充延迟、解码延迟和并发请求吞吐量），而第四个场景则同时平衡这三者。在15种前沿代理配置中，代理可靠地超越了一个简单的PyTorch基线（最高可达$8.08 imes$），并且通常与默认设置的服务引擎相匹配或超越（vLLM的$4.05 imes$），但在相同的时间预算下仍低于简单的超参数搜索（最高可达$11.53 imes$）。对代理轨迹的定性分析表明，尽管代理列举了许多相关的优化技术，但它们主要集中在单一的推理框架上。它们仅测试了少数不同的配置，并将剩余的预算用于重新测量、修复或优化超参数，而不是探索实质上不同的策略。这表明瓶颈并不在于领域知识，而在于提出多样化配置、系统性评估这些配置并提交最佳识别解决方案的能力。总体而言，InferenceBench反映了代理在开放式AI工程环境中操作的能力，在这种环境中，记忆化的解决方案导致了有限的改进。

View on arXiv Download PDF AI Translation

cs.AI / 8 / 2607.20469

DecodeShare: Tracing the Shared Subspace of LLM Decode-Time Decisions

DecodeShare：追踪大语言模型解码时决策的共享子空间

Shao, Zishan, Zhang, Lixun, Cui, Kangning, Wang, Yixiao, Jiang, Ting, Ye, Hancheng, Wang, Qinsi, Du, Zhixu, Fu, Yuzhe, Yang, Fan, Zhuo, Danyang, Chen, Yiran, Li, Hai Helen

Abstract

Large language models (LLMs) handle many tasks with one set of parameters, but under KV-cached inference it is unclear what task-general structure, if any, is used at decode time rather than during prefill. We propose DecodeShare, a protocol that identifies a low-dimensional subspace consistently shared across tasks in decode-time hidden states, and then tests its causal role by removing that subspace only during decoding. In our experiments, disturbing the discovered shared subspace degrades decision performance far more than disturbing either a prefill-derived or random subspace under the same intervention budget. We further show this decode-shared subspace has practical consequences for activation steering: common steering directions can overlap the task-general decode channel. Projecting out this shared subspace directly separates the functional roles of the two components, while evaluating steering vectors at decode-time yields more reliable signal for downstream deployment than prefill-based proxies. Despite its compactness, the shared subspace can serve as a high-leverage causal channel at decode time. Code is available at: https://github.com/Zishan-Shao/decodeshare.git.

Chinese Translation

大型语言模型（LLMs）使用一组参数处理许多任务，但在KV缓存推理下，尚不清楚在解码时使用了何种任务通用结构（如果有的话），而不是在预填充阶段。我们提出了DecodeShare，一种协议，用于识别在解码时隐藏状态中跨任务一致共享的低维子空间，并通过在解码过程中仅移除该子空间来测试其因果作用。在我们的实验中，干扰发现的共享子空间会显著降低决策性能，远超在相同干预预算下干扰预填充导出的或随机子空间的影响。我们进一步表明，这一解码共享子空间对激活引导具有实际影响：常见的引导方向可以与任务通用的解码通道重叠。投影出这一共享子空间可以直接分离两个组件的功能角色，而在解码时评估引导向量比基于预填充的代理提供了更可靠的下游部署信号。尽管其紧凑性，共享子空间在解码时仍可以作为一个高杠杆的因果通道。代码可在：https://github.com/Zishan-Shao/decodeshare.git获取。

View on arXiv Download PDF AI Translation

cs.AI / 9 / 2607.20470

PlanE: Meta Planning of Data, Tuning, and Inference for Extractive-based LLMs

PlanE：基于提取的语言模型的数据、调优和推理的元规划

Wang, Jiacheng, Zhang, Weiyan, Yu, Guangya

Abstract

Enhancing the task-specific capabilities of Large Language Models (LLMs) primarily requires substantial instruction-tuning datasets. However, the sheer volume of such data imposes a considerable annotation cost, and a lack of optimization methods for tailoring LLMs to specific tasks. To address the above issues, we propose a \textbf{Plan}ning framework for constructing \textbf{E}xtractive-based LLMs called \textbf{PlanE}, which includes data decomposition, instruction tuning, and prompt inference. Additionally, we introduce a Data-Tuning-Inference (DTI) planner, aimed at selecting the optimal base-LLM and its DTI combinations for specific datasets to improve construction efficiency. The experimental results demonstrate the effectiveness of our PlanE from two views: (1) across different datasets using the same base-LLM, and (2) on the same dataset using different base-LLMs. Furthermore, we validate the generalizability of the proposed DTI planner under different optimization objectives. The codes are publicly available at https://github.com/gugugu-469/PlanE.

Chinese Translation

增强大型语言模型（LLMs）的任务特定能力主要需要大量的指令调优数据集。然而，这些数据的庞大体量带来了相当高的标注成本，并且缺乏优化方法来将LLMs调整到特定任务。为了解决上述问题，我们提出了一种用于构建基于提取的LLMs的 extbf{Plan}ning框架，称为 extbf{PlanE}，该框架包括数据分解、指令调优和提示推理。此外，我们引入了一种数据-调优-推理（Data-Tuning-Inference, DTI）规划器，旨在为特定数据集选择最佳的基础LLM及其DTI组合，以提高构建效率。实验结果从两个方面证明了我们PlanE的有效性：（1）在使用相同基础LLM的不同数据集上，以及（2）在使用不同基础LLM的相同数据集上。此外，我们验证了所提出的DTI规划器在不同优化目标下的可推广性。代码已公开可用，地址为 https://github.com/gugugu-469/PlanE。

View on arXiv Download PDF AI Translation

cs.AI / 10 / 2607.20471

Benchmarking the Personalization Capabilities of Large Language Models

大型语言模型个性化能力的基准测试

Srivastava, Ashutosh, Yedlapati, Siddharth, Aggarwal, Vinay, Singla, Yaman Kumar, Dixit, Shashwat, Ajmera, Jitendra, Krishnamurthy, Balaji

Abstract

Personalization, the act of varying a message to induce action from a specific receiver while keeping sender, channel, and time fixed, has a long tradition in psychology and marketing as a two-party problem in which sender and receiver have independent objectives. Large language models remove the bounded-inventory constraint of classical retrieval-and-ranking approaches by generating a continuum of message variants conditioned on inferred receiver state, raising the question of how well current models perform personalization in the classical sense. Existing LLM personalization benchmarks measure sender-side adaptation, in which the receiver is the same user the model is serving. The two-party question, whether a generated message induces its intended action in a third party, has been investigated only through A/B tests and small-scale human studies that cannot be re-run against a new model on demand. We adapt the Bayesian Persuasion framework of Kamenica and Gentzkow (2011) to generative agents and instantiate the formulation in sales, where receiver actions are routinely logged against the outreach that induced them. We release SDR-Bench, a public corpus of 6,279 customer success stories spanning 22 industries and approximately 200 enterprises, served through a temporally constrained simulation that prevents future-data leakage. Across frontier LLMs and deep-research agents, we observe a consistent personalization plateau and on a Fortune 100 tech cohort no model statistically separates successful from unsuccessful outreach. A field deployment with 12 professional sales representatives validates the framework, with 48 percent of model-generated content rated immediately useful and senior-expert agreement at Pearson 0.82. We release SDR-Arena and SDR-Bench publicly to support reproducible study of generative personalization at scale.

Chinese Translation

个性化是指在保持发送者、渠道和时间不变的情况下，调整信息以引导特定接收者采取行动的行为。这在心理学和市场营销中有着悠久的传统，作为一个双方问题，发送者和接收者有独立的目标。大型语言模型通过生成基于推断接收者状态的连续信息变体，消除了经典检索和排序方法的库存限制，这引发了一个问题：当前模型在经典意义上的个性化表现如何。现有的LLM个性化基准测量发送者侧的适应性，其中接收者是模型所服务的同一用户。关于生成的信息是否能在第三方中引发预期的行动的双方问题，仅通过A/B测试和小规模的人类研究进行探讨，而这些研究无法按需针对新模型重新进行。我们将Kamenica和Gentzkow（2011）的贝叶斯劝说框架适应于生成代理，并在销售领域实例化该公式，其中接收者的行为会定期记录，以便与引发这些行为的外展进行对比。我们发布了SDR-Bench，这是一个包含6,279个客户成功案例的公共语料库，涵盖22个行业和约200家企业，通过一个时间受限的模拟进行服务，以防止未来数据泄漏。在前沿的LLM和深度研究代理中，我们观察到个性化表现的持续平台，而在财富100强科技公司中，没有模型在统计上区分成功和不成功的外展。与12名专业销售代表的现场部署验证了该框架，其中48%的模型生成内容被评为立即有用，资深专家的一致性达到Pearson 0.82。我们公开发布SDR-Arena和SDR-Bench，以支持大规模生成个性化的可重复研究。

View on arXiv Download PDF AI Translation

cs.AI / 11 / 2607.20472

Robust Critics: Defending LLMs Against Multi-Turn Attacks

稳健的批评者：保护大型语言模型免受多轮攻击

Belaire, Roman, Sinha, Arunesh, Varakantham, Pradeep

Abstract

When a user asks a language model something harmful, is it a genuine attack or a misunderstood but well-meaning question? This ambiguity is one of the central challenges of LLM safety. A model that assumes the worst harms legitimate users; one that assumes the best is easily exploited. The problem is compounded in multi-turn dialogue, where an attacker's true intent may only reveal itself gradually across many exchanges, yet existing safety frameworks apply a contextual bandit treatment, ignoring the trajectory of the conversation. To that end, we propose Dialogue Critic Guided Sampling (DCGS), a framework that addresses this by inferring user intent at every turn of dialogue. Instead of applying a fixed rule about what is or is not safe, DCGS learns what the user's intent is likely to be based on the full conversational history and generates responses accordingly. Formally, we model adversarial dialogue as a Markov Decision Process and learn value and regret-based critics at both the individual token and utterance (full response) levels, scoring candidate responses via an action-value critic. We prove that this inference-time reweighting approximates exponential tilting of the base policy, guaranteeing improvement in expected return for any finite candidate pool, a property that group-relative objectives do not exhibit. Evaluated on CARES-18k, WildJailbreak, Redbench, and Harmbench, DCGS outperforms strong robust baselines and frontier models on adversarial dialogue tasks. DCGS also transfers to frontier models, improving their robustness without fine-tuning.

Chinese Translation

当用户向语言模型提出有害问题时，这是一种真正的攻击，还是一个误解但出于好意的问题？这种模糊性是大型语言模型安全性面临的核心挑战之一。假设最坏情况的模型会伤害到合法用户；而假设最好情况的模型则容易被利用。在多轮对话中，这个问题更加复杂，因为攻击者的真实意图可能仅在多次交流中逐渐显露，而现有的安全框架则采用上下文赌博者的处理方式，忽视了对话的轨迹。为此，我们提出了对话批评者引导采样（Dialogue Critic Guided Sampling, DCGS），该框架通过推断每轮对话中的用户意图来解决这一问题。DCGS并不是应用固定的安全规则，而是基于完整的对话历史学习用户意图的可能性，并相应生成响应。我们正式地将对抗性对话建模为马尔可夫决策过程，并在单个标记和发言（完整响应）层面学习基于价值和遗憾的批评者，通过动作-价值批评者对候选响应进行评分。我们证明了这种推断时的重加权近似于基础策略的指数倾斜，确保在任何有限候选池中期望回报的改善，而这一特性在群体相对目标中并不存在。在CARES-18k、WildJailbreak、Redbench和Harmbench上的评估表明，DCGS在对抗性对话任务中超越了强大的稳健基线和前沿模型。DCGS还可以迁移到前沿模型，提高其稳健性而无需微调。

View on arXiv Download PDF AI Translation

cs.AI / 12 / 2607.20473

Incomplete Prompt Jailbreaks in Large Language Models

大型语言模型中的不完整提示越狱

Kim, Yeonjea, Park, Bumjin, Choi, Jaesik

Abstract

Large language models (LLMs) are increasingly released as open-weight models with safeguards against harmful requests. Nevertheless, sentence completion remains vulnerable to incomplete harmful prompts. In this work, we formalize this phenomenon as incomplete prompt jailbreaks (IPJ) and provide a systematic empirical characterization of when and how incomplete prompts elicit harmful continuations. We analyze diverse attractor types associated with incomplete sentence continuation and show that LLMs systematically delay refusal until sentence termination. We further demonstrate that training models to refuse incomplete harmful prompts via parameter tuning is insufficient, failing to generalize across both content domains and attractor types. To enable fine-grained control, we identify two functional neurons: termination and continuation neurons. By clarifying their roles in sentence completion, we highlight the potential of neuron-level interventions for more precise and robust IPJ defenses.

Chinese Translation

大型语言模型（LLMs）越来越多地以开放权重模型的形式发布，并配备了防止有害请求的安全措施。然而，句子完成仍然容易受到不完整有害提示的影响。在本研究中，我们将这一现象形式化为不完整提示越狱（IPJ），并系统性地实证描述了何时以及如何不完整提示引发有害的续写。我们分析了与不完整句子续写相关的多种吸引子类型，并表明LLMs系统性地延迟拒绝，直到句子结束。我们进一步证明，通过参数调优训练模型拒绝不完整有害提示是不够的，无法在内容领域和吸引子类型之间进行泛化。为了实现细粒度控制，我们识别了两个功能神经元：终止神经元和续写神经元。通过阐明它们在句子完成中的作用，我们强调了神经元级干预在更精确和稳健的IPJ防御中的潜力。

View on arXiv Download PDF AI Translation

cs.AI / 13 / 2607.20474

VeriSimpl: Robust Optimization Modeling from Natural Language using Simplification-based Verification

VeriSimpl：基于简化验证的自然语言鲁棒优化建模

Rahman, Sumaya Abdul, Cuellar, Seckhen Ariel Andrade, Raissov, Ghani, Raza, Mohammad

Abstract

Natural language interfaces can greatly benefit the accessibility and usability of optimization modeling, and recent advances in large language models (LLMs) show promise in automatically translating textual problem descriptions into executable solver formulations. However, a key challenge for existing approaches is to ensure that the inferred formulation correctly implements the intended task, even if it may execute without errors. We introduce VeriSimpl, a solver LLM framework for robust natural-language-to-optimization formalization. Our approach is based on the idea of simplification-based verification, where the optimization solver is leveraged to generate simplified diagnostic queries about a candidate formulation to allow the LLM to tractably reason about the correctness of the formulation with respect to the task description. We present such simplification strategies along different dimensions with respect to problem constraints and decision variables, which allow the LLM to reason locally under fixed global contexts. Evaluations on a range of optimization benchmarks show how our approach provides consistent improvements in accuracy over existing methods, while also providing a novel high-precision self-verification signal.

Chinese Translation

自然语言接口可以极大地提升优化建模的可访问性和可用性，而近期大型语言模型（LLMs）的进展显示出自动将文本问题描述转换为可执行求解器公式的潜力。然而，现有方法面临的一个关键挑战是确保推导出的公式正确地实现了预期任务，即使它可能在没有错误的情况下执行。我们引入了VeriSimpl，一个用于鲁棒自然语言到优化形式化的求解器LLM框架。我们的方法基于简化验证的思想，通过利用优化求解器生成关于候选公式的简化诊断查询，使LLM能够在固定的全局上下文下，局部推理公式相对于任务描述的正确性。我们展示了在问题约束和决策变量等不同维度上的简化策略，使LLM能够在固定的全局上下文下进行局部推理。对一系列优化基准的评估表明，我们的方法在准确性上相较于现有方法提供了一致的提升，同时也提供了一种新颖的高精度自我验证信号。

View on arXiv Download PDF AI Translation

cs.AI / 14 / 2607.20475

SonicSampler: Unified Tile-Aware Kernels for LLM Sampling and Speculative Verification

SonicSampler：用于LLM采样和推测验证的统一块感知内核

Ponnusamy, Pragaash, Sahni, Shivam, Wang, Jue, Dao, Tri

Abstract

Sampling in LLM inference comprises a combinatorial set of logit processing, token selection, and verification operations for speculative decoding. However, existing implementations either accelerate only subsets of this pipeline, rely on multiple kernel launches, or assume homogeneous sampling behavior across a batch, limiting support for dynamic serving workloads and preventing efficient CUDA Graph execution. We present $\textbf{SonicSampler}$, a unified suite of tile-aware Triton kernels that vertically fuses the complete sampling pipeline into a fixed, workload-aware execution model. Our kernels support dynamic per-request sampling behaviors, including grammar-constrained decoding, repetition, frequency and presence penalties, logit bias, temperature scaling, top-$k$ / top-$p$ / min-$p$ filtering, and speculative verification - within a single batched kernel while remaining fully CUDA Graph-compatible. Central to our approach is a novel hierarchical two-stage top-$k$ algorithm that achieves up to $\textbf{10x speedup}$ over competitive baselines and exploits the low-entropy structure of LLM outputs to enable efficient selection over large vocabularies. Across heterogeneous speculative decoding workloads, SonicSampler achieves up to $\textbf{16x speedup}$ over state-of-the-art baselines while preserving flexible batched execution.

Chinese Translation

在LLM推理中，采样包括一组组合的logit处理、标记选择和推测解码的验证操作。然而，现有的实现要么仅加速该管道的某些子集，要么依赖多个内核启动，或者假设批次中的采样行为是同质的，这限制了对动态服务工作负载的支持，并阻碍了高效的CUDA图执行。我们提出了$ extbf{SonicSampler}$，这是一个统一的块感知Triton内核套件，它将完整的采样管道垂直融合为一个固定的、工作负载感知的执行模型。我们的内核支持动态的每请求采样行为，包括语法约束解码、重复、频率和存在惩罚、logit偏置、温度缩放、top-$k$ / top-$p$ / min-$p$过滤以及推测验证——在一个单一的批处理内核中，同时保持完全的CUDA图兼容性。我们方法的核心是一个新颖的分层两阶段top-$k$算法，该算法在竞争基准上实现了高达$ extbf{10x加速}$，并利用LLM输出的低熵结构来实现对大词汇表的高效选择。在异构推测解码工作负载中，SonicSampler在保持灵活的批处理执行的同时，实现了高达$ extbf{16x加速}$，相较于最先进的基准。

View on arXiv Download PDF AI Translation

cs.AI / 15 / 2607.20476

Benchmarking Large Language Models on Multi-Sensor Physical Hazard Assessment

大型语言模型在多传感器物理危害评估中的基准测试

Iqbal, Faizan

Abstract

We present an empirical benchmark evaluating how five large language models assess multisensor physical hazard data. Testing 60 scenarios across three categories - multi-sensor joint assessment, response proportionality, and pattern disambiguation - with 1,800 API calls at temperature 0.0, we find that all tested models consistently produced no precautionary warning signal across the tested scenarios where multiple sensors are simultaneously elevated below their individual safety limits, while achieving near-perfect accuracy on single-sensor threshold violations. All five models (ChatGPT-4o, Gemini 2.5 Flash, DeepSeek, Kimi, Llama 3.1 8B) score near zero on Category A multi-sensor scenarios (Q2: 0.000-0.208; Q3: 0.000-0.592) compared to strong performance on single-sensor scenarios (Category B Q1: 0.975-1.000). Structured tabular formatting shows no consistent advantage over plain prose; ChatGPT-4o performs significantly better under prose (p = 0.001). These findings have direct implications for practitioners deploying the tested models in physical safety monitoring systems.

Chinese Translation

我们提出了一项实证基准，评估五种大型语言模型如何评估多传感器物理危害数据。在三个类别（多传感器联合评估、响应比例和模式消歧）中测试了60个场景，进行了1,800次API调用，温度设置为0.0。我们发现，在多个传感器同时在各自安全限制以下升高的测试场景中，所有测试模型一致未产生任何预警信号，而在单传感器阈值违规方面则达到了近乎完美的准确率。所有五个模型（ChatGPT-4o、Gemini 2.5 Flash、DeepSeek、Kimi、Llama 3.1 8B）在类别A的多传感器场景中得分接近零（Q2: 0.000-0.208；Q3: 0.000-0.592），而在单传感器场景（类别B Q1: 0.975-1.000）中表现强劲。结构化表格格式未显示出相较于普通文本的一致优势；ChatGPT-4o在文本格式下表现显著更好（p = 0.001）。这些发现对在物理安全监测系统中部署测试模型的从业者具有直接的影响。

View on arXiv Download PDF AI Translation

cs.AI / 16 / 2607.20477

Semi-Supervised Text-Attributed Graph Distillation

半监督文本属性图蒸馏

Lai, Yurui, Moustafa, Samir, Yang, Renchi, Chan, Tsz Nam

Abstract

{\em Text-Attributed Graphs} (TAGs) have emerged as an expressive data model for integrating graph topology with rich textual semantics. Existing representation learning methods over TAGs suffer from severe scalability bottlenecks, particularly together with {\em Large Language Models} (LLMs). While data distillation offers a promising data-centric solution, existing methods fail to capture the complex interplay between graph and text modalities, struggle with the label scarcity inherent in semi-supervised settings, and lack the ability to produce the human-readable textual attributes required for downstream LLM-based tasks. To address these challenges, we propose \algo{}, a unified semi-supervised framework guided by the {\em Wasserstein Distance} (WSD). Grounded in our empirical findings on real TAGs, \algo{} introduces a graph-text collaborative encoding module that utilizes dual-pathway encoders (graph-aware and -free) within a collaborative self-training scheme to harvest reliable pseudo-labels and fuse complementary graph-text features. Furthermore, we develop a theoretically grounded WSD-based graph sketching algorithm and a cost-effective LLM text synthesis module, which leverages cluster-based keyword extraction to generate coherent, human-readable summaries for condensed nodes. Extensive experiments on benchmark datasets demonstrate that \algo{} achieves a state-of-the-art performance-compression trade-off in terms of both GNN- and LLM-based downstream tasks, enabling effective and efficient TAG learning or analytics.

Chinese Translation

文本属性图（Text-Attributed Graphs, TAGs）作为一种表达丰富的数据模型，已成为将图拓扑与丰富文本语义相结合的重要工具。然而，现有的 TAG 表示学习方法在与大型语言模型（Large Language Models, LLMs）结合时遭遇了严重的可扩展性瓶颈。尽管数据蒸馏提供了一种有前景的数据中心解决方案，但现有方法未能捕捉图与文本模态之间复杂的相互作用，难以应对半监督设置中固有的标签稀缺问题，并且缺乏生成用于下游 LLM 任务所需的人类可读文本属性的能力。为了解决这些挑战，我们提出了 extit{algo}，一个由 Wasserstein 距离（Wasserstein Distance, WSD）指导的统一半监督框架。基于我们在真实 TAG 上的实证发现， extit{algo} 引入了一个图-文本协同编码模块，该模块在协同自我训练方案中利用双通道编码器（图感知和无图）来获取可靠的伪标签并融合互补的图-文本特征。此外，我们开发了一种理论基础的基于 WSD 的图草图算法和一种具有成本效益的 LLM 文本合成模块，该模块利用基于聚类的关键词提取生成连贯的人类可读摘要，以便于凝缩节点的表示。大量在基准数据集上的实验表明， extit{algo} 在 GNN 和 LLM 基于的下游任务中实现了性能与压缩的最佳平衡，从而有效且高效地促进了 TAG 的学习或分析。

View on arXiv Download PDF AI Translation

cs.AI / 17 / 2607.20479

Beyond Liars' Bench: The Impact of Lie Typology, Depth, and Sparsity on Deception Detection in LLMs

超越谎言的法庭：谎言类型、深度和稀疏性对大型语言模型中欺骗检测的影响

Moustafa, Amr, Feser, Max, Mai, Florian

Abstract

Training probes to detect deceptive outputs from large language models is still an open problem. Recent work has demonstrated that detection probes fail especially in out-of-domain scenarios -- training on one type of lie does not transfer well to deception scenarios involving other types of lies. In this work, we conduct a systematic study on how various factors impact detection performance: representation depth, probe expressivity, sparse feature representations, and the lie typology of the training data. To this end, we augment standard benchmark training data with a supplementary dataset containing diverse types of deception, including fabrication, omission, and exaggeration examples. Analyzing these factors across seven probe types, our experimental results show that the optimal representation depth is highly dataset-dependent, more expressive probes provide only selective gains over linear baselines, and sparse autoencoder features perform similarly to dense hidden states. Ultimately, we demonstrate that the choice of training data and lie typology substantially changes detectability, highlighting that deception detection is a highly representation-dependent problem.

Chinese Translation

训练探测器以检测大型语言模型的欺骗输出仍然是一个未解决的问题。近期的研究表明，探测器在域外场景中尤其失败——在一种类型的谎言上训练并不能很好地迁移到涉及其他类型谎言的欺骗场景中。在本研究中，我们系统地研究了各种因素如何影响检测性能：表示深度、探测器表现力、稀疏特征表示以及训练数据的谎言类型。为此，我们用一个包含多种欺骗类型的补充数据集增强了标准基准训练数据，其中包括捏造、遗漏和夸大等示例。在七种探测器类型上分析这些因素后，我们的实验结果表明，最佳表示深度高度依赖于数据集，更具表现力的探测器仅在有限的情况下优于线性基线，而稀疏自编码器特征的表现与密集隐状态相似。最终，我们证明了训练数据和谎言类型的选择显著改变了可检测性，强调了欺骗检测是一个高度依赖表示的问题。

View on arXiv Download PDF AI Translation

cs.AI / 18 / 2607.20480

Enabling Scalable Topology Inference in Distribution Systems via Constrained Multi-Source Inference

通过约束多源推理实现配电系统可扩展拓扑推断

Li, Haoran, Mai, Lihao, Guo, Muhao, Wu, Jiaqi, Weng, Yang

Abstract

Accurate distribution system topology is essential for outage localization, voltage analytics, and operation of distribution grids, yet maintaining reliable connectivity records remains challenging in practice due to heterogeneous and imperfect utility data. Existing topology identification methods often rely primarily on electrical similarity or spatial records alone, which become unreliable in dense feeders and under inconsistent metadata conditions. This paper formulates distribution topology identification as a constrained inference problem that refines a utility-provided base topology using heterogeneous evidence while enforcing spatial feasibility and physical operational constraints. Instead of reconstructing connectivity from scratch, the proposed framework detects inconsistent assignments, performs localized reconnection within constrained neighborhoods to ensure scalability, and iteratively enforces physical feasibility to produce operationally consistent topology estimates. In addition, a falsification-driven reliability metric evaluates how strongly each inferred connection is supported relative to alternative feasible assignments, enabling utilities to prioritize verification efforts while preserving system-wide observability. The framework is validated using operational data from three feeders comprising more than $8{,}000$ AMI meters in collaboration with a large U.S. utility. Results demonstrate over $95\%$ topology reconstruction accuracy while significantly reducing computational effort compared with global inference approaches. The study further shows that correlation-based methods alone produce ambiguous assignments in dense urban feeders, whereas combining electrical measurements with spatial and operational constraints enables robust and scalable topology recovery under realistic deployment conditions.

Chinese Translation

准确的配电系统拓扑对于故障定位、电压分析和配电网的运行至关重要，但由于公用事业数据的异质性和不完备性，维护可靠的连接记录在实践中仍然具有挑战性。现有的拓扑识别方法通常主要依赖于电气相似性或空间记录，这在密集馈线和不一致的元数据条件下变得不可靠。本文将配电拓扑识别表述为一个约束推理问题，该问题利用异质证据对公用事业提供的基础拓扑进行精炼，同时强制执行空间可行性和物理操作约束。提出的框架不是从头开始重建连接，而是检测不一致的分配，在约束邻域内执行局部重新连接以确保可扩展性，并迭代地强制执行物理可行性，以生成操作一致的拓扑估计。此外，基于伪造驱动的可靠性指标评估每个推断连接相对于替代可行分配的支持强度，使公用事业能够优先考虑验证工作，同时保持系统范围内的可观察性。该框架使用来自三条馈线的操作数据进行验证，这些馈线包含超过8,000个AMI计量表，并与一家大型美国公用事业公司合作。结果表明，与全球推理方法相比，拓扑重建的准确率超过95%，同时显著减少了计算工作量。研究进一步表明，仅基于相关性的方法在密集的城市馈线中会产生模糊的分配，而将电气测量与空间和操作约束结合起来则能够在现实部署条件下实现稳健且可扩展的拓扑恢复。

View on arXiv Download PDF AI Translation

cs.AI / 19 / 2607.20481

Routing Without Training: Controllable-Ratio LLM Offloading via Reliability Gating

无需训练的路由：通过可靠性门控实现可控比例的LLM卸载

Chen, Evan, Wang, Shiqiang, Chan, Kevin S, Wang, Su, Brinton, Christopher

Abstract

Local-cloud collaboration is a practical way to deploy large language models under resource constraints, but existing methods often rely on trained routers or collaboration-aware finetuning that tie routing behavior to a particular operating regime. In this work, we show that such training may be unnecessary: the local model's own inference-time agreement across sampled responses already provides a strong signal for deciding when to trust local execution and when to offload to a stronger cloud model. We propose CARGO, a training-free routing framework that estimates this agreement through prompt-varied sampling, applies Bayesian early stopping for sample-efficient uncertainty control, and supports arbitrary target collaboration ratios through lightweight deployment-time calibration. Across diverse reasoning and question-answering tasks, multiple local LLM families and scales, and both pretrained and finetuned local models, CARGO consistently outperforms other training-free baselines and in several settings surpasses supervised learned routers. These results suggest that effective and adaptable local-cloud collaboration can emerge directly from the local model's intrinsic response behavior, without requiring an additional trained router.

Chinese Translation

本地云协作是一种在资源限制下部署大型语言模型的实用方法，但现有方法往往依赖于训练过的路由器或关注协作的微调，这将路由行为与特定的操作模式绑定。在本研究中，我们表明这种训练可能是多余的：本地模型在采样响应中的推理时间一致性已经提供了一个强有力的信号，用于决定何时信任本地执行，何时卸载到更强大的云模型。我们提出了CARGO，一个无训练的路由框架，通过提示变化采样来估计这种一致性，应用贝叶斯早停以实现样本高效的不确定性控制，并通过轻量级的部署时校准支持任意目标协作比例。在多种推理和问答任务中，多个本地LLM家族和规模，以及预训练和微调的本地模型，CARGO始终优于其他无训练的基线，并在多个设置中超越了监督学习的路由器。这些结果表明，有效且可适应的本地云协作可以直接从本地模型的内在响应行为中产生，而无需额外的训练路由器。

View on arXiv Download PDF AI Translation

cs.AI / 20 / 2607.20482

PersonaTrail: Benchmarking Personalized Web Agents through Browsing Trails

PersonaTrail：通过浏览轨迹评估个性化网络代理

Yang, Seungbin, Ki, Chaewoon, Lee, Dohyun, Choo, Jaegul, Park, ChaeHun

Abstract

Recent advances in large language models have enabled web agents to autonomously execute complex tasks. In practice, users frequently provide underspecified instructions, requiring agents to infer the missing context from their raw browsing histories. Existing benchmarks fail to capture this form of personalization, as they either restrict tasks to fully explicit prompts or abstract web interaction history into simplified forms. To bridge this gap, we introduce PersonaTrail, a benchmark for personalized web agents operating in a managed open web environment. By leveraging realistic browsing trajectories as user history, PersonaTrail evaluates an agent's ability to infer user preferences and recall information from past browsing sessions. We further propose Preference-Aware Contextual Memory (PACMem), a framework that decomposes raw browsing histories into two types of structured memory: factual memories that summarize individual sessions and preference memories that distill recurring behavioral patterns. At inference time, the agent retrieves the most relevant entries from these memories to guide personalized navigation. Extensive experiments show that PACMem consistently outperforms existing memory-based baselines on both tasks.

Chinese Translation

近期大型语言模型的进展使得网络代理能够自主执行复杂任务。在实际应用中，用户经常提供不够明确的指令，这要求代理从用户的原始浏览历史中推断缺失的上下文。现有的基准测试未能捕捉这种个性化形式，因为它们要么将任务限制为完全明确的提示，要么将网络交互历史抽象为简化形式。为了解决这一问题，我们提出了PersonaTrail，一个在受控开放网络环境中运行的个性化网络代理基准。通过利用现实的浏览轨迹作为用户历史，PersonaTrail评估代理推断用户偏好和回忆过去浏览会话信息的能力。我们进一步提出了偏好感知上下文记忆（Preference-Aware Contextual Memory, PACMem）框架，该框架将原始浏览历史分解为两种类型的结构化记忆：总结个别会话的事实记忆和提炼重复行为模式的偏好记忆。在推理时，代理从这些记忆中检索最相关的条目以指导个性化导航。大量实验表明，PACMem在这两项任务上始终优于现有的基于记忆的基线。

View on arXiv Download PDF AI Translation

cs.AI / 21 / 2607.20483

Tractable Hierarchical Control of Autoregressive Language Models

可处理的自回归语言模型的层次控制

Scribner, Max, Vergari, Antonio, Belle, Vaishak

Abstract

Constraining the generation of autoregressive large language models (LLMs) is an important component of integrating language models into formal systems. In the generation of code and data for tasks like program synthesis, ensuring that language models produce syntactically valid output is a prerequisite for processing such output. These languages (such as SQL or JSON) are often designed as $LR(k)$ context-free grammars. By distilling the LLM to a tractable probabilistic model, its autoregressive generation can be steered and masked to incorporate the probability of satisfying logical constraints, ensuring high quality output that is guaranteed to be valid. This paper demonstrates that the satisfaction of any $LR(k)$ grammar of finite duration can be calculated in polynomial time, an improvement over the exponential time of applying previous methods to such grammars. This result enables efficient constraint and steering of LLM generation towards output that better satisfies formal syntactic constraints.

Chinese Translation

约束自回归大型语言模型（LLMs）的生成是将语言模型集成到形式系统中的重要组成部分。在代码和数据生成任务（如程序合成）中，确保语言模型生成语法上有效的输出是处理该输出的前提。这些语言（如 SQL 或 JSON）通常被设计为 $LR(k)$ 上下文无关文法。通过将 LLM 提炼为一个可处理的概率模型，可以引导和屏蔽其自回归生成，以纳入满足逻辑约束的概率，从而确保生成高质量且有效的输出。本文证明了任何有限持续时间的 $LR(k)$ 文法的满足性可以在多项式时间内计算，这一结果比以前方法在处理此类文法时所需的指数时间有了显著改进。该结果使得能够高效地约束和引导 LLM 的生成，以产生更好地满足形式语法约束的输出。

View on arXiv Download PDF AI Translation

cs.AI / 22 / 2607.20484

The Devil is in the Spectrum: Mitigating Representation Collapse in LLMs via Topologically Regularized Side-Path

魔鬼藏在光谱中：通过拓扑正则化侧路径减轻大型语言模型中的表示崩溃

Tao, Yiheng, Cheng, Kaiwen, Lu, Yao, Liu, Chang, Chen, Jie

Abstract

Large Language Models (LLMs) are fundamentally limited by representation collapse, a bottleneck that severely degrades long-context performance. We identify that existing approaches risk drifting into one of two pathological extremes: homogenization collapse (e.g., attention sinks causing rank deficiency) and isolation collapse (e.g., local attention causing context disconnection). Through spectral analysis of attention dynamics, we derive an intrinsic trade-off between mixing efficiency (spectral gap) and information capacity (effective rank) that standard mechanisms struggle to balance. To resolve this dilemma, we propose the Topologically Regularized Side-Path (TRSP), a non-invasive architectural intervention that achieves spectral balance. TRSP employs a parameter-free Triangular Box mechanism, scaled by a lightweight, length-aware gate, to regularize the token interaction topology. By integrating proximal coupling to preserve effective rank and distal propagation to support non-degenerate mixing, TRSP promotes a geometrically healthier transition operator without altering core attention. Experiments show significant improvements across general capabilities and long-context benchmarks. Notably, on NoLiMa at $8\times$ the training length, TRSP retains $83\%$ accuracy and surpasses the Differential Transformer and Gated Attention by approximately 30 and 50 percentage points, respectively. Code available at: https://github.com/Eziotao-tyd/TRSP.

Chinese Translation

大型语言模型（LLMs）在根本上受到表示崩溃的限制，这一瓶颈严重降低了其在长上下文中的表现。我们发现现有的方法存在漂移至两种病态极端的风险：同质化崩溃（例如，注意力沉没导致的秩缺失）和孤立崩溃（例如，局部注意力导致的上下文断连）。通过对注意力动态的谱分析，我们推导出混合效率（谱间隙）与信息容量（有效秩）之间的内在权衡，而标准机制难以平衡这一点。为了解决这一困境，我们提出了拓扑正则化侧路径（Topologically Regularized Side-Path, TRSP），这是一种非侵入性的架构干预，能够实现谱平衡。TRSP采用无参数的三角盒机制，通过轻量级的、长度感知的门控进行缩放，以正则化令牌交互的拓扑结构。通过整合近端耦合以保持有效秩和远端传播以支持非退化混合，TRSP促进了几何上更健康的过渡算子，而不改变核心注意力。实验结果显示，在一般能力和长上下文基准测试中有显著改善。值得注意的是，在NoLiMa上，训练长度为$8 imes$时，TRSP保持了$83\%$的准确率，分别超过了差异变换器（Differential Transformer）和门控注意力（Gated Attention）约30和50个百分点。代码可在以下链接获取：https://github.com/Eziotao-tyd/TRSP。

View on arXiv Download PDF AI Translation

cs.AI / 23 / 2607.20485

Expectation Alignment of Language Models for Real-World User Expectations

语言模型的期望对齐与现实用户期望

Li, Miaomiao, Wang, Yang, Liang, Bin, Liu, Shudong, Zhang, Zhiwei, Wong, Kam-Fai

Abstract

Large language models (LLMs) have demonstrated remarkable performance on standard benchmarks, yet it remains largely unexplored whether they truly meet user expectations. Existing evaluation approaches, relying on model heuristics, expert rubrics, or user simulation, fail to capture the diversity and subtlety of real human expectations, causing models to appear competent while misaligning with what users actually seek. We present the first systematic study of user expectations in real-world LLM interactions, proposing a principled procedure to extract semantically rich expectations and introducing ExpectBench, a benchmark grounded in real user expectations. Analyses reveal that current LLMs struggle to satisfy and anticipate what users hope to obtain, highlighting a fundamental source of misalignment. Building on these observations, we propose LENS, a lightweight latent expectation-aware response generation framework. LENS enables models to internalize user expectations and generate better-aligned responses, consistently improving expectation satisfaction and underscoring the importance of explicitly modeling user expectations for realistic human-AI alignment.

Chinese Translation

大型语言模型（LLMs）在标准基准测试中表现出色，但它们是否真正满足用户期望仍然 largely 未被探索。现有的评估方法依赖于模型启发式、专家评分标准或用户模拟，未能捕捉到真实人类期望的多样性和细微差别，导致模型看似具备能力，但与用户实际需求不一致。我们提出了首个系统性研究真实世界 LLM 交互中用户期望的工作，提出了一种原则性程序来提取语义丰富的期望，并引入了 ExpectBench，这是一个基于真实用户期望的基准测试。分析结果显示，当前的 LLM 在满足和预测用户希望获得的内容方面存在困难，突显了不对齐的根本来源。基于这些观察，我们提出了 LENS，一个轻量级的潜在期望感知响应生成框架。LENS 使模型能够内化用户期望并生成更好对齐的响应，持续提高期望满意度，并强调明确建模用户期望对于实现现实的人机对齐的重要性。

View on arXiv Download PDF AI Translation

cs.AI / 24 / 2607.20486

OPTScientist: Multi-Agent Discovery of Typed Optimizer Programs for Transformer Pretraining

OPTScientist：用于变换器预训练的类型优化器程序的多智能体发现

Li, Zhongzheng, Feng, Tiancan, Li, Wenhao, Ran, Qingsong, Feng, Shikun, Zhang, Xiaoyuan, Wang, Yue, Zhao, Xiaoguang

Abstract

Designing optimizers for modern deep learning remains a challenging scientific problem, requiring the joint consideration of optimization geometry, state dynamics, numerical stability, implementation constraints, and empirical generalization. Existing automated optimizer discovery methods typically search either over unconstrained code spaces or within narrowly parameterized optimizer families. The former is flexible but often produces invalid or uninterpretable programs, while the latter is stable but limits novelty. We introduce OPTScientist, a theory-guided multi-agent framework for optimizer discovery in a typed domain-specific language (DSL). OPTScientist formulates optimizer design as a constrained scientific search process, where candidate updates are expressed through direction, scaling, preconditioning, regularization, state, and grouping modules. Four role agents, Theorist, Designer, Engineer, and Reviewer, collaborate within a single orchestration loop to propose hypotheses, synthesize DSL candidates, compile and evaluate optimizers, and critique results. To overcome the limitations of a fixed search space, OPTScientist combines evolutionary search over optimizer programs with a second-stage mechanism that proposes small DSL extensions when repeated failures reveal representational bottlenecks. Using this framework, we discover RS-MR, a reduced-state matrix optimizer that improves transformer pretraining over strong baselines under our native evaluation protocol. Our results suggest a path toward automated optimizer science grounded in theory, typed programs, compiler validation, and closed-loop experimentation.

Chinese Translation

为现代深度学习设计优化器仍然是一个具有挑战性的科学问题，需要共同考虑优化几何、状态动态、数值稳定性、实现约束和经验泛化。现有的自动优化器发现方法通常在无约束的代码空间或狭窄参数化的优化器家族中进行搜索。前者灵活但常常产生无效或难以解释的程序，而后者稳定但限制了新颖性。我们提出了OPTScientist，一个基于理论指导的多智能体框架，用于在类型特定领域语言（DSL）中发现优化器。OPTScientist将优化器设计公式化为一个受限的科学搜索过程，其中候选更新通过方向、缩放、预处理、正则化、状态和分组模块进行表达。四个角色代理——理论家、设计师、工程师和审阅者——在单一的协调循环中协作，提出假设、合成DSL候选、编译和评估优化器，并批评结果。为了克服固定搜索空间的局限性，OPTScientist结合了对优化器程序的进化搜索和一个第二阶段机制，当重复失败揭示出表示瓶颈时，提出小的DSL扩展。利用该框架，我们发现了RS-MR，一种减少状态的矩阵优化器，在我们的本地评估协议下，改善了变换器预训练的表现，超越了强基线。我们的结果表明，朝着基于理论、类型程序、编译器验证和闭环实验的自动优化器科学的方向迈出了步伐。

View on arXiv Download PDF AI Translation

cs.AI / 25 / 2607.20487

Directional Hallucinations: Ideological Drift in News-Grounded LLM Question Answering

定向幻觉：基于新闻的LLM问答中的意识形态漂移

Wang, Chendi, Cunningham, Liam, Yishay, Tom, Chen, Jieying

Abstract

Large language models (LLMs) are increasingly used to answer questions about political information, including in election-adjacent information settings where factual errors and ideological distortions are high-stakes. We present a reproducible measurement framework that treats hallucinations, unsupported statements in document-grounded QA, as diagnostic signals of ideological drift. Using 21,727 expert-labeled U.S. political news articles from QBias spanning left, center, and right sources, we (i) generate an article-specific question, (ii) elicit document-grounded answers from three open-weight LLMs and one proprietary model, (iii) detect sentence-level hallucinations via reference-based comparison, (iv) classify the ideological valence of hallucinated sentences with a fine-tuned stance classifier, and (v) probe output logits to relate token-level uncertainty to hallucination and drift. Hallucination rates vary substantially across models and concentrate in contentious topics, while source-ideology differences in hallucination frequency are modest. In contrast, hallucination content exhibits robust leftward drift: a majority of hallucinated sentences are classified as left-leaning, including among hallucinations generated from right-leaning sources. Logit-level analysis shows hallucinations arise in high-entropy generation contexts, and in some models uncertainty also predicts leftward drift, consistent with an "uncertainty to guessing" mechanism. We discuss implications for auditing AI-mediated political information and for designing safeguards in election-relevant deployments.

Chinese Translation

大型语言模型（LLMs）越来越多地被用于回答有关政治信息的问题，包括在与选举相关的信息环境中，在这些环境中，事实错误和意识形态扭曲的风险很高。我们提出了一种可重复的测量框架，将幻觉（在文档基础的问答中未被支持的陈述）视为意识形态漂移的诊断信号。利用来自QBias的21,727篇专家标记的美国政治新闻文章，这些文章涵盖了左、中、右不同来源，我们（i）生成特定于文章的问题，（ii）从三个开放权重的LLM和一个专有模型中引出基于文档的答案，（iii）通过基于参考的比较检测句子级幻觉，（iv）使用微调的立场分类器对幻觉句子的意识形态倾向进行分类，以及（v）探测输出对数值以将令牌级的不确定性与幻觉和漂移相关联。幻觉率在不同模型之间变化显著，并集中在有争议的话题上，而源意识形态在幻觉频率上的差异则相对较小。相比之下，幻觉内容表现出明显的左倾漂移：大多数幻觉句子被分类为左倾，包括来自右倾来源生成的幻觉。对数值级分析表明，幻觉出现在高熵生成环境中，并且在某些模型中，不确定性也预测左倾漂移，这与“从不确定性到猜测”的机制一致。我们讨论了对审计AI介导的政治信息的影响，以及在与选举相关的部署中设计保护措施的必要性。

View on arXiv Download PDF AI Translation

cs.AI / 26 / 2607.20488

Autonomous Topology Mutation: Safe Runtime Restructuring for Multi-Agent LLM Systems with Capability, State, and Shadow Invariants

自主拓扑突变：具有能力、状态和影子不变性的多智能体大语言模型系统的安全运行时重构

Sidik, Bronislav, Levi, Chaya, Kimhi, Nizzan

Abstract

Multi-agent LLM frameworks typically fix their team topology at boot time. When an individual agent becomes overloaded at runtime, for example by mixing too many action categories, accumulating tool errors, or queueing behind too many calls, the system has no mechanism to restructure itself. We introduce Autonomous Topology Mutation (ATM), a runtime team-mutation mechanism for multi-agent LLM frameworks. ATM combines telemetry-driven overload detection with three safety invariants that gate each structural change: capability monotonicity, state-routing completeness, and shadow-before-live validation. ATM monitors a six-signal Bottleneck Index that includes queue depth, context thrash, tool-error rate, role entropy, retry-loop rate, and cross-agent wait time. When a warmup-calibrated threshold is breached for multiple consecutive ticks, ATM factorises the overloaded agent into specialised sub-agents and hot-swaps the parent into a coordinator role while preserving its external identity. State transfer is controlled by privacy-level-aware routing: each memory atom is routed only to a permitted child set, or explicitly dropped with a logged reason. No candidate topology receives live traffic until it has passed a shadow validation window. On 720 DeepSeek-V3-driven task runs with deterministic tool stubs across four ablation conditions and three workloads, the ATM factoriser split lifts code-task success from 3.3% to 61.7%. The full rail-and-distillation system reduces detected high-privacy memory exposure under a regex classifier from 2.0 to 0.0 events per task while preserving task quality. The runtime rails carrying ATM's invariants add less than 500 microseconds of p99 latency on the agent hot path. A small live-tool probe with real Python execution is included as an external-validity check. The implementation, benchmark harness, and traces are open-sourced.

Chinese Translation

多智能体大语言模型框架通常在启动时固定其团队拓扑。当个别智能体在运行时过载，例如由于混合过多的动作类别、累积工具错误或在过多调用后排队时，系统没有机制进行自我重构。我们提出了自主拓扑突变（Autonomous Topology Mutation, ATM），这是一种用于多智能体大语言模型框架的运行时团队突变机制。ATM结合了基于遥测的过载检测与三种安全不变性，这些不变性对每个结构变化进行限制：能力单调性、状态路由完整性和影子优先于实时验证。ATM监控一个六信号瓶颈指数，其中包括队列深度、上下文抖动、工具错误率、角色熵、重试循环率和跨智能体等待时间。当多个连续时刻超过热身校准阈值时，ATM将过载的智能体分解为专门的子智能体，并将父智能体热插拔为协调者角色，同时保留其外部身份。状态转移由隐私级别感知路由控制：每个记忆原子仅路由到允许的子集，或明确丢弃并记录原因。没有候选拓扑在通过影子验证窗口之前接收实时流量。在720个基于DeepSeek-V3的任务运行中，针对四种消融条件和三种工作负载，ATM分解器将代码任务成功率从3.3%提升至61.7%。完整的轨道和蒸馏系统在正则表达式分类器下将检测到的高隐私内存暴露从每个任务2.0事件减少到0.0事件，同时保持任务质量。承载ATM不变性的运行时轨道在智能体热路径上增加的延迟不到500微秒的p99。包括一个小型实时工具探针，具有真实的Python执行，作为外部有效性检查。该实现、基准测试工具和跟踪数据已开源。

View on arXiv Download PDF AI Translation

cs.AI / 27 / 2607.20489

EvoSQL: Memory-Augmented Critic-Generator Co-Evolution for Text-to-SQL

EvoSQL：用于文本到 SQL 的记忆增强型评论生成器共进化

Zhou, Jiawei, Wang, Jianwei, Zhou, Chenyu, Shi, Chaojian, Dong, Ming, Wang, Kai

Abstract

Text-to-SQL has advanced rapidly with large language models, but complex database queries still require reasoning beyond one-shot generation, including multi-step decomposition, execution-based diagnosis, and targeted correction. We present EvoSQL, a co-evolution framework that formulates SQL synthesis as an iterative interaction between a generator and a critic. EvoSQL maintains a contextualized candidate memory, verifies SQL candidates with both execution signals and LLM-based critique, and updates its memory through utility-guided aggregation. To strengthen the underlying generator-critic pair, we further introduce a Self-Distillation Policy Optimization (SDPO) fine-tuning stage that injects execution-aware supervision into modern coding LLM backbones. Experiments on Spider and BIRD show that EvoSQL consistently improves open-source models over Maj@16 baselines, with particularly large gains on BIRD-Dev, ranging from +1.37% for Qwen3-4B to +9.19% for Qwen2.5-Coder-3B. SDPO initialization further improves selected backbones on Spider-Test and BIRD-Dev. These results suggest that memory-grounded co-evolution is an effective path toward more reliable and generalizable Text-to-SQL systems. Code is available at https://github.com/valleysprings/EvoSQL.

Chinese Translation

文本到 SQL 的转换在大型语言模型的推动下迅速发展，但复杂的数据库查询仍然需要超越一次性生成的推理，包括多步骤分解、基于执行的诊断和有针对性的修正。我们提出了 EvoSQL，这是一种将 SQL 合成形式化为生成器和评论者之间迭代交互的共进化框架。EvoSQL 维护一个上下文化的候选记忆，通过执行信号和基于 LLM 的评论验证 SQL 候选，并通过效用引导的聚合更新其记忆。为了增强基础的生成器-评论者对，我们进一步引入了自蒸馏策略优化 (Self-Distillation Policy Optimization, SDPO) 微调阶段，将执行感知的监督注入现代编码 LLM 主干。针对 Spider 和 BIRD 的实验表明，EvoSQL 在 Maj@16 基准上持续提升开源模型，特别是在 BIRD-Dev 上取得了显著提升，从 Qwen3-4B 的 +1.37% 到 Qwen2.5-Coder-3B 的 +9.19%。SDPO 初始化进一步改善了 Spider-Test 和 BIRD-Dev 上选定主干的表现。这些结果表明，基于记忆的共进化是实现更可靠和可泛化的文本到 SQL 系统的有效途径。代码可在 https://github.com/valleysprings/EvoSQL 获取。

View on arXiv Download PDF AI Translation

cs.AI / 28 / 2607.20490

CRAWO: Custom Resources for Adaptive Workload Orchestration

CRAWO：自定义资源用于自适应工作负载编排

Santos, Eugênio, Maia, Daniel, Loss, Stefano, Silva, José Manoel, Neto, Aluizio Rocha, Batista, Thais, Cavalcante, Everton, Cacho, Nélio, Nogueira, Eduardo, Araújo, Daniel, Lopes, Frederico

Abstract

Edge Intelligence has emerged as a key paradigm for enabling real-time applications in smart cities by shifting computation from centralized cloud data centers to the network edge, thereby reducing latency and bandwidth consumption. However, deploying Artificial Intelligence (AI) pipelines across heterogeneous edge infrastructures remains challenging due to the wide range of device capabilities, from low-power microcontrollers to accelerator-equipped systems. Existing edge orchestration platforms primarily focus on deployment automation and infrastructure management, but these approaches are often inefficient and limit the ability to adaptively allocate resources under dynamic conditions. To tackle these issues, this paper introduces CRAWO (Custom Resources for Adaptive Workload Orchestration), an architectural framework for coordinating AI pipelines across distributed edge environments. CRAWO follows a control-loop-based model that separates allocation intelligence from execution by managing placement decisions, state management, and inter-stage data flows while instantiating services on edge nodes. The framework incorporates a hardware-aware allocator with a pluggable multi-criteria decision layer that leverages real-time infrastructure metrics to enable adaptive workload placement. The reference implementation adopts a microservices architecture deployed on a lightweight Kubernetes distribution (K3s), using Custom Resource Definitions (CRDs) for domain modeling and a dedicated operator for state reconciliation. Evaluation in a vehicle surveillance scenario using license plate recognition demonstrates improved workload distribution and reduced reliance on centralized cloud processing in latency-sensitive environments.

Chinese Translation

边缘智能已成为支持智能城市实时应用的关键范式，通过将计算从集中式云数据中心转移到网络边缘，从而降低延迟和带宽消耗。然而，由于设备能力范围广泛，从低功耗微控制器到配备加速器的系统，在异构边缘基础设施上部署人工智能（AI）管道仍然面临挑战。现有的边缘编排平台主要集中在部署自动化和基础设施管理上，但这些方法往往效率低下，并限制了在动态条件下自适应分配资源的能力。为了解决这些问题，本文提出了CRAWO（自定义资源用于自适应工作负载编排），这是一个用于协调分布式边缘环境中AI管道的架构框架。CRAWO遵循基于控制回路的模型，通过管理放置决策、状态管理和阶段间数据流，同时在边缘节点上实例化服务，将分配智能与执行分离。该框架结合了一个硬件感知的分配器和一个可插拔的多标准决策层，利用实时基础设施指标实现自适应工作负载放置。参考实现采用在轻量级Kubernetes发行版（K3s）上部署的微服务架构，使用自定义资源定义（CRDs）进行领域建模，并使用专用操作员进行状态协调。在使用车牌识别的车辆监控场景中的评估表明，工作负载分配得到了改善，并减少了在延迟敏感环境中对集中式云处理的依赖。

View on arXiv Download PDF AI Translation

cs.AI / 29 / 2607.20491

DFAH-Bench: Benchmarking Observable Agent Instability in Financial Decision-Making

DFAH-Bench：金融决策中可观察代理不稳定性的基准评估

Khatchadourian, Raffi

Abstract

Standard evaluation benchmarks measure what a tool-using agent decides, not whether it arrives at that decision through the same process each time. We introduce DFAH-Bench, a replay benchmark that measures observable behavioral instability in financial agent decision-making across three channels -- tool-call trajectories, evidence contacts, and decision concentration -- none of which require access to hidden reasoning text. Across 8,127 replay episodes spanning 10 models and 3 financial tasks, we find that outcome agreement alone is an incomplete stability signal: frontier models can agree on decisions 95% of the time while following the same tool path only 77% of the time -- an 18-percentage-point gap (95% CI: [0.14, 0.22]) that outcome-only evaluation misses entirely. Among frontier-model case groups with high decision agreement, over 55% exhibit meaningful trajectory divergence. We identify three behavioral profiles: pattern matchers that achieve near-perfect agreement by collapsing to a single output regardless of input, stable executors with relatively consistent tool-use processes, and trajectory divergers that reach the same conclusions through materially different tool paths and evidence contacts. The benchmark code, metric scripts, replay logs, benchmark card, dataset README, and release manifest are released in the accompanying repository.

Chinese Translation

标准评估基准测量工具使用代理的决策，而不是其是否每次都通过相同的过程得出该决策。我们引入了DFAH-Bench，一个重放基准，测量金融代理决策中的可观察行为不稳定性，涵盖三个渠道——工具调用轨迹、证据接触和决策集中度——这些都不需要访问隐藏的推理文本。在涵盖10个模型和3个金融任务的8,127个重放回合中，我们发现仅靠结果一致性是一个不完整的稳定性信号：前沿模型在95%的情况下可以达成一致决策，但仅在77%的情况下遵循相同的工具路径——这一18个百分点的差距（95%置信区间：[0.14, 0.22]）在仅基于结果的评估中完全被忽视。在高决策一致性的前沿模型案例组中，超过55%表现出显著的轨迹分歧。我们识别出三种行为特征：模式匹配者通过无论输入如何都收敛到单一输出而实现近乎完美的一致性，稳定执行者具有相对一致的工具使用过程，以及轨迹分歧者通过实质上不同的工具路径和证据接触得出相同结论。基准代码、度量脚本、重放日志、基准卡、数据集自述文件和发布清单已在随附的代码库中发布。

View on arXiv Download PDF AI Translation

cs.AI / 30 / 2607.20493

Attention-based Experience Replay Framework for Continual Learning of Agnostic Time Series Forecasting Models

基于注意力的经验重放框架用于无关时间序列预测模型的持续学习

Besnard, Quentin, Ragot, Nicolas

Abstract

Deep learning has led to remarkable progress in artificial intelligence, particularly in robotics, imaging and sound processing. However, a major limitation of neural networks remains their strong dependence on large and stationary datasets. In many real-world applications, these conditions are rarely met due to evolving and dynamic environments where data distributions change over time. Continual learning aims to address this challenge by developing models capable of adapting incrementally while maintaining a balance between stability and plasticity under computational constraints. In this work, we introduce a novel framework for continual time series forecasting, designed to extend existing static forecasting models commonly used in the literature by incorporating an Experience Replay strategy guided by Attention mechanisms. This approach allows the model to adapt dynamically to new contexts while preserving prior knowledge, effectively mitigating catastrophic forgetting. The framework is evaluated on standard forecasting benchmarks as well as on a piezometric dataset exhibiting diverse temporal behaviors. Results show that our approach effectively increases or maintains predictive performance over time while reducing retraining costs and data requirements, thus facilitating the deployment of forecasting models in dynamic and real-world settings.

Chinese Translation

深度学习在人工智能领域取得了显著进展，特别是在机器人技术、图像处理和声音处理方面。然而，神经网络的一个主要限制是其对大型和静态数据集的强依赖。在许多现实世界的应用中，由于数据分布随时间变化的动态环境，这些条件很少得到满足。持续学习旨在通过开发能够在计算约束下逐步适应的模型来应对这一挑战，同时保持稳定性与可塑性之间的平衡。在本研究中，我们提出了一种新的持续时间序列预测框架，旨在通过结合基于注意力机制的经验重放策略，扩展文献中常用的现有静态预测模型。这种方法使模型能够动态适应新环境，同时保留先前的知识，有效减轻灾难性遗忘。该框架在标准预测基准和展现多样时间行为的压电数据集上进行了评估。结果表明，我们的方法有效地提高或维持了预测性能，同时降低了再训练成本和数据需求，从而促进了预测模型在动态和现实世界环境中的部署。

View on arXiv Download PDF AI Translation

cs.AI / 31 / 2607.20494

Isolating LLM Alignment from Regex: Zero Coverage and Metric-Dependent Divergence Under Adversarial Mutation

将 LLM 对齐与正则表达式隔离：在对抗性变异下的零覆盖率和度量依赖性差异

Maiorano, Alexandre Cristovão

Abstract

Production LLM applications commonly stack a regex filter in front of model-side alignment; prior work found no measurable coverage gain from adding a live Gemini backend behind an active regex filter. We ask whether that ceiling holds when the corpus is \emph{designed to bypass the regex}. We introduce $L_5$-no-regex -- identical to $L_4$-real (Gemini-2.5-flash, token-budget cap, rate limit, output scrub) but with the nine-pattern filter disabled -- and evaluate it against $N{=}45$ adversarial probes across three sub-corpora (carry-forward, regex-bypass, alignment-isolate), amplified by Gemini paraphrase and PAIR to ${\sim}1{,}555$ probe-run pairs over $N{=}5$ replications. Under the primary substring classifier, H1 is refuted: $L_5$ block rate is $0,%$ across all five OWASP LLM Top-10 categories ($\Delta\text{pp}{=}0$ vs.\ $L_0$, $p{=}1.00$; Wilson upper bound ${<}5,%$). A secondary LLM-judge metric on PAIR variants shows $56$--$100,%$ block rates ($p{<}0.01$), revealing alignment does respond to adversarially-framed probes -- but produces refusals too nuanced for substring matching. The sub-corpus differential prediction is not supported ($p{=}1.00$). Alignment's contribution is \emph{metric-dependent}: on natural-language harmful-request probes, it adds zero observed coverage beyond the regex; on adversarially-framed variants, an LLM judge detects refusals the substring classifier misses. The locked corpus, mutation artifacts, and export scripts are released for replication.

Chinese Translation

生产 LLM 应用通常在模型侧对齐前堆叠一个正则表达式过滤器；先前的研究发现，在活跃的正则表达式过滤器后添加一个实时的 Gemini 后端并没有带来可测量的覆盖率提升。我们询问当语料库 extit{旨在绕过正则表达式}时，这一上限是否依然成立。我们引入 $L_5$-no-regex——与 $L_4$-real（Gemini-2.5-flash、令牌预算上限、速率限制、输出清理）相同，但禁用了九种模式过滤器——并在三个子语料库（carry-forward、regex-bypass、alignment-isolate）上评估其对 $N{=}45$ 个对抗性探针的表现，通过 Gemini 释义和 PAIR 扩大至约 ${ ilde}1{,}555$ 个探针运行对，进行了 $N{=}5$ 次重复实验。在主要子字符串分类器下，H1 被驳斥：$L_5$ 的阻断率在所有五个 OWASP LLM 前十类别中均为 $0 ext{%}$（$ ext{pp} ext{差异}{=}0$ 对比 $L_0$，$p{=}1.00$；Wilson 上限 ${<}5 ext{%}$）。对 PAIR 变体的次要 LLM-judge 度量显示 $56$--$100 ext{%}$ 的阻断率（$p{<}0.01$），揭示对齐确实对以对抗性框架提出的探针做出反应——但产生的拒绝反应过于细微，无法通过子字符串匹配来捕捉。子语料库的差异预测未得到支持（$p{=}1.00$）。对齐的贡献是 extit{度量依赖性的}：在自然语言有害请求探针上，它在正则表达式之外没有增加任何观察到的覆盖率；在以对抗性框架提出的变体上，LLM judge 检测到的拒绝反应是子字符串分类器所遗漏的。锁定的语料库、变异伪影和导出脚本已发布以供复制。

View on arXiv Download PDF AI Translation

cs.AI / 32 / 2607.20495

Workload-Aware Caching for Multi-Agent Systems

面向工作负载的多智能体系统缓存

Mohamed, Anas, Haque, Kaizan, Khan, Azal Ahmad, Sharma, Chetan, Ge, Shuwen, Anwar, Ali

Abstract

Multi-agent systems decompose complex tasks into directed acyclic graphs (DAGs) of specialized agent executions, creating natural opportunities for caching intermediate results across queries. However, existing cache eviction policies treat all cached entries uniformly based on access history, ignoring structural and workload signals uniquely available in agentic execution environments. We present a workload-aware eviction policy that combines three signals, namely recomputation cost, DAG dependency count, and agent invocation frequency, into a unified scoring function that retains the most valuable entries under memory constraints. Evaluated across three multi-agent benchmarks spanning diverse reuse regimes, our policy reduces latency by up to 64.7% relative to the uncached baseline and achieves on average a 31.1% latency reduction over the next best finite-capacity baseline, while approaching the performance of an unbounded cache and maintaining accuracy on par with or exceeding all competing finite-capacity methods. We further show that workload-aware content caching is complementary to other agentic system optimization methods, including plan-level caching and parallel agent execution, with each technique targeting a distinct efficiency bottleneck in multi-agent pipelines.

Chinese Translation

多智能体系统将复杂任务分解为有向无环图（DAG）形式的专门智能体执行，从而为跨查询缓存中间结果创造了自然的机会。然而，现有的缓存驱逐策略基于访问历史对所有缓存条目进行统一处理，忽视了在智能体执行环境中独特可用的结构和工作负载信号。我们提出了一种面向工作负载的驱逐策略，将重计算成本、DAG依赖计数和智能体调用频率这三种信号结合成一个统一的评分函数，以在内存限制下保留最有价值的条目。在涵盖多种重用模式的三个多智能体基准测试中进行评估，我们的策略相较于未缓存基线将延迟降低了高达64.7%，并且在与下一个最佳有限容量基线相比时平均实现了31.1%的延迟降低，同时接近无限缓存的性能，并保持与所有竞争有限容量方法相当或更高的准确性。我们进一步表明，面向工作负载的内容缓存与其他智能体系统优化方法（包括计划级缓存和并行智能体执行）是互补的，每种技术针对多智能体管道中的不同效率瓶颈。

View on arXiv Download PDF AI Translation

cs.AI / 33 / 2607.20497

From Errors to Rules: Iterative Prompt Optimization for Text Classification

从错误到规则：文本分类的迭代提示优化

Cui, Yueying, Xue, Renhao, Zhang, Yi, Prasad, Mukul

Abstract

Prompt optimization for text classification spans diverse approaches, from demonstration selection to exploration-based search to error-driven diagnosis, each with known but incompletely characterized strengths and limitations. We conduct a comprehensive empirical study across diverse classification benchmarks (2 to 150 classes) comparing these paradigms through both quantitative evaluation and qualitative analysis of optimization traces, revealing that each paradigm excels on structurally different task types and that no single method dominates. Guided by these insights, we propose Error-Guided Optimization (ERGO), an error-driven method that iterates over the full training set in non-overlapping batches, diagnoses classification failures, and generates targeted decision rules through a diagnose-prescribe-rewrite feedback loop. ERGO achieves the best accuracy on tasks where errors concentrate in specific confused label pairs (which we term boundary-learnable tasks): TREC: 90.0%, CLINC150: 94.4%, converges in 3-5 iterations, and produces interpretable decision rules. While ERGO does not achieve the highest overall average, it fills a complementary role: demonstration-based ICL wins on coverage-dependent tasks, exploration-based search wins on many-class intent, and ERGO wins where decision boundaries are learnable from error patterns. We provide a complementarity framework linking task characteristics to optimal paradigm selection, offering practical guidance for practitioners.

Chinese Translation

文本分类的提示优化涵盖了多种方法，从示例选择到基于探索的搜索，再到基于错误的诊断，每种方法都有已知但尚未完全表征的优缺点。我们在多样的分类基准（2到150个类别）上进行了一项全面的实证研究，通过定量评估和优化轨迹的定性分析比较这些范式，揭示了每种范式在结构上不同的任务类型中表现优异，且没有单一方法占据主导地位。基于这些见解，我们提出了错误引导优化（Error-Guided Optimization, ERGO），这是一种基于错误的方法，通过非重叠批次迭代整个训练集，诊断分类失败，并通过诊断-规定-重写的反馈循环生成针对性的决策规则。ERGO在错误集中在特定混淆标签对的任务上（我们称之为边界可学习任务）实现了最佳准确率：TREC: 90.0%，CLINC150: 94.4%，在3-5次迭代中收敛，并生成可解释的决策规则。尽管ERGO未能达到最高的整体平均水平，但它填补了互补的角色：基于示例的ICl在依赖覆盖的任务上表现优异，基于探索的搜索在多类别意图上表现良好，而ERGO在可以从错误模式中学习决策边界的任务中表现突出。我们提供了一个互补性框架，将任务特征与最佳范式选择联系起来，为从业者提供实用指导。

View on arXiv Download PDF AI Translation

cs.AI / 34 / 2607.20498

AISE-Bench: A Full-Cycle Curated Benchmark for Information Seeking on Academic Knowledge Graphs

AISE-Bench：一个针对学术知识图谱的信息检索全周期精心设计的基准

Zhang, Fanjin, Wang, Zhengyang, Huang, Ruixuan, Zhang, Kefan, Xin, Amy, Wang, Yuanchun, Zhao, Shu, Kharlamov, Evgeny, Tang, Jie, Li, Juanzi

Abstract

Large language models (LLMs) augmented with tools are emerging as autonomous agents capable of using Web engine, APIs, and code to solve complex, long-horizon tasks. Current tool-using benchmarks for information seeking on academic graphs rely on synthetic templates, simplified solution spaces, or narrow tasks such as paper-centric tasks, leaving key challenges underexplored - realistic user intent, complex multi-step API planning, rich parameter filling for APIs, grounded answers with references, and comprehensive evaluation of both the process and the outcome. We introduce AISE-Bench, a real-world, full-cycle annotated benchmark for information seeking on academic knowledge graphs. AISE-Bench release contains 1,133 QA pairs, including query taxonomies, full API execution trajectories, validated parameters, and source-grounded answers with reference links. To support high-quality annotation, we design a customized agent workflow to enable annotators to plan, execute, and revise complex API workflows efficiently. We develop a comprehensive evaluation protocol measuring answer quality, reference grounding, API-planning correctness, and execution success. Among the 14 evaluated methods, even the strongest model (PLAY2PROMPT with Gemini-3-Pro) achieves only moderate performance and often struggles with API planning and execution. AISE-Bench establishes a challenging new testbed for quantitatively evaluating and improving the stepwise correctness, grounded summarization, and traceable reasoning of multi-step API-using LLM agents. Our code and data are available at https://aise-bench.github.io/.

Chinese Translation

增强工具的大型语言模型（LLMs）正在成为能够利用网络引擎、API和代码解决复杂的长期任务的自主代理。目前针对学术图谱的信息检索的工具使用基准依赖于合成模板、简化的解决空间或狭窄的任务（如以论文为中心的任务），导致一些关键挑战未得到充分探讨——如现实用户意图、复杂的多步骤API规划、API的丰富参数填充、带有参考文献的有根据的答案，以及对过程和结果的全面评估。我们介绍了AISE-Bench，这是一个针对学术知识图谱的信息检索的真实世界全周期注释基准。AISE-Bench的发布包含1,133对问答对，包括查询分类、完整的API执行轨迹、验证的参数，以及带有参考链接的源基础答案。为了支持高质量的注释，我们设计了一个定制的代理工作流程，使注释者能够高效地规划、执行和修订复杂的API工作流程。我们开发了一个全面的评估协议，衡量答案质量、参考基础、API规划的正确性和执行成功率。在14种评估方法中，即使是最强的模型（PLAY2PROMPT与Gemini-3-Pro）也仅取得中等表现，并且在API规划和执行方面常常面临困难。AISE-Bench建立了一个具有挑战性的新的测试平台，用于定量评估和改进多步骤API使用的LLM代理的逐步正确性、有根据的总结和可追溯的推理。我们的代码和数据可在https://aise-bench.github.io/获取。

View on arXiv Download PDF AI Translation

cs.AI / 35 / 2607.20499

ExecuGraph: A Multi-Agent, Execution-Grounded Framework for Reliable Backend Code Synthesis with Large Language Models

ExecuGraph：一个基于执行的多智能体框架，用于与大型语言模型可靠的后端代码合成

Lekkala, Sai Deekshith, Appadurai, Jothi Prabha, Bellibatlu, Rohith Reddy, Singh, Manpreet

Abstract

Large Language Models generate plausible backend code, but a single-pass paradigm provides no guarantee of correctness or runtime reliability. We present ExecuGraph, a multi-agent framework that places execution-based validation at the center of backend code synthesis. Six specialized agents (Planner, Code Generator, Logical Reviewer, Evaluator, Optimizer, and Explainer) are coordinated by a typed directed workflow with a bounded retry budget, implemented on LangGraph with locally hosted models (Ollama) and an optional retrieval layer for algorithmic technique recall. A subprocess-isolated sandbox with a wall-clock timeout guards every evaluation. We evaluate on a curated 30-problem DSA suite (internal-30), HumanEval (n=64), and an APPS-introductory subset, contrasting ExecuGraph against a single-agent one-shot baseline and a single-agent execution-retry baseline (a Reflexion-style ablation that isolates the contribution of multi-agent decomposition). On internal-30, the three conditions are statistically indistinguishable (n=30; paired Wilcoxon p=0.59 MF vs. SO, p=0.08 SR vs. SO); 95% bootstrap confidence intervals on all pairwise mean differences include zero. On HumanEval, multi-full edges ahead by +3.1 pp. The strongest signal is cross-model: with DeepSeekCoder V2 Lite, graph-category accuracy improves from 57.5% (oneshot) to 80.0% (multi-full), a +22.5 pp jump that supports a scaling hypothesis: the value of multi-agent decomposition grows with base-model capability. The framework's primary contribution is methodological: a single codebase that collapses by configuration into one-shot, execution-retry, and per-agent ablation conditions, enabling controlled measurement of each lever's marginal contribution. A per-agent ablation, retry-budget sweep, error-class taxonomy, and test-source audit are reported.

Chinese Translation

大型语言模型能够生成合理的后端代码，但单次生成的范式无法保证正确性或运行时的可靠性。我们提出了ExecuGraph，一个将基于执行的验证置于后端代码合成中心的多智能体框架。六个专门的智能体（规划者、代码生成器、逻辑审查者、评估者、优化器和解释者）通过一个类型化的有向工作流进行协调，该工作流具有有限的重试预算，基于LangGraph实现，使用本地托管的模型（Ollama）和一个可选的检索层用于算法技术的回忆。每次评估都在一个与子进程隔离的沙箱中进行，并设有墙钟超时限制。我们在一个精心挑选的30题数据结构与算法（DSA）套件（internal-30）、HumanEval（n=64）和一个APPS引导子集上进行评估，将ExecuGraph与单智能体一次性基线和单智能体执行重试基线进行对比（这是一个Reflexion风格的消融实验，旨在孤立多智能体分解的贡献）。在internal-30上，三种条件在统计上没有显著差异（n=30；配对Wilcoxon p=0.59 MF vs. SO, p=0.08 SR vs. SO）；所有成对均值差异的95%自助置信区间均包含零。在HumanEval上，多全连接边领先+3.1个百分点。最强的信号是跨模型的：使用DeepSeekCoder V2 Lite时，图类别的准确率从57.5%（一次性）提高到80.0%（多全连接），这是一个+22.5个百分点的跃升，支持了一个扩展假设：多智能体分解的价值随着基础模型能力的增强而增长。该框架的主要贡献在于方法论：一个单一的代码库，通过配置可以折叠为一次性、执行重试和每个智能体的消融条件，从而实现对每个杠杆边际贡献的控制测量。报告了每个智能体的消融、重试预算扫描、错误类别分类法和测试源审计。

View on arXiv Download PDF AI Translation

cs.AI / 36 / 2607.20500

FlowEdit: Information-Theoretic Control of LLM Reasoning Flows for Ill-posed Problems Involving Conflicts

FlowEdit：针对涉及冲突的病态问题的信息论控制大语言模型推理流

Tang, Sizhe, Jiang, Guangyu, Li, Yu, Chen, Rongqian, Kevrekidis, Ioannis G., Lan, Tian

Abstract

Large Language Models (LLMs) perform strongly on well-specified reasoning tasks with a feasible answer. However, problems encountered in the open world can become ill-posed due to inconsistent conditions, conflicting statements, or mutually incompatible requirements, admitting no valid responses. We argue that reasoning of such ill-posed problems involving conflicts require novel LLM capabilities to make hidden conflicts explicit, maintain competing hypotheses via multiple reasoning branches, and generate alternative responses in a single pass, all of which are challenging due to the limitation of the next-token prediction mechanism in LLMs. To this end, we propose FlowEdit, a novel framework that leverages information-theoretic principles to quantify and regulate internal reasoning flows of LLMs, for generating a full set of alternative responses under valid hypotheses. FlowEdit can be viewed as enforcing a branch-aware reasoning process using two dual information-theoretic objectives on the model's internal reasoning representations: maximizing the information flow from each selected hypothesis to the branch outcome, while minimizing the overlap and conditional dependence across sibling branches, to provide a diverse, informative set of responses with broad coverage. We show that this is achieved through tractable variational bounds under boundary embeddings being {\epsilon}-sufficient, optimizing the underlying conditional mutual information in LLM reasoning process. Extensive experiments demonstrate that FlowEdit outperforms leading proprietary models, improving exact-set-match accuracy by 68%, while boosting overall response informativeness by 24%. We further show that flow regulation surfaces in the token stream as a redistribution of next-token entropy that concentrates inside each branch, amplifies at flow boundaries, and scales with the number of flows the problem requires.

Chinese Translation

大型语言模型（LLMs）在具有可行答案的明确推理任务中表现出色。然而，在开放世界中遇到的问题可能由于不一致的条件、相互冲突的陈述或相互不兼容的要求而变得病态，导致没有有效的响应。我们认为，涉及冲突的此类病态问题的推理需要大语言模型具备新的能力，以使隐藏的冲突显性化，通过多个推理分支维持竞争假设，并在单次推理中生成替代响应，这些都是由于大语言模型的下一个标记预测机制的局限性而面临的挑战。为此，我们提出了FlowEdit，一个新颖的框架，利用信息论原理来量化和调节大语言模型的内部推理流，以在有效假设下生成一整套替代响应。FlowEdit可以被视为通过对模型内部推理表示施加两个对偶的信息论目标来强制执行一种关注分支的推理过程：最大化从每个选定假设到分支结果的信息流，同时最小化兄弟分支之间的重叠和条件依赖，以提供多样化、信息丰富且覆盖面广的响应集。我们展示了这一点是通过在边界嵌入下实现可处理的变分界限来实现的，该界限是{ extepsilon}-充分的，优化大语言模型推理过程中的潜在条件互信息。大量实验表明，FlowEdit的表现优于领先的专有模型，准确集匹配率提高了68%，同时整体响应的信息性提升了24%。我们进一步展示，流的调节在标记流中表现为下一个标记熵的重新分配，这种熵集中在每个分支内部，在流边界处放大，并随着问题所需流的数量而扩展。

View on arXiv Download PDF AI Translation

cs.AI / 37 / 2607.20501

MKEvolve: A Modular Multi-Agent Framework for Kernel Code Generation

MKEvolve：一种模块化的多智能体框架用于内核代码生成

Yoo, Jason, Saha, Rajarshi, Zhu, Shaowei, Yu, Tao, Tang, Wei, Park, Youngsuk

Abstract

Despite rapid progress in LLM-based code generation, writing correct and performant kernels for hardware accelerators remains a key bottleneck in scaling modern ML workloads. We present MKEvolve (Modular Kernel Evolve), a framework that iteratively co-evolves a modular decomposition of complex PyTorch modules and the LLM-generated kernel for each submodule, refining the decomposition by splitting and fusing across iterations while independently improving each subkernel via LLM-driven beam search. The resulting kernels are programmatic compositions of independently verified subkernels, making them configurable (subkernel implementations are swappable), interpretable (errors and speedups are traceable to specific subkernels), and readily adaptable to related model architectures. Experiments with Triton on KernelBench L2 and L3, spanning multi-operator sequences and full model architectures, show that MKEvolve improves both correctness and speedup over end-to-end direct synthesis baselines while reducing LLM token usage by up to 35%.

Chinese Translation

尽管基于大规模语言模型（LLM）的代码生成取得了快速进展，但为硬件加速器编写正确且高效的内核仍然是扩展现代机器学习工作负载的关键瓶颈。我们提出了MKEvolve（模块化内核演化），这是一个框架，能够迭代地共同演化复杂PyTorch模块的模块化分解和为每个子模块生成的LLM内核，通过在迭代中拆分和融合来优化分解，同时通过基于LLM的束搜索独立改进每个子内核。生成的内核是独立验证的子内核的程序化组合，使其可配置（子内核实现可以互换）、可解释（错误和加速可以追溯到特定子内核），并且能够轻松适应相关模型架构。在KernelBench L2和L3上使用Triton进行的实验，涵盖多操作符序列和完整模型架构，表明MKEvolve在提高正确性和加速方面优于端到端直接合成基线，同时将LLM的令牌使用量减少了多达35%。

View on arXiv Download PDF AI Translation

cs.AI / 38 / 2607.20502

Inducing Comparability of Factorised Probability Distributions

诱导因子化概率分布的可比性

Speller, Jan, Luttermann, Malte, Gehrke, Marcel, Braun, Tanya

Abstract

To allow for principled comparison between two probabilistic graphical models defined over non-identical variable sets, they have to be lifted to a common measurable space. To this end, we propose an extension scheme for any two given models and establish the formal foundation: Unmatched components are completed using conditionally uniform (Laplace) extensions such that the resulting joint distributions differ from the original ones only by multiplicative constants and coincide under projection. This preserves the probabilistic semantics while enabling the application of well-defined distributional discrepancy measures. We establish the invariance of the induced joint under projection and use the extensions to provide a minimal structural extension of two factor graphs to the smalles common measurable space as well as to a common graphical structure by a deterministic algorithm. In addition, we discuss structural and measure-theoretic properties and identify promising criteria for comparison methodologies.

Chinese Translation

为了在定义于非相同变量集的两个概率图模型之间进行原则性的比较，必须将它们提升到一个共同的可测空间。为此，我们提出了一种扩展方案，适用于任何两个给定模型，并建立了正式基础：使用条件均匀（拉普拉斯）扩展来补全不匹配的组件，使得得到的联合分布仅与原始分布在乘法常数上有所不同，并在投影下重合。这保持了概率语义，同时使得应用明确定义的分布差异度量成为可能。我们建立了在投影下诱导的联合分布的不变性，并利用这些扩展提供了两个因子图的最小结构扩展，以达到最小的共同可测空间，以及通过确定性算法达到共同的图结构。此外，我们讨论了结构和测度理论属性，并确定了比较方法的有希望的标准。

View on arXiv Download PDF AI Translation

cs.AI / 39 / 2607.20503

LeanFlow: A Case Study in Workflow-Driven Lean Autoformalization

LeanFlow：基于工作流驱动的精益自动形式化案例研究

Milikic, Lazar, Guilloud, Simon, Nguyen, Khanh, Kuncak, Viktor

Abstract

We present and evaluate LeanFlow, an LLM agent system specialized for translating mathematical papers into buildable Lean projects. Recent verifier-in-the-loop systems show that large formal artifacts can be produced, but it remains unclear which runtime mechanisms affect completion, auditability, or efficiency in document-to-project formalization. We study this question through case studies on two previously unformalized mathematical papers in number theory and measure theory, using model, proof-workflow, and toolset ablations with Kimi2.6 and GPT5.5; we report task outcome, API calls, input tokens, and output tokens. With Kimi2.6, the full workflow completes both document-level projects within the 2000-call budget, while no-queue variants reach the budget limit; with GPT5.5, all document-level variants complete, and the full workflow has the lowest or tied-lowest input-token cost on both sources. As complementary calibration, LeanFlow reaches 75.7% BEq+ on the PFR slice of RLM25 and solves all five ICML 2026 AI for Math TCS challenge projects in our GPT5.5 runs.

Chinese Translation

我们提出并评估了LeanFlow，一个专门用于将数学论文翻译为可构建的Lean项目的LLM代理系统。最近的验证者在环系统表明可以生成大型正式文档，但尚不清楚哪些运行机制会影响文档到项目形式化的完成、可审计性或效率。我们通过对两篇在数论和测度论中未形式化的数学论文进行案例研究，使用Kimi2.6和GPT5.5进行模型、证明工作流和工具集的消融实验，研究了这个问题；我们报告了任务结果、API调用、输入标记和输出标记。使用Kimi2.6，完整工作流在2000次调用预算内完成了两个文档级项目，而无队列变体达到了预算上限；使用GPT5.5，所有文档级变体均完成，且完整工作流在两个来源上具有最低或并列最低的输入标记成本。作为补充校准，LeanFlow在RLM25的PFR切片上达到了75.7%的BEq+，并在我们的GPT5.5运行中解决了所有五个ICML 2026 AI for Math TCS挑战项目。

View on arXiv Download PDF AI Translation

cs.AI / 40 / 2607.20506

Optimizing Hypergraph-Based RAG: Toward Better Fact Extraction and Chunk Retrieval

优化基于超图的 RAG：朝着更好的事实提取和块检索

Khrouf, Houda, Fillastre, Pedro, Correia, Sebastiao

Abstract

GraphRAG enables deeper reasoning by structuring knowledge as graphs but struggles with n-ary facts. HyperGraphRAG uses hypergraphs for richer semantics, improving accuracy, yet relies on error-prone LLM extraction and inefficient standard chunk retrieval. We address this by employing self-consistency prompting to improve the extraction, and Personalized PageRank algorithm over hypergraph to enhance chunk retrieval.

Chinese Translation

GraphRAG 通过将知识结构化为图形，实现了更深层次的推理，但在处理 n-元事实时存在困难。HyperGraphRAG 利用超图提供更丰富的语义，提高了准确性，但依赖于容易出错的 LLM 提取和低效的标准块检索。我们通过采用自一致性提示来改善提取，并在超图上应用个性化 PageRank 算法来增强块检索，解决了这一问题。

View on arXiv Download PDF AI Translation

cs.AI / 41 / 2607.20507

MiniCache: Reusable Program Caching with Small Model Interfaces for Efficient LLM Inference

MiniCache：具有小模型接口的可重用程序缓存框架以实现高效的LLM推理

Chen, Jingquan, Piao, Jinghua, Feng, Jie, Hu, Shaogang, Li, Yong

Abstract

Large language models (LLMs) are increasingly used for program-aided reasoning, agentic decision making, and structured task execution, but these applications often incur high inference cost. We present MiniCache, a reusable program caching framework that transforms Program-of-Thought (PoT) programs into parameterized cache objects, enabling reusable computation across structurally similar requests. MiniCache reuses the same small model for semantic variable extraction on cache-hit requests and speculative drafting during target-LLM generation, reducing expensive target-LLM invocations while preserving task quality. Experiments on shopping-style request datasets, WebShop, Formula, and CodeTAT-QA demonstrate that MiniCache improves the trade-off between inference latency, cache reuse, and accuracy, achieving up to 3.1x lower latency and 2.8x higher throughput under parallel serving. These results show that small models are most effective not as replacements for large models, but as lightweight interface models that enable reliable and efficient reusable program caching.

Chinese Translation

大型语言模型（LLMs）越来越多地用于程序辅助推理、代理决策和结构化任务执行，但这些应用通常会产生高昂的推理成本。我们提出了MiniCache，一种可重用的程序缓存框架，它将思维程序（Program-of-Thought，PoT）转化为参数化缓存对象，从而在结构相似的请求之间实现可重用计算。MiniCache在缓存命中请求上重用相同的小模型进行语义变量提取，并在目标LLM生成过程中进行推测性草拟，从而减少了昂贵的目标LLM调用，同时保持任务质量。在购物风格请求数据集WebShop、Formula和CodeTAT-QA上的实验表明，MiniCache改善了推理延迟、缓存重用和准确性之间的权衡，在并行服务下实现了高达3.1倍的延迟降低和2.8倍的吞吐量提升。这些结果表明，小模型最有效的方式不是替代大型模型，而是作为轻量级接口模型，能够实现可靠和高效的可重用程序缓存。

View on arXiv Download PDF AI Translation

cs.AI / 42 / 2607.20510

Telco-GAIA: Bilingual Benchmark for Agents in Telecom Domain

Telco-GAIA：电信领域代理的双语基准测试

Khizbullin, Dmitrii, Alyafeai, Zaid, Eldesokey, Abdelrahman, AlSultan, Nourah, Alshalan, Raghad, Pugh, David R., Ghanem, Bernard

Abstract

We introduce Telco-GAIA, a bilingual, multi-modal benchmark for evaluating tool-using agents on the data of a real-world telecommunications operator. Telco-GAIA comprises 100 human-verified question-answering tasks, in English and Arabic, that each demand multi-hop reasoning (4.2 hops on average) over three heterogeneous sources: a static website snapshot (HTML, images, and linked PDFs), a synthetic relational SQL database, and external web archives, spanning text, image, and tabular modalities. The benchmark is delivered as a sandboxed Docker environment and scored by normalized exact string matching, making evaluation objective, deterministic, and reproducible over time without any LLM-as-a-Judge. Evaluating a purpose-built reference agent across twelve commercial and open LLMs, we find Telco-GAIA challenging: even the strongest model solves only 71% of tasks; under a moderate cost budget, this falls to about 40%, and the visually grounded categories remain the weakest, where the average backend scores below 30%, leaving substantial headroom in document and image understanding. Telco-GAIA offers a rigorous, reproducible testbed for enterprise agents and a template for constructing closed-domain benchmarks.

Chinese Translation

我们介绍了Telco-GAIA，这是一个双语的多模态基准，用于评估在真实世界电信运营商数据上使用工具的代理。Telco-GAIA包含100个经过人工验证的问答任务，分别用英语和阿拉伯语表述，每个任务都需要在三个异构来源上进行多跳推理（平均4.2跳）：一个静态网站快照（HTML、图片和链接的PDF）、一个合成的关系型SQL数据库，以及外部网页档案，涵盖文本、图像和表格等多种模态。该基准以沙箱化的Docker环境提供，并通过标准化的精确字符串匹配进行评分，使评估过程客观、确定且可重复，且不依赖于任何大型语言模型（LLM）作为评判者。在对十二个商业和开放的LLM进行评估时，我们发现Telco-GAIA具有挑战性：即使是最强的模型也仅能解决71%的任务；在适度的成本预算下，这一比例降至约40%，而视觉基础的类别仍然是最弱的，平均后端得分低于30%，在文档和图像理解方面留有相当大的提升空间。Telco-GAIA为企业代理提供了一个严格且可重复的测试平台，并为构建封闭域基准测试提供了模板。

View on arXiv Download PDF AI Translation

cs.AI / 43 / 2607.20511

SiGMA: Sign-Guided Merging and Adaptation for Multimodal Continual Instruction Tuning

SiGMA：基于信号的合并与适应用于多模态持续指令调优

Park, Keonhee, Kim, Gunhee

Abstract

Multimodal Continual Instruction Tuning (MCIT) is crucial for adapting Multimodal Large Language Models (MLLMs) to evolving a sequence of downstream tasks. Prior methods mostly utilize Mixture of Experts or expansion merge approach, primarily focusing on catastrophic forgetting, yet they still suffer from negative interference during inference, where newly learned updates overwrite useful prior knowledge and degrade overall performance. To address this, we propose SiGMA (Sign Guided Merging and Adaptation), a simple yet effective framework that mitigates negative interference with two components: sign guided adaptive tuning during training and sign guided merging at inference. Sign guided adaptive tuning reduces collisions with past knowledge and learns the current task with minimal drift, mitigating severe forgetting. Sign guided merging further improves consolidation by selectively scaling salient parameters to preserve and amplify useful task specific knowledge. Experiments on UCIT and DCL benchmarks show that SiGMA significantly reduces negative interference and outperforms state of the art MCIT methods. Our code is available at SiGMA.

Chinese Translation

多模态持续指令调优（MCIT）对于将多模态大型语言模型（MLLMs）适应于不断演变的下游任务序列至关重要。先前的方法主要利用专家混合（Mixture of Experts）或扩展合并方法，主要关注灾难性遗忘，但在推理过程中仍然遭受负干扰，新学习的更新覆盖了有用的先前知识，导致整体性能下降。为了解决这个问题，我们提出了SiGMA（基于信号的合并与适应），这是一个简单而有效的框架，通过两个组件来减轻负干扰：训练过程中的基于信号的自适应调优和推理过程中的基于信号的合并。基于信号的自适应调优减少了与过去知识的冲突，并以最小的漂移学习当前任务，从而减轻了严重遗忘。基于信号的合并通过选择性地缩放显著参数进一步改善了知识的巩固，以保留和增强有用的任务特定知识。在UCIT和DCL基准上的实验表明，SiGMA显著减少了负干扰，并超越了最先进的MCIT方法。我们的代码可在SiGMA获取。

View on arXiv Download PDF AI Translation

cs.AI / 44 / 2607.20515

Reliability-Aware LLM Alignment from Inconsistent Human Feedback

基于可靠性的LLM对齐方法：应对不一致的人类反馈

Huang, Jingyi, Zong, Ruohan, Feng, Yujun, Ma, Liran, Shang, Lanyu, Zhang, Yang

Abstract

Reinforcement Learning from Human Feedback (RLHF) is critical for aligning Large Language Models (LLMs) with human preferences. However, its efficacy is often compromised by the inherent inconsistency and subjectivity of human annotations. Existing preference optimization frameworks, such as Direct Preference Optimization (DPO), typically treat ambiguous pairs with high annotator disagreement identically to those with unanimous consensus, forcing models to overfit to inconsistent supervision signals and leading to suboptimal alignment. In this work, we propose Reliability-Guided Preference Optimization (RGPO), a robust framework designed to mitigate the impact of inconsistent human feedback. RGPO estimates annotator reliability and infers latent ground truth labels from noisy human feedback to identify robust preferences. Furthermore, we introduce a reliability-aware consistency optimization that dynamically modulates the training objective based on the consensus level of annotations, ensuring the model prioritizes high-consensus supervision signals. Extensive experiments on LLM alignment benchmarks demonstrate that RGPO effectively reduces inconsistency and noise in training data and achieves superior performance compared to widely adopted RLHF baselines. Our code and configurations are available at https://github.com/GenieHuang/RGPO.

Chinese Translation

基于人类反馈的强化学习（RLHF）对于将大型语言模型（LLMs）与人类偏好对齐至关重要。然而，其有效性常常受到人类注释固有的不一致性和主观性的影响。现有的偏好优化框架，如直接偏好优化（DPO），通常将高注释者分歧的模糊对与一致共识的对等同对待，迫使模型过度拟合不一致的监督信号，从而导致次优的对齐效果。在本研究中，我们提出了一种可靠性引导的偏好优化（RGPO）框架，旨在减轻不一致的人类反馈的影响。RGPO通过估计注释者的可靠性，并从嘈杂的人类反馈中推断潜在的真实标签，以识别稳健的偏好。此外，我们引入了一种基于可靠性的连续性优化方法，该方法根据注释的一致性水平动态调整训练目标，确保模型优先考虑高共识的监督信号。在大型语言模型对齐基准上的大量实验表明，RGPO有效减少了训练数据中的不一致性和噪声，并且相比广泛采用的RLHF基线取得了更优的性能。我们的代码和配置可在 https://github.com/GenieHuang/RGPO 获取。

View on arXiv Download PDF AI Translation

cs.AI / 45 / 2607.20518

CANN Bench: Benchmarking Agent Generated Kernels against Real NPU and Algorithmic Limits

CANN Bench：针对真实NPU和算法限制的代理生成内核基准测试

Gao, Xue-Jian, Pan, Deng, Su, Yueming, Li, Jiasheng, Du, Bin, Zhu, Fengming, Ma, Chengdi, Fan, Junyi, Liao, Qichen, Hu, Chengqiu, Chen, Xinxian, Zheng, Lingchao, Li, Jun, Yang, Jiwei, Fan, Yuwei

Abstract

AI agents are now capable of writing, compiling, and iteratively optimizing low-level operator kernels on different hardware platforms. Existing benchmarks, however, focus almost exclusively on CUDA and Triton, leaving hardware ecosystems with less-exposed programming models without a common evaluation baseline. We present CANN Bench, an open benchmark for AI-generated operator code on Huawei's Ascend NPU. The current release covers 53 operators and 1060 test cases organized into four difficulty tiers -- from simple elementwise primitives to MoE dispatch and FlashAttention kernels -- spanning FP16, BF16, FP32, and INT8 precision formats. Evaluation adopts a \textbf{three-dimensional weighted composite score} that treats compilation, functional correctness, and performance as independent axes, providing a principled reward signal for kernel-generation agents. Performance is graded against an out-of-the-box PyTorch-on-Ascend baseline and an analytical per-case Hardware-Anchored Performance (HAP) limit on real NPU hardware, ensuring scores reflect genuine optimization headroom rather than measurement artifacts. The evaluation harness is designed to resist reward hacking from the ground up. CANN Bench is versioned within the official CANN repository and is designed for long-term community co-construction, providing the Ascend ecosystem with a quantitative, reproducible, and sustainably maintained yardstick for AI operator-authoring capability.

Chinese Translation

AI代理现在能够在不同的硬件平台上编写、编译和迭代优化低级操作内核。然而，现有的基准测试几乎完全专注于CUDA和Triton，导致那些编程模型曝光较少的硬件生态系统缺乏共同的评估基准。我们提出了CANN Bench，这是一个针对华为Ascend NPU上AI生成操作代码的开放基准测试。当前版本涵盖53个操作符和1060个测试用例，分为四个难度等级——从简单的逐元素原语到MoE调度和FlashAttention内核——涵盖FP16、BF16、FP32和INT8精度格式。评估采用 extbf{三维加权复合评分}，将编译、功能正确性和性能视为独立的轴，为内核生成代理提供了一个原则性的奖励信号。性能评分是基于开箱即用的PyTorch-on-Ascend基准和真实NPU硬件上的分析性逐案例硬件锚定性能（HAP）限制，确保得分反映真实的优化空间，而非测量伪影。评估工具的设计旨在从根本上抵抗奖励黑客行为。CANN Bench在官方CANN库中进行版本管理，旨在长期促进社区共同建设，为Ascend生态系统提供一个定量、可重复和可持续维护的AI操作符创作能力的衡量标准。

View on arXiv Download PDF AI Translation

cs.AI / 46 / 2607.20520

Representation Robustness Under Executable Reasoning Constraints in Large Language Models for Mathematical Problem Solving

在可执行推理约束下的大型语言模型的表示鲁棒性：针对数学问题求解的研究

Nath, Sagnik, Graf, Edith Aurora, Zhang, Liang, Zapata-Rivera, Diego

Abstract

Large language models (LLMs) are increasingly evaluated on mathematical problem solving, yet prior work often treats representationally equivalent formulations as interchangeable and conflates reasoning errors with interface failures. This paper investigates representation robustness in LLM-based mathematical problem solving by systematically varying surface representations of the same underlying problems, including story problems, word-equations, symbolic equations, and isomorphic paraphrases. Using a curated dataset of mathematically equivalent problems, we evaluate five contemporary LLMs under a direct answer generation condition. We find substantial representational sensitivity: models frequently change correctness across equivalent formulations, with nontrivial flip rates across story, symbolic, and word-equation variants. We also observe systematic regressions under isomorphic reformulations, showing that even subtle paraphrase-level changes can degrade performance despite preserved mathematical structure. We then evaluate a code-augmented condition in which models externalize reasoning as executable Python code that is run locally for validation. This interface reveals strong latent reasoning capability in some models that perform poorly under direct prompting, but it does not uniformly improve robustness. Instead, failures shift across interaction layers, from opaque reasoning errors to protocol violations and execution failures. Even when executable reasoning succeeds, representation sensitivity often persists. Overall, our results show that reasoning scaffolds do not eliminate representational brittleness, but expose new tradeoffs among correctness, reliability, latency, and cost. We argue that representation should be treated as a first-class interface design variable in LLM evaluation and deployment, especially for AI-assisted problem-solving systems.

Chinese Translation

大型语言模型（LLMs）在数学问题求解方面的评估日益增多，但以往的研究往往将表示等价的表述视为可互换，并将推理错误与接口失败混为一谈。本文通过系统地变换相同基础问题的表面表示形式，包括故事问题、文字方程、符号方程和同构释义，研究基于LLM的数学问题求解中的表示鲁棒性。我们使用一个经过精心策划的数学等价问题数据集，在直接答案生成条件下评估五个当代LLM。我们发现显著的表示敏感性：模型在等价表述之间的正确性经常发生变化，故事、符号和文字方程变体之间的翻转率非平凡。我们还观察到在同构重述下的系统性退化，表明即使是微妙的释义级变化也会降低性能，尽管数学结构得以保留。随后，我们评估了一种代码增强条件，其中模型将推理外化为可执行的Python代码，并在本地运行以进行验证。该接口揭示了一些在直接提示下表现不佳的模型具有强大的潜在推理能力，但并未普遍提高鲁棒性。相反，失败在交互层次之间转移，从不透明的推理错误到协议违规和执行失败。即使在可执行推理成功的情况下，表示敏感性往往仍然存在。总体而言，我们的结果表明，推理支架并未消除表示脆弱性，而是揭示了正确性、可靠性、延迟和成本之间的新权衡。我们认为，在LLM的评估和部署中，表示应被视为一项重要的接口设计变量，尤其是在AI辅助问题解决系统中。

View on arXiv Download PDF AI Translation

cs.AI / 47 / 2607.20524

Attention Degradation, Function Token Anchoring, and the Limits of Attention-Based Intervention in Large Language Models

注意力退化、功能标记锚定及基于注意力的干预在大型语言模型中的局限性

Dangal, Sagar, Shakya, Manoj

Abstract

Mean cross-positional attention degradation is widely reported in transformer interpretability, yet whether it causally limits contextual retrieval remains untested. We present six coordinated experiments across GPT-2, LLaMA-3.2-1B/3B, OPT-1.3B, and distilgpt2. We first characterise short-term (5-100 token) attention degradation, finding a universal exponential-then-plateau pattern whose rate is inversely correlated with depth, with distinct layer-wise entropy signatures per architecture. Function token anchoring proves architecture-dependent: OPT-1.3B (absolute positional encoding) shows distance-dependent preposition specificity, GPT-2 shows uniform non-specific dependence, and LLaMA (RoPE) shows reversal at long distances. Strategic comma insertion at clause boundaries causally reduces prediction degradation in the 40-80 token range, with the benefit tied to syntactic boundary alignment rather than token density. We then test the mechanism causally: Relay-Aware Attention (RAA), which biases attention logits toward function token positions, verifiably increases attention mass by 16-24% yet yields null effects on GPT-2 and LLaMA-1B, preliminary harm on LLaMA-3B, and a mixed effect on OPT-1.3B that nets to approximately zero. Multi-fact retrieval probes further show that degradation rate does not predict retrieval accuracy across models. We conclude that mean attention degradation is largely descriptive rather than prescriptive: function tokens contribute through what their hidden states compute, not through the attention they receive -- with implications for interpretability methodology and attention-score-based inference optimisations such as KV-cache eviction.

Chinese Translation

在变换器可解释性研究中，平均跨位置注意力退化被广泛报道，但其是否在因果上限制了上下文检索尚未得到验证。我们在GPT-2、LLaMA-3.2-1B/3B、OPT-1.3B和distilgpt2上进行了六项协调实验。我们首先描述了短期（5-100个标记）注意力退化，发现其具有普遍的指数-再平稳模式，其速率与深度呈反相关，并且每种架构具有不同的层级熵特征。功能标记锚定依赖于架构：OPT-1.3B（绝对位置编码）显示出距离依赖的介词特异性，GPT-2显示出均匀的非特异性依赖，而LLaMA（RoPE）在长距离时则表现出反转。在子句边界处的战略性逗号插入因果性地减少了40-80个标记范围内的预测退化，其益处与句法边界对齐而非标记密度相关。然后我们因果性地测试了这一机制：Relay-Aware Attention（RAA），它将注意力对数偏向于功能标记位置，验证性地增加了16-24%的注意力质量，但对GPT-2和LLaMA-1B没有显著影响，对LLaMA-3B初步产生了负面影响，而对OPT-1.3B的影响则呈现混合效果，最终净效应约为零。多事实检索探针进一步显示，退化率并不能预测各模型的检索准确性。我们得出结论，平均注意力退化在很大程度上是描述性的而非规定性的：功能标记通过其隐藏状态计算的内容而非所接收的注意力做出贡献，这对可解释性方法论和基于注意力分数的推理优化（如KV-cache驱逐）具有重要意义。

View on arXiv Download PDF AI Translation

cs.AI / 48 / 2607.20525

Autonomous disproofs of the sum-product conjecture over $\mathbb R$ with GPT-5.5 Pro

基于GPT-5.5 Pro的$ extbf{R}$上和积猜想的自主反驳

Huang, Yichen

Abstract

OpenAI's recent disproof of the Erd\H{o}s unit distance conjecture marked a milestone for AI in mathematics. It also inspired another breakthrough: a human disproof of the Erd\H{o}s--Szemer\'edi sum-product conjecture over $\mathbb R$. In this paper, we present a simple agent built on GPT-5.5 Pro. Using a problem-agnostic, three-stage prompting pipeline -- proof-plan proposal, proof construction, and review -- the agent autonomously generated correct proofs that the sum-product conjecture is false over $\mathbb R$ in 7 of 8 independent trials; in the remaining trial, it identified an unresolved gap in its argument. The seven proofs are diverse: some are close to existing unit-based constructions, while others avoid units by using $L^p$-type regions of algebraic integers. The system used an average of 132.4k reasoning tokens per trial. We release the code, intermediate outputs, and generated proofs, providing a reproducible, data-contamination-free case study in autonomous proof generation.

Chinese Translation

OpenAI最近对Erd ext{ö}s单位距离猜想的反驳标志着人工智能在数学领域的一个里程碑。这也激发了另一个突破：人类对Erd ext{ö}s--Szemerédi和积猜想在$ extbf{R}$上的反驳。本文介绍了一个基于GPT-5.5 Pro构建的简单代理。该代理使用一个与问题无关的三阶段提示管道——证明计划提议、证明构建和审查——自主生成了7个独立试验中8个的和积猜想在$ extbf{R}$上是错误的正确证明；在剩余的试验中，它识别出了论证中的一个未解决的空白。这七个证明各具特色：有些接近现有的基于单位的构造，而另一些则通过使用$L^p$类型的代数整数区域来避免单位。该系统在每次试验中平均使用了132.4k推理令牌。我们发布了代码、中间输出和生成的证明，提供了一个可重复的、无数据污染的自主证明生成案例研究。

View on arXiv Download PDF AI Translation

cs.AI / 49 / 2607.20526

ConfidenceBench: Evaluating Confidence Calibration in Large Language Models

ConfidenceBench：评估大型语言模型中的置信度校准

ffrench-Constant, Matthew, Yang, Daniel, Huang, Xinmeng, Kapoor, Sanyam

Abstract

Large language models (LLMs) are increasingly deployed in settings where fluent but incorrect answers can be costly. In these settings, accuracy alone is insufficient: models must also know when they are likely to be wrong. We present ConfidenceBench, a calibration benchmark that evaluates verbalized confidence estimates in 15 frontier LLMs using the Brier score, a proper scoring rule that incentivises truthful probability reporting. Confidence is elicited via prompting, requiring no access to model logits and making the framework applicable to both closed-source and open-source systems. The benchmark comprises 200 private multiple-choice questions across four categories: spatial reasoning, high-precision mathematics, word lookup, and unknowable questions. Across three independent runs, Claude Opus 4.6 and Gemini 3.1 Pro Preview achieve the lowest Brier scores, both reported as 0.103. Both substantially outperform the calibrated-random baseline of 0.1875, while Gemini 3.1 Flash-Lite scores 0.367, indicating severe miscalibration. Accuracy and calibration diverge substantially across model families: the most accurate model is not the best-calibrated, and several models perform worse than the calibrated-random Brier baseline despite reasonable accuracy. These results show that verbalized confidence calibration is a distinct and practically important axis of LLM reliability, complementary to standard accuracy-based evaluation.

Chinese Translation

大型语言模型（LLMs）越来越多地应用于流利但错误的答案可能代价高昂的场景。在这些场景中，仅仅依靠准确性是不够的：模型还必须知道何时它们可能会出错。我们提出了ConfidenceBench，这是一个校准基准，使用Brier分数评估15个前沿LLM的口头置信度估计，Brier分数是一种激励真实概率报告的适当评分规则。通过提示获取置信度，无需访问模型的logits，使得该框架适用于闭源和开源系统。该基准包含200个私人多项选择题，分为四个类别：空间推理、高精度数学、单词查找和不可知问题。在三次独立运行中，Claude Opus 4.6和Gemini 3.1 Pro Preview的Brier分数最低，均为0.103。两者的表现均显著优于校准随机基线0.1875，而Gemini 3.1 Flash-Lite的得分为0.367，显示出严重的校准失调。不同模型家族的准确性和校准存在显著差异：最准确的模型并不是校准最好的模型，尽管准确性合理，仍有多个模型的表现低于校准随机Brier基线。这些结果表明，口头置信度校准是LLM可靠性的一条独特且实际重要的维度，补充了基于标准准确性的评估。

View on arXiv Download PDF AI Translation

cs.AI / 50 / 2607.20527

Evaluating and Guarding Citation Faithfulness in Agentic Scientific Synthesis

评估与保护代理科学综合中的引用真实性

Goo, Taewan, Kim, Junsik, Han, Kyulhee, Jo, GwonYul, Kim, Jong-Soo, Kim, Tae-Hyung

Abstract

Agentic LLM systems such as OpenScholar and PaperQA2 read the scientific literature and return cited answers, and both they and their benchmarks already check whether those citations hold, with a fixed attribution model or human graders. Neither audits the reliability of that check itself. We show it is not reliable, and that this matters. On identical agent outputs the measured unsupported-citation rate ranges from about 3% to about 18% depending only on the verifier's strictness, and although verifiers agree on which citations are supported, they disagree on which to flag (negative-specific agreement 0.27 to 0.30), so no single flag set is trustworthy and cross-paper comparison is invalid without a named verifier and protocol. We present a gold-anchored evaluation protocol and a deployable guard that make this behavior measurable and bounded. The protocol validates the verifier, measures re-attribution, and calibrates a guarantee against human gold rather than another model's verdict; the verifier is a swappable instrument chosen on cost (recall 0.94 on the supported class, held out), and re-attribution is a commodity step where a deterministic BM25 matches the best open generator. The guard adds a split-conformal layer placing a distribution-free, finite-sample bound on truly unsupported citations that slip past a chosen flagging rule, a guarantee on catch rate rather than conclusion correctness. The bound holds on held-out gold, and we identify and quantify the condition governing its transfer to deployment, calibration-negative difficulty, with a concrete recalibration recipe, left untested by prior conformal-factuality work. Validated across four open 27-35B models and three agentic pipelines on public benchmarks (SciFact, QASA, PubMedQA), with confidence intervals on every headline number, the protocol and guard ship as an open single-GPU kit.

Chinese Translation

代理型大语言模型（LLM）系统，如 OpenScholar 和 PaperQA2，能够阅读科学文献并返回引用的答案，这些系统及其基准测试已经检查了这些引用的有效性，使用固定的归属模型或人工评分。然而，现有的审计并未验证这一检查的可靠性。我们展示了这一点并指出其重要性。在相同的代理输出中，测得的无支持引用率在约 3% 到 18% 之间变化，仅取决于验证者的严格程度。尽管验证者在哪些引用是被支持的方面达成一致，但在标记哪些引用方面存在分歧（负特异性一致性为 0.27 到 0.30），因此没有单一的标记集是可信的，跨论文比较在没有指定验证者和协议的情况下是无效的。我们提出了一种基于黄金标准的评估协议和可部署的保护措施，使这一行为可测量且有界。该协议验证验证者，测量重新归属，并针对人类黄金标准而非其他模型的判决进行校准；验证者是一个可更换的工具，依据成本选择（在支持类上的召回率为 0.94，保留数据），而重新归属是一个商品化步骤，其中确定性 BM25 匹配最佳的开放生成器。保护措施增加了一个分裂一致性层，为真正的无支持引用提供了一个无分布、有限样本的界限，这些引用在选择的标记规则下滑过，保证了捕获率而非结论的正确性。该界限在保留的黄金标准上成立，我们识别并量化了其转移到部署的条件，即校准负难度，并提供了一个具体的重新校准方案，这是之前一致性事实工作未曾测试的。通过四个开放的 27-35B 模型和三个代理管道在公共基准（SciFact、QASA、PubMedQA）上进行了验证，并对每个主要数字提供了置信区间，该协议和保护措施作为一个开放的单 GPU 套件发布。

View on arXiv Download PDF AI Translation

cs.AI / 51 / 2607.20528

PromptPack: Scaling LLM Annotation Agents for Online Recommendation

PromptPack：扩展在线推荐的 LLM 注释代理

Koralewski, Sebastian, Barlier, Merwan, Stolin, Yulia, Škrlj, Blaž

Abstract

Online recommendation platforms increasingly use Large Language Models (LLMs) to extract structured features from ad creatives. While deploying a single-call LLM annotation agent yields significant Click-Through Rate (CTR) improvements in our live production environment, per-creative prompting is prohibitively expensive to scale. The redundant system instructions sent in every request account for 94% of billed input tokens. To break this cost bottleneck, we introduce PromptPack, a scalable, high-throughput LLM annotation agent. PromptPack achieves this scale via in-context batching, combining a shared system prompt, a strict XML structural envelope, and an output correction layer to ensure deterministic, pipeline-ready feature extraction across multiple creatives simultaneously. We evaluate PromptPack via an offline retrieval benchmark using a downstream logistic-regression ranker. To deeply profile the agent's behavior, we measure AUC and introduce Volume-Weighted Absolute Lift (VWAL), a novel metric capturing the signal quality of the generated features. Compared to our live, unbatched production baseline, PromptPack at batch size 20 cuts our LLM costs by 89% and accelerates throughput by 2.5x while fully preserving AUC.

Chinese Translation

在线推荐平台越来越多地使用大型语言模型（LLMs）从广告创意中提取结构化特征。虽然在我们的实时生产环境中部署单次调用的 LLM 注释代理显著提高了点击率（CTR），但逐个创意的提示在规模化时代价过高。每个请求中发送的冗余系统指令占据了 94% 的计费输入令牌。为了打破这一成本瓶颈，我们引入了 PromptPack，一个可扩展的高吞吐量 LLM 注释代理。PromptPack 通过上下文批处理实现了这一规模，结合了共享系统提示、严格的 XML 结构封装和输出校正层，以确保在多个创意之间同时进行确定性、管道就绪的特征提取。我们通过使用下游逻辑回归排序器的离线检索基准评估 PromptPack。为了深入分析代理的行为，我们测量了 AUC，并引入了体积加权绝对提升（VWAL），这一新颖的指标捕捉了生成特征的信号质量。与我们的实时、未批处理的生产基线相比，PromptPack 在批量大小为 20 时将我们的 LLM 成本降低了 89%，并将吞吐量提高了 2.5 倍，同时完全保持了 AUC。

View on arXiv Download PDF AI Translation

cs.AI / 52 / 2607.20531

DynamicMCPBench: A Trace-Grounded, Effect-Scored Benchmark for LLM Agents over Live MCP Servers

DynamicMCPBench：针对实时MCP服务器的基于追踪、效果评分的LLM代理基准

Kamiński, Jerzy, Galyukshev, Ilya, Kuznetsov, Artem, Chuprin, Sergey, Redko, Kirill, Shumbalov, Aidar, Kalyuzhnaya, Anna

Abstract

Large language model (LLM) agents are increasingly deployed over Model Context Protocol (MCP) servers, yet the benchmarks used to evaluate them score the final answer or a fixed "ground-truth" list of tools, both of which are fragile once the underlying data is live and stateful. We present DynamicMCPBench, a reusable framework rather than a fixed dataset. A practitioner can run it on their own MCP servers to test models on their own tasks, or let it collect servers automatically to measure a model's general ability to solve agentic tasks. Given the servers and any set of models, it generates realistic goals, pursues each one live to record a successful trajectory, distills that trajectory into path-agnostic effect checkpoints, and scores an agent on whether it reproduces those effects, never on the final answer. To show what the framework reveals, we run it at scale: 24 models over 121 servers and 750 tasks spread evenly over 15 task categories (50 each), where each category targets a distinct tool-use challenge of the generated questions. Each task is scored by pass^3: it counts as solved only if all three independent attempts succeed. Even the strongest agents solve only about half of the tasks, 31% of tasks are solved by no model at all, and accuracy collapses as the required tool chain grows longer (from 39% on the shortest chains to 13% on the longest). A human validation study confirms the automatic scoring is reliable (chance-corrected agreement of 0.76). DynamicMCPBench thus turns benchmark construction into something practitioners can rerun on their own servers and models, while exposing a consistent inability of current agents to handle long, multi-step agentic tasks.

Chinese Translation

大型语言模型（LLM）代理越来越多地部署在模型上下文协议（MCP）服务器上，但用于评估它们的基准仅对最终答案或固定的“真实”工具列表进行评分，这在底层数据为实时和有状态时显得脆弱。我们提出了DynamicMCPBench，这是一种可重用的框架，而非固定的数据集。实践者可以在自己的MCP服务器上运行它，以测试模型在自己任务上的表现，或者让它自动收集服务器，以测量模型解决代理任务的整体能力。给定服务器和任何模型集，它生成现实的目标，实时追踪每个目标以记录成功的轨迹，将该轨迹提炼为与路径无关的效果检查点，并根据代理是否重现这些效果进行评分，而不是最终答案。为了展示该框架的揭示，我们进行了大规模运行：24个模型在121个服务器和750个任务上均匀分布于15个任务类别（每类50个），每个类别针对生成问题的不同工具使用挑战。每个任务由pass^3评分：只有当三个独立尝试均成功时，才算解决。即使是最强的代理也仅解决了大约一半的任务，31%的任务没有任何模型解决，随着所需工具链的增长，准确率急剧下降（从最短链的39%降至最长链的13%）。一项人工验证研究确认自动评分的可靠性（机会校正一致性为0.76）。因此，DynamicMCPBench将基准构建转变为实践者可以在自己的服务器和模型上重新运行的过程，同时揭示了当前代理在处理长的多步骤代理任务时的一致性无能。

View on arXiv Download PDF AI Translation

cs.AI / 53 / 2607.20536

AppWorld-UL: Benchmarking Diverse Agent-User Interactions for Tool-Use

AppWorld-UL：多样化代理-用户交互的工具使用基准测试

Chen, Junzhi, Trivedi, Harsh, Pan, Jane, Zhang, Michael JQ, Srinivasan, Tejas, Balasubramanian, Niranjan, Sabharwal, Ashish

Abstract

Tool-use agents that address day-to-day digital tasks such as ordering groceries must not only operate applications, but also interact with the user, e.g., to ask clarification questions, prompt for confirmation, and inform the user when the instruction is infeasible. However, current benchmarks for evaluating agent-user interactions do not capture the diversity of such interactions. Further, they operate in small environments with few, often non-state-changing, APIs. To address this gap, we introduce AppWorld-UL, a ``user-in-the-loop'' benchmark of 516 challenging tasks requiring diverse agent-user interactions. Building upon the AppWorld framework with 9 popular simulated apps like Amazon and Spotify, we systematically modify original tasks to introduce ambiguities and constraints that necessitate various types of agent-user interaction. User behavior is simulated by an LLM prompted to respond with carefully designed knowledge boundaries, offering more reliable simulation than the unconstrained or overly rigid alternatives used in prior work. Our evaluation reveals that a state-of-the-art LLM, Claude Opus 4.7, achieves only 48.6% success on AppWorld-UL, and only 35.7% on the harder, compositional subset. On the stricter, scenario-level metric, compositional task performance drops to only 21.3%. Our analysis reveals that correct user-interaction is crucial for success. This demonstrates the benchmark's difficulty and its potential to advance research on user-in-the-loop tool-use agents.

Chinese Translation

应对日常数字任务（如订购杂货）的工具使用代理不仅需要操作应用程序，还需与用户进行交互，例如询问澄清问题、提示确认以及在指令不可行时通知用户。然而，目前评估代理-用户交互的基准测试并未捕捉到这种交互的多样性。此外，它们通常在小环境中运行，涉及的API数量少且往往不改变状态。为了解决这一空白，我们引入了AppWorld-UL，这是一个包含516个具有挑战性的任务的“用户参与”基准，要求多样化的代理-用户交互。我们在AppWorld框架的基础上，结合了9个流行的模拟应用（如亚马逊和Spotify），系统性地修改原始任务，引入模糊性和约束，迫使代理与用户进行多种类型的交互。用户行为通过一个大型语言模型（LLM）进行模拟，该模型被提示在精心设计的知识边界内作出响应，提供比以往研究中使用的无约束或过于僵化的替代方案更可靠的模拟。我们的评估显示，最先进的LLM Claude Opus 4.7在AppWorld-UL上的成功率仅为48.6%，而在更具挑战性的组合子集上仅为35.7%。在更严格的情景级指标下，组合任务的表现降至仅21.3%。我们的分析表明，正确的用户交互对成功至关重要。这表明该基准的难度及其在推动用户参与的工具使用代理研究方面的潜力。

View on arXiv Download PDF AI Translation

cs.AI / 54 / 2607.20545

StrideDiffusion: Accelerating Diffusion Models for Time-series Generation

StrideDiffusion：加速时间序列生成的扩散模型

Yin, Du, He, Estrid, Bañuelos, Julián Jerónimo, Yang, Yang, Hu, Feng, Luo, Yuchen, Xue, Hao, Sigg, Stephan, Salim, Flora

Abstract

Diffusion models have become competitive generators for time series, but their practical use is limited by the large number of sequential denoising steps required at inference time. Existing fast samplers typically use fixed or generic timestep schedules, overlooking a distinctive property of time-series diffusion: different spectral bands evolve at different rates during the reverse process. We introduce StrideDiffusion, a training-free spectral-aware sampler that adaptively selects the denoising stride from band-level activity. At each step, StrideDiffusion monitors relative band energy, log-power drift, and phase velocity to identify whether high- frequency dynamics remain active or whether the trajectory is dominated by stable low-frequency structure. It then takes fine steps when rapidly varying bands are active and larger jumps once only coarse components remain. A bandwise stability analysis shows that inactive frequency bands change only linearly with the jump size under deterministic affine reverse updates, providing a local justification for spectral activity as a step-size indicator. Across six unconditional time-series generation benchmarks, StrideDiffusion uses only 14-66 function evaluations instead of 500/1000 denoising steps, achieving up to 18.9x wall-clock speedup while preserving or improving generation quality. On conditional imputation and forecasting, it further delivers 5-14x average acceleration with comparable predictive accuracy. These results show that spectral evolution provides a practical and principled signal for fast time-series diffusion sampling. Our code is available at https://anonymous.4open.science/r/stridediff-ts.

Chinese Translation

扩散模型已经成为时间序列生成的竞争性生成器，但其实际应用受到推理时所需的大量顺序去噪步骤的限制。现有的快速采样器通常使用固定或通用的时间步调度，忽视了时间序列扩散的一个独特特性：不同的频谱带在反向过程中以不同的速率演变。我们提出了StrideDiffusion，这是一种无训练的频谱感知采样器，能够根据频带级活动自适应选择去噪步幅。在每一步中，StrideDiffusion监测相对频带能量、对数功率漂移和相位速度，以识别高频动态是否仍然活跃，或轨迹是否被稳定的低频结构所主导。当快速变化的频带处于活跃状态时，它会采取细微的步骤，而一旦只剩下粗糙成分，则会进行更大的跳跃。频带稳定性分析表明，在确定性仿射反向更新下，非活跃频带的变化仅与跳跃大小呈线性关系，为频谱活动作为步长指示器提供了局部依据。在六个无条件时间序列生成基准测试中，StrideDiffusion仅使用14-66次函数评估，而不是500/1000次去噪步骤，实现了高达18.9倍的墙钟加速，同时保持或提高了生成质量。在条件插补和预测方面，它进一步提供了5-14倍的平均加速，同时保持了可比的预测准确性。这些结果表明，频谱演变为快速时间序列扩散采样提供了一个实用且有原则的信号。我们的代码可在 https://anonymous.4open.science/r/stridediff-ts 获取。

View on arXiv Download PDF AI Translation

cs.AI / 55 / 2607.20553

CMI-Mem: Toward Generalizable Long-Term Memory Management via CMI-Augmented Reinforcement Learning

CMI-Mem：通过CMI增强的强化学习实现可泛化的长期记忆管理

Wang, Yubo, Zhao, Qiuyu, Sun, Zenghui, Dong, Shichao, Lan, Jinsong, Zhu, Xiaoyong, Li, Haoyang, Zheng, Bo, Chen, Lei

Abstract

Memory Manager models are pivotal in agent systems. Existing methods rely predominantly on LLM-judged synthetic question-answer (QA) pairs, making memory valuation dependent on sampled queries and the downstream reader. To address this limitation, we propose \textbf{CMI-Mem}, a reinforcement learning(RL)-based lightweight memory manager model with a hybrid reward that combines downstream QA correctness and intrinsic Conditional Mutual Information (CMI). CMI evaluates the information contributed by new conversational inputs relative to the current memory state without conditioning on a sampled QA query, thereby complementing rather than replacing QA grounding. Our codes are available at: https://github.com/Wyb0627/CMIMem , and the CMI-Mem-4B model checkpoint is available at: https://www.modelscope.cn/models/wyb0627/CMIMem-4B

Chinese Translation

记忆管理模型在智能体系统中至关重要。现有方法主要依赖于大型语言模型（LLM）判断的合成问答（QA）对，使得记忆评估依赖于采样查询和下游阅读器。为了解决这一局限性，我们提出了 extbf{CMI-Mem}，一种基于强化学习（RL）的轻量级记忆管理模型，采用结合下游QA正确性和内在条件互信息（CMI）的混合奖励。CMI评估新对话输入相对于当前记忆状态所贡献的信息，而不依赖于采样的QA查询，从而补充而非替代QA基础。我们的代码可在以下链接获取：https://github.com/Wyb0627/CMIMem，CMI-Mem-4B模型检查点可在：https://www.modelscope.cn/models/wyb0627/CMIMem-4B获取。

View on arXiv Download PDF AI Translation

cs.AI / 56 / 2607.20554

AI-Driven Multi-Hop Relay Selection for Smart Urban NR-V2X Networks via Learning-to-Optimize Graph Neural Networks

基于人工智能的多跳中继选择在智能城市NR-V2X网络中的优化学习图神经网络

Amati, Giambattista, Mangiatordi, Federica, Angelini, Simone, Pallotti, Emiliano, Salvo, Pierpaolo

Abstract

Reliable and low-latency NR-V2X communications are essential for smart mobility in dense urban environments. However, limited Road-Side Unit (RSU) density, frequent non-line-of-sight conditions, and highly dynamic vehicular topologies often prevent many Connected and Automated Vehicles (CAVs) from maintaining stable single-hop connectivity. Although multi-hop relay-assisted communication can extend infrastructure coverage, selecting relay links in real time under practical flow, capacity, and connectivity constraints remains challenging. Mixed-Integer Linear Programming (MILP) yields optimal multi-hop relay decisions, but its computational complexity scales sharply with network density, limiting real-time applicability. To address this, we propose a Learning-to-Optimise (L2O) framework based on Graph Neural Networks (GNNs) for real-time NR-V2X relay selection. Vehicular communication states are modeled as attributed graphs, where CAVs and RSUs are nodes and candidate radio links are enriched with propagation-aware features. An offline MILP oracle provides optimal supervision, while an edge-aware Graph Isomorphism Network (GINE) approximates oracle decisions with near-constant inference latency. Experiments on large-scale urban datasets generated by an integrated SUMO--GEMV2 simulation pipeline show that the proposed approach achieves connectivity comparable to that of the MILP oracle while reducing execution time by orders of magnitude. The framework enables cost-effective enhancement of urban V2X connectivity by leveraging existing vehicular assets and supporting scalable, real-time NR-V2X operation in smart city environments.

Chinese Translation

可靠且低延迟的NR-V2X通信对于密集城市环境中的智能出行至关重要。然而，有限的路边单元（RSU）密度、频繁的非视距条件以及高度动态的车辆拓扑常常导致许多连接和自动化车辆（CAV）无法维持稳定的单跳连接。尽管多跳中继辅助通信可以扩展基础设施覆盖范围，但在实际流量、容量和连接性约束下实时选择中继链路仍然具有挑战性。混合整数线性规划（MILP）能够产生最优的多跳中继决策，但其计算复杂度随着网络密度的增加而急剧上升，限制了其实时应用性。为了解决这一问题，我们提出了一种基于图神经网络（GNN）的学习优化（L2O）框架，用于实时NR-V2X中继选择。车辆通信状态被建模为带属性的图，其中CAV和RSU为节点，候选无线链路则通过传播感知特征进行丰富。离线MILP预言机提供最优监督，而边缘感知图同构网络（GINE）则以近乎恒定的推理延迟近似预言机决策。在由集成的SUMO-GEMV2仿真管道生成的大规模城市数据集上的实验表明，所提出的方法在实现与MILP预言机相当的连接性的同时，执行时间减少了几个数量级。该框架通过利用现有的车辆资产，支持可扩展的实时NR-V2X操作，从而实现城市V2X连接的成本效益提升，适应智能城市环境。

View on arXiv Download PDF AI Translation

cs.AI / 57 / 2607.20556

KeySI: An Interaction Framework for Tuning Text Embeddings Based on Human Feedback

KeySI：基于人类反馈调整文本嵌入的交互框架

Zhu, Yan, Chen, Y., Faust, Rebecca

Abstract

In large-scale text analysis tasks, pre-trained language models are often used to embed text corpora for downstream analysis. However, such models may struggle to capture domain-specific semantics and adapting them typically requires large amounts of labeled data and technical expertise to implement training pipelines. Recent approaches have demonstrated how visual interactions in document projections can capture human feedback as training signals for model tuning. However, these methods operate on document-level feedback, which requires users to open and assess individual documents in order to provide effective feedback. In this paper, we propose KeySI, an interaction framework that enables feature-level feedback through keyword-based concept specification. Users specify feedback by organizing extracted keywords into groups representing concepts, which KeySI translates into document-level supervision for subsequent tuning. By operating on keywords as the primary interaction medium, KeySI reduces the need for manual document inspection and labeling and lowers the barrier to adapting embedding models. We present a prototype implementation that, given a corpus, curates representative keywords, visualizes keywords and document embeddings via dimensionality reduction, allows interactive specification of keyword groups, and supports iterative refinement through system feedback. We evaluate KeySI through a user study, usage scenarios, and quantitative experiments demonstrating its effectiveness in capturing user intent and improving embedding alignment.

Chinese Translation

在大规模文本分析任务中，预训练语言模型通常用于为下游分析嵌入文本语料。然而，这些模型可能难以捕捉特定领域的语义，适应这些模型通常需要大量标注数据和技术专长来实施训练流程。近期的方法展示了如何在文档投影中通过视觉交互捕捉人类反馈作为模型调整的训练信号。然而，这些方法依赖于文档级反馈，要求用户打开并评估单个文档以提供有效反馈。本文提出了KeySI，一个通过基于关键词的概念规范实现特征级反馈的交互框架。用户通过将提取的关键词组织成代表概念的组来指定反馈，KeySI将其转化为文档级监督以便后续调整。通过将关键词作为主要交互媒介，KeySI减少了手动文档检查和标注的需求，降低了调整嵌入模型的门槛。我们展示了一个原型实现，该实现能够在给定语料的情况下，策划代表性关键词，通过降维可视化关键词和文档嵌入，允许交互式指定关键词组，并通过系统反馈支持迭代优化。我们通过用户研究、使用场景和定量实验评估KeySI，证明其在捕捉用户意图和改善嵌入对齐方面的有效性。

View on arXiv Download PDF AI Translation

cs.AI / 58 / 2607.20638

WaveformQA: Benchmarking LLM Temporal Reasoning on Digital Waveforms

WaveformQA：基于数字波形的LLM时间推理基准测试

Liu, Yichuan, Cummings, Daniel, Vadlamudi, Nick

Abstract

Large Language Models (LLMs) have demonstrated strong capabilities in code generation and reasoning, yet their ability to perform temporal reasoning over digital waveform data remains largely unexplored. Although reasoning over digital waveforms is a critical bottleneck in design verification, existing benchmarks primarily evaluate hardware description language (HDL) code generation and use waveforms only as supplementary context. This paper presents WaveformQA, an open-source question-answering benchmark for evaluating LLM temporal reasoning over digital waveforms. The benchmark comprises 360 questions with programmatically generated ground truths across eight categories of varying difficulty, including questions targeting multi-signal correlation and event ordering. Waveforms are generated from open-source design implementations, ensuring reproducibility and grounding the benchmark in real hardware behavior. Evaluation of frontier LLMs reveals that while models achieve reasonable accuracy on simple queries, performance degrades due to context window limitations and reasoning difficulties on complex temporal and multi-step questions. In addition, we show that an event-time JSON representation of waveforms improves LLM reasoning accuracy versus the standardized value change dump (VCD) format. The open-source framework supports extending to new question categories and importing new waveform sources, enabling researchers to rapidly prototype temporal reasoning experiments.

Chinese Translation

大型语言模型（LLMs）在代码生成和推理方面表现出强大能力，但其在数字波形数据上的时间推理能力尚未得到充分探索。尽管数字波形推理是设计验证中的关键瓶颈，现有基准测试主要评估硬件描述语言（HDL）代码生成，仅将波形作为辅助上下文。本文提出了WaveformQA，一种用于评估LLM在数字波形时间推理能力的开源问答基准。该基准包含360个问题，涵盖八个不同难度类别，问题包括多信号关联和事件排序等。波形数据来源于开源设计实现，确保可复现性并基于真实硬件行为。对前沿LLM的评测表明，模型在简单查询上表现合理，但在复杂时间和多步骤问题上由于上下文窗口限制和推理难度，性能显著下降。此外，我们展示了基于事件时间的JSON波形表示相比标准的值变化转储（VCD）格式能提升LLM的推理准确率。该开源框架支持扩展新的问题类别和导入新的波形数据源，助力研究者快速构建时间推理实验。

View on arXiv Download PDF AI Translation

cs.AI / 59 / 2607.20709

NVIDIA-labs OO Agents: Native Python Object-Oriented Agents

NVIDIA实验室OO代理：原生Python面向对象代理

Furgale, Paul, Klingler, Severin, Nolan, James, Staats, Matt, Di Lorenzo, Gaia, Abad, Elisa Martinez, Schüller, Christian, Dinu, Razvan, Devoto, Alessio, Berard, Pascal, Kaplun, Gal, Sarafian, Elad, Roveri, Riccardo, Derczynski, Leon, Cabral, Ricardo Silveira

Abstract

Traditional agent development is split across prompt templates, tool schemas, callback code, and workflow graphs. We present NVIDIA Object-Oriented Agents (NOOA), a model-agnostic Python framework for building reliable AI agents. NOOA takes a simpler approach: an agent is a Python object. Its methods are the actions the model can take, fields are its state, docstrings are its prompts, and its type annotations are contracts. A method whose code body consists of "..." is completed at runtime by an LLM-driven agent loop, while methods with normal bodies remain standard deterministic Python. This gives developers and agents the same interface, so agent behavior can be tested, traced, refactored, and improved just like other software. This paper makes three contributions. (1) We present the agent-as-a-Python-object programming model and the design principles behind it. Where Python has existing abstractions, we adopt them directly. Agent-specific capabilities--context, events, state rendering, long-term memory, and validated LLM loops--are exposed through simple Pythonic APIs, so both developers and agents share one familiar programming model. (2) We identify six model-facing ideas that NOOA is, to our knowledge, the first to combine on a single surface: typed input/output, pass-by-reference over live objects, code as action, programmable loop engineering, explicit object state, and model-callable harness APIs for context and events. We find the community already converging on several of these ideas--often as experimental or partial features--and present the comparison to encourage further adoption. (3) We demonstrate that current models use this interface effectively, both in targeted capability tests and on agentic and reasoning benchmarks such as SWE-bench Verified and Terminal-Bench 2.0 and ARC-AGI-3.

Chinese Translation

传统的代理开发分散在提示模板、工具架构、回调代码和工作流图中。我们提出了NVIDIA面向对象代理（NOOA），这是一个与模型无关的Python框架，用于构建可靠的AI代理。NOOA采用了一种更简单的方法：代理是一个Python对象。它的方法是模型可以采取的动作，字段是其状态，文档字符串是其提示，而类型注解是契约。代码主体由“...”组成的方法在运行时由一个基于LLM（大语言模型）的代理循环完成，而具有正常主体的方法保持标准的确定性Python。这使得开发者和代理拥有相同的接口，因此代理行为可以像其他软件一样被测试、追踪、重构和改进。本文做出了三项贡献。(1) 我们提出了代理作为Python对象的编程模型及其背后的设计原则。在Python已有的抽象中，我们直接采用它们。代理特定的能力——上下文、事件、状态渲染、长期记忆和经过验证的LLM循环——通过简单的Pythonic API暴露出来，因此开发者和代理共享一个熟悉的编程模型。(2) 我们识别出六个面向模型的思想，NOOA在我们所知的范围内首次将它们结合在一个单一的表面上：类型化输入/输出、对实时对象的引用传递、代码作为动作、可编程循环工程、显式对象状态，以及可供模型调用的上下文和事件的API。我们发现社区已经在多个这些思想上趋于一致——通常作为实验性或部分特性——并呈现比较以鼓励进一步采用。(3) 我们展示了当前模型有效地使用这一接口，无论是在针对能力的测试中，还是在代理和推理基准测试中，如SWE-bench Verified、Terminal-Bench 2.0和ARC-AGI-3。

View on arXiv Download PDF AI Translation

cs.AI / 60 / 2607.20764

ArbiGraph: Arbitrarily Scalable Verifiable Task Graphs for Evaluating Context Management

ArbiGraph：用于评估上下文管理的任意可扩展可验证任务图

Golikov, Pavel, Opryshko, Evgenii, Pekhimenko, Gennady, Jeffrey, Mark C.

Abstract

We introduce ARBIGRAPH, a benchmark generator for evaluating whether tool-assisted language agents can retain, update, compose, and discard task-relevant context across extended reasoning workflows. ARBIGRAPH represents each task as a natural-language problem with an executable Python solver, and composes tasks through typed intermediate states, instantiated here as scalar and list values. This design enables controllable task graphs whose length, dependency structure, distractor count, and value type can be varied while preserving exact automatic verification. We instantiate ARBIGRAPH with math, GSM-style word-problems, and Python-tracing task categories, and evaluate a Qwen3.5-27B tool-assisted agent across four topologies. The results show high accuracy on isolated tasks but substantial degradation on more complex dependent tasks: accuracy drops by up to 33.3% on branching chains of dependent math tasks. This shows that ARBIGRAPH exposes failures that are not visible from single-task evaluation alone. Our code, generated datasets, and evaluation results are available at https://github.com/pavelgolikov/ArbiGraph.git

Chinese Translation

我们介绍了 ARBIGRAPH，这是一个基准生成器，用于评估工具辅助语言代理是否能够在扩展推理工作流中保留、更新、组合和丢弃与任务相关的上下文。ARBIGRAPH 将每个任务表示为具有可执行 Python 求解器的自然语言问题，并通过类型化的中间状态组合任务，这里实例化为标量值和列表值。这种设计使得可控的任务图得以实现，其长度、依赖结构、干扰项数量和数值类型可以变化，同时保持精确的自动验证。我们用数学、GSM 风格的文字题和 Python 跟踪任务类别实例化了 ARBIGRAPH，并在四种拓扑结构中评估了 Qwen3.5-27B 工具辅助代理。结果显示，在孤立任务上具有高准确率，但在更复杂的依赖任务上显著下降：在分支链的依赖数学任务上，准确率下降高达 33.3%。这表明 ARBIGRAPH 揭示了单任务评估中不可见的失败。我们的代码、生成的数据集和评估结果可在 https://github.com/pavelgolikov/ArbiGraph.git 获取。

View on arXiv Download PDF AI Translation

cs.AI / 61 / 2607.20781

The Human-AI Substitution Principle: When will you be replaced by AI in your organization?

人类-人工智能替代原则：你何时会在组织中被人工智能取代？

Banerjee, Bonny, Singh, Shreya

Abstract

Artificial Intelligence (AI) is rapidly transforming organizations, raising a fundamental organizational and economic question: when will a human employee be replaced by AI? We present an analytical model for studying Human--AI Task Allocation (HAT) in hierarchical organizations. A central feature of the HAT model is that it formally encodes the economic asymmetry between human skill acquisition and AI capability scaling. The HAT model allows us to derive how risk-adjusted costs, skills, organizational depth, deployment scale, strategic adaptation, and risk jointly determine when, where, why, and under what structural conditions human--AI replacement occurs. A key result is the Human--AI Substitution Principle, which provides a precise condition --- grounded in the formal asymmetry assumption --- under which AI replaces human labor. Building on this result, we show that AI adoption can produce abrupt workforce transitions, hybrid human--AI organizations, including cases where risk heterogeneity sustains human and AI roles without requiring a minimum-human-fraction constraint, and flatter managerial hierarchies with wider spans of control. The HAT model identifies structural conditions under which middle-management roles exhibit elevated vulnerability to automation, and shows that the vulnerability of highly skilled workers depends on a skill threshold shaped by organizational depth, baseline costs, and risk differentials. More broadly, the paper connects automation economics, organizational design, AI governance, and workforce planning into a unified theory of AI-driven organizational transformation.

Chinese Translation

人工智能（AI）正在迅速改变组织，这引发了一个基本的组织和经济问题：人类员工何时会被人工智能取代？我们提出了一个分析模型，用于研究层级组织中的人类-人工智能任务分配（HAT）。HAT模型的一个核心特征是它正式编码了人类技能获取与人工智能能力扩展之间的经济不对称性。HAT模型使我们能够推导出风险调整成本、技能、组织深度、部署规模、战略适应性和风险如何共同决定人类-人工智能替代发生的时间、地点、原因以及在何种结构条件下发生。一个关键结果是人类-人工智能替代原则，它提供了一个精确条件——基于正式的不对称假设——在此条件下人工智能取代人类劳动。基于这一结果，我们展示了人工智能的采用可以导致突发的劳动力转型、混合的人类-人工智能组织，包括在风险异质性支持人类和人工智能角色而不需要最低人类比例约束的情况下，以及更扁平的管理层级和更广泛的控制跨度。HAT模型识别了中层管理角色在自动化中表现出较高脆弱性的结构条件，并表明高技能工人的脆弱性取决于由组织深度、基线成本和风险差异塑造的技能阈值。更广泛地说，本文将自动化经济学、组织设计、人工智能治理和劳动力规划连接成一个统一的人工智能驱动的组织转型理论。

View on arXiv Download PDF AI Translation

cs.AI / 62 / 2607.20791

Refusal-Gated Decoding: Preserving Refusal Behavior Under High-Temperature Sampling

拒绝门控解码：在高温采样下保持拒绝行为

Howard, Phillip, Su, Xin, Roush, Allen, Ravikiran, Manikandan, Abdullah, Amir

Abstract

High-temperature sampling is one of the primary mechanisms for increasing diversity in LLMs. Recent advances in truncation-based sampling techniques have helped mitigate drawbacks of high-temperature sampling such as neural text degeneration, thereby enabling greater diversity in LLM outputs without sacrificing coherence. However, increasing the entropy of the token probability distribution via high temperatures has also been shown to weaken model guardrails by reducing the model's refusal response in the presence of harmful prompts. Despite the potential benefits of high-temperature sampling and the importance of maintaining model safety, there is a lack of existing solutions for maintaining the refusal behavior of LLMs under a higher entropy regime. To address this gap, we systematically study how temperature influences refusal behavior in LLMs and propose an efficient sequential decoding approach which preserves a model's greedy decoding refusal response at high temperatures while incurring minimal additional latency. Through extensive experiments, we show that our approach preserves 91-99% of the greedy decoding refusal behavior across three benchmark datasets without compromising the model's high-temperature response for safe prompts. Our work demonstrates how refusal behavior can be maintained in an efficient manner for applications which require high-temperature sampling.

Chinese Translation

高温采样是增加大型语言模型（LLMs）多样性的主要机制之一。最近基于截断的采样技术的进展帮助减轻了高温采样的缺点，如神经文本退化，从而在不牺牲连贯性的情况下实现了更大的多样性。然而，通过高温增加标记概率分布的熵也被证明会削弱模型的保护机制，减少模型在面对有害提示时的拒绝响应。尽管高温采样具有潜在的好处，并且保持模型安全性至关重要，但在更高熵状态下保持LLMs的拒绝行为的现有解决方案仍然缺乏。为了解决这一问题，我们系统地研究了温度如何影响LLMs的拒绝行为，并提出了一种高效的序列解码方法，该方法在高温下保持模型的贪婪解码拒绝响应，同时引入的额外延迟最小。通过大量实验，我们展示了我们的方法在三个基准数据集上保留了91-99%的贪婪解码拒绝行为，而不影响模型对安全提示的高温响应。我们的工作展示了如何以高效的方式保持拒绝行为，以满足需要高温采样的应用。

View on arXiv Download PDF AI Translation

cs.AI / 63 / 2607.20796

Can an AI System Be Creative? A Critical Perspective from Art and Engineering

人工智能系统能否具备创造力？来自艺术与工程的批判视角

Magrin-Chagnolleau, Ivan

Abstract

This paper examines the question of whether artificial intelligence (AI) systems can be creative, approached from the dual perspective of a researcher trained in electrical engineering, pattern recognition, machine learning, and neural networks, who has also spent most of his life engaged in the arts as actor, stage and film director, writer, composer, and visual artist, and in philosophy. Drawing on Margaret Boden's foundational framework, both her three properties of creativity (novelty, surprise, and value) and her three types of creative processes (combinatorial, exploratory, and transformational), the paper argues that AI systems are structurally incapable of creativity in its strongest sense. While they exhibit genuine capability in the domain of combinatorial creativity, they are significantly bounded in exploratory creativity, and fundamentally incapable of transformational creativity. The paper further argues that the most important limitation of current AI systems is not the absence of novelty per se, but the absence of any mechanism for serendipity, accident, or the unexpected, all of which play a central role in the phenomenology of creativity, and the absence of any subject position from which to recognize and welcome such chance events. The paper concludes by proposing a model of human, AI creative collaboration that is both realistic and generative, illustrated by several concrete experiments. The paper is itself a demonstration of the thesis it advances: it was composed through a deliberate human AI collaborative process, which is described in the methodological note that opens it.

Chinese Translation

本文探讨了人工智能（AI）系统是否能够具备创造力这一问题，从一位受过电气工程、模式识别、机器学习和神经网络训练的研究者的双重视角出发，该研究者在艺术领域也有着丰富的经历，包括演员、舞台和电影导演、作家、作曲家以及视觉艺术家，并且在哲学方面有所研究。基于玛格丽特·博登（Margaret Boden）的基础框架，本文分析了她所提出的创造力的三种特性（新颖性、惊讶性和价值）以及三种创造性过程（组合性、探索性和变革性），认为AI系统在其最强意义上的创造力是结构上无法实现的。尽管它们在组合性创造力领域展现了真实的能力，但在探索性创造力方面却受到显著限制，而在变革性创造力方面则根本无法实现。本文进一步论证了当前AI系统最重要的限制并非缺乏新颖性本身，而是缺乏任何偶然性、意外性或不可预见性的机制，这些因素在创造力的现象学中扮演着核心角色，同时也缺乏任何能够识别和欢迎这些偶然事件的主体位置。最后，本文提出了一种人类与AI创造性协作的模型，该模型既现实又具生成性，并通过若干具体实验进行说明。本文本身就是其所提出论点的一个示范：它是通过一个有意的人类与AI协作的过程创作而成，相关方法论说明在开头部分进行了描述。

View on arXiv Download PDF AI Translation

cs.AI / 64 / 2607.20806

Profiling Lightweight Large Language Models

轻量级大语言模型的特征分析

Harada, Tomohiro, Alba, Enrique, Luque, Gabriel

Abstract

Lightweight large language models (LLMs) are increasingly being deployed locally on personal computers and are expected to play a growing role in resource-constrained edge and mobile environments. In such settings, energy consumption, execution time, and memory usage directly affect practical usability, yet existing evaluations of LLM efficiency largely rely on proxy descriptors such as parameter count or FLOPs, often decoupled from task precision. This paper introduces a PTME-based experimental framework for the precision-aware profiling of lightweight LLM inference, jointly measuring Precision, execution Time, peak Memory usage, and Energy consumption through direct hardware-level measurements. The methodology is applied to a representative set of lightweight LLMs executed locally under edge-class resource envelopes on a controlled desktop platform, using benchmarks spanning code generation, mathematical reasoning, and multi-task understanding. We find that static proxy descriptors approximate inference cost well but fail to predict precision. Tightening the resource envelope increases cost without affecting precision, amplifying execution time more strongly than energy and penalizing larger models the most. Moreover, no single model dominates across all PTME dimensions, and a Pareto analysis reveals non-dominated configurations that would be hidden by accuracy-only or efficiency-only assessments, providing practical guidance for selecting models under different resource envelopes. These results show that selecting lightweight LLMs by size, FLOPs, latency, or accuracy alone can select the wrong deployment candidate; PTME profiling exposes configurations that preserve useful accuracy at lower physical cost.

Chinese Translation

轻量级大语言模型（LLMs）越来越多地在个人计算机上本地部署，并预计将在资源受限的边缘和移动环境中发挥越来越重要的作用。在这种环境中，能耗、执行时间和内存使用直接影响实际可用性，但现有的LLM效率评估主要依赖于参数数量或FLOPs等代理描述符，这些描述符往往与任务精度脱节。本文提出了一种基于PTME的实验框架，用于精度感知的轻量级LLM推理特征分析，通过直接的硬件级测量共同测量精度、执行时间、峰值内存使用和能耗。该方法应用于一组代表性的轻量级LLM，这些模型在受控桌面平台上以边缘级资源限制进行本地执行，基准测试涵盖代码生成、数学推理和多任务理解。我们发现，静态代理描述符能够很好地近似推理成本，但无法预测精度。收紧资源限制会增加成本而不影响精度，执行时间的增加幅度明显高于能耗，并对较大的模型影响最大。此外，没有单一模型在所有PTME维度上占据主导地位，帕累托分析揭示了在仅依赖精度或效率评估时会被忽视的非支配配置，为在不同资源限制下选择模型提供了实用指导。这些结果表明，仅通过模型大小、FLOPs、延迟或精度选择轻量级LLM可能会选错部署候选；PTME特征分析揭示了在较低物理成本下保持有用精度的配置。

View on arXiv Download PDF AI Translation

cs.AI / 65 / 2607.20814

Enhancing Explainable Cardiac Diagnosis with Guide-Grounded Multimodal LLMs

通过引导基础的多模态大语言模型增强可解释的心脏诊断

Vuong, Hai-Nam Duy, Bui, Duy-Anh, Nguyen, Trong-Nghia, Nguyen, Kim-Ngan Thi, Xuan, Trang Mai, Nguyen, Tien-Cuong, Pham, Van-Dem, Van Luong, Thien

Abstract

The electrocardiogram (ECG) is a cornerstone of cardiac as- sessment, yet clinical deployment of deep learning models remains con- strained by limited interpretability and the hallucination risk of large language models (LLMs). Existing CNN+Grad-CAM+multimodal LLM frameworks can generate ECG reports, but their explanations are often only weakly grounded in established diagnostic criteria, reducing trust- worthiness and reproducibility. We propose a guide-grounded multimodal framework that explicitly anchors report generation in curated clinical knowledge. A convolutional neural network (CNN) and Grad-CAM first produce class probabilities and class-specific heatmaps from 12-lead ECG images. In parallel, authoritative ECG textbooks and guideline materials are distilled offline into a structured ECG Interpretation Guide, which is injected as a fixed knowledge block for every sample. Conditioned on the ECG image, Grad-CAM overlay, CNN-derived fact pack, and the in- jected guide, a multimodal LLM generates structured diagnostic reports with guideline-consistent terminology and criteria usage. Experiments on the full PTB-XL test set demonstrate that guide grounding improves se- mantic quality and perceived consistency of generated reports while pre- serving competitive classification performance. In particular, our method increases the average BERTScore of generated impressions from 0.818 to 0.953 relative to a strong CNN+Grad-CAM+MLLM baseline, indicat- ing closer alignment with reference reports. These findings suggest that injecting a distilled interpretation guide into the multimodal prompting pipeline offers a practical pathway to reduce hallucinations and enhance the clinical plausibility of LLM-based ECG explanations, bringing ex- plainable cardiac diagnosis closer to real-world deployment.

Chinese Translation

心电图（ECG）是心脏评估的基石，但深度学习模型在临床应用中的部署仍受到有限可解释性和大型语言模型（LLMs）幻觉风险的限制。现有的CNN+Grad-CAM+多模态LLM框架能够生成ECG报告，但其解释往往仅在既定诊断标准中弱相关，从而降低了可信度和可重复性。我们提出了一种引导基础的多模态框架，明确将报告生成锚定在经过筛选的临床知识上。卷积神经网络（CNN）和Grad-CAM首先从12导联ECG图像中生成类别概率和类别特定热图。同时，权威的ECG教科书和指南材料被离线提炼为结构化的ECG解读指南，并作为固定知识块注入每个样本中。在ECG图像、Grad-CAM叠加图、CNN生成的事实包和注入的指南的条件下，多模态LLM生成结构化的诊断报告，使用与指南一致的术语和标准。对完整PTB-XL测试集的实验表明，引导基础提高了生成报告的语义质量和感知一致性，同时保持了竞争性的分类性能。特别是，我们的方法使生成印象的平均BERTScore从0.818提高到0.953，相较于强大的CNN+Grad-CAM+MLLM基线，表明与参考报告的对齐更为紧密。这些发现表明，将提炼的解读指南注入多模态提示管道提供了一条实用路径，以减少幻觉并增强基于LLM的ECG解释的临床合理性，使可解释的心脏诊断更接近于实际应用。

View on arXiv Download PDF AI Translation

cs.AI / 66 / 2607.20820

Efficient and Interpretable Body-Based Emotion Recognition with Lightweight Temporal Convolutional Networks

基于身体的高效且可解释的情感识别：轻量级时间卷积网络的应用

Cruz, Christian Arzate, Gkikas, Stefanos, Asadi, Houshyar

Abstract

Body-based emotion recognition is important for real-time affective systems, but graph-based skeleton models can be computationally expensive. This paper studies whether lightweight temporal convolutional networks (TCNs) can provide an efficient and interpretable alternative for body-based emotion classification. We evaluate a family of TCN models on DIEM-A and compare them with a graph-based time-series graph (G-TSG) baseline using accuracy, macro-F1, parameter count, and inference latency. Although G-TSG achieves the highest mean performance, TCN-Base remains within $1.58$ accuracy points and $1.25$ macro-F1 points while using $79.18\%$ fewer parameters and reducing classifier latency by approximately $12.5\times$. We also analyze body-region contributions using region-specific TCN models, zero-based occlusion, and G-TSG gradient saliency. The results show that upper-body motion provides the strongest standalone regional cue, that the usefulness of body regions varies across emotions, and that different interpretability methods capture distinct aspects of model behavior. These findings suggest that lightweight TCNs can support efficient body-based emotion recognition while also providing practical insight into how motion cues contribute to classification.

Chinese Translation

基于身体的情感识别对于实时情感系统至关重要，但基于图的骨架模型可能计算开销较大。本文研究轻量级时间卷积网络（Temporal Convolutional Networks, TCNs）是否可以为基于身体的情感分类提供一种高效且可解释的替代方案。我们在DIEM-A数据集上评估了一系列TCN模型，并将其与基于图的时间序列图（Graph-based Time-Series Graph, G-TSG）基线进行比较，使用准确率、宏观F1值、参数数量和推理延迟作为评估指标。尽管G-TSG实现了最高的平均性能，但TCN-Base的准确率仅比其低$1.58$点，宏观F1值低$1.25$点，同时使用的参数减少了$79.18 ext{ extperthousand}$，分类器延迟大约减少了$12.5 imes$。我们还使用区域特定的TCN模型、零基础遮挡和G-TSG梯度显著性分析了身体区域的贡献。结果表明，上半身运动提供了最强的独立区域线索，身体区域的有效性在不同情感间存在差异，并且不同的可解释性方法捕捉到模型行为的不同方面。这些发现表明，轻量级TCNs可以支持高效的基于身体的情感识别，同时提供关于运动线索如何贡献于分类的实际见解。

View on arXiv Download PDF AI Translation

cs.AI / 67 / 2607.20827

Auditing Provenance Sensitivity in LLM Agent Action Selection

审计大型语言模型代理行动选择中的来源敏感性

Liao, Junchi

Abstract

LLM agents choose tools and arguments from context that mixes user requests, tool outputs, retrieved records, memory, and untrusted text. Evidence can be relevant without being authorized to determine a decision, so a correct action need not be grounded only in permitted evidence. We introduce a target-specific authorization audit that labels context factors separately for each tool and argument target. Its primary test holds the task, proposition, position, and policy fixed while changing only the proposition's source authority. We then test behavior when valid evidence is weakened and use context-subset interactions as a secondary localization diagnostic. Across 450 controlled next-action tasks and multiple open-weight LLM families, trusted and untrusted variants produce different actions in 5.4 percent of competing cases versus 1.7 percent of supporting cases. Under controlled degradation, unauthorized competition is retained in a full-correct, mixed-error, clean-correct pattern in 2.4 percent of comparisons, with a 95 percent confidence interval from 2.1 to 3.0 percent. These are controlled stress-set rates, not deployment prevalence. The models respond to textual source-authority cues, but this does not prevent untrusted evidence from influencing their actions.

Chinese Translation

大型语言模型（LLM）代理从混合用户请求、工具输出、检索记录、记忆和不可信文本的上下文中选择工具和论据。证据可能与决策相关，但并不一定被授权，因此正确的行动不必仅基于被允许的证据。我们引入了一种针对特定目标的授权审计，分别为每个工具和论据目标标记上下文因素。其主要测试保持任务、命题、立场和政策不变，仅改变命题的来源权威。然后，我们测试在有效证据减弱时的行为，并使用上下文子集交互作为次要定位诊断。在450个受控的下一步行动任务和多个开放权重的LLM家族中，可信和不可信的变体在5.4%的竞争案例中产生不同的行动，而在1.7%的支持案例中则如此。在受控降级下，未经授权的竞争在2.4%的比较中保持全正确、混合错误、干净正确的模式，95%的置信区间为2.1%至3.0%。这些是受控压力集的比率，而非部署普遍性。模型对文本来源权威线索作出反应，但这并不阻止不可信证据影响其行动。

View on arXiv Download PDF AI Translation

cs.AI / 68 / 2607.20848

Auditing Evidence Use in Medical LLM Diagnosis

医学大型语言模型诊断中的证据使用审计

Liao, Junchi, Deng, Jiawen, Ren, Fuji

Abstract

Medical LLMs are often evaluated by whether they select the correct diagnosis, but diagnostic accuracy alone does not show whether the model used the case evidence appropriately. We present a behavioral audit of evidence use in medical diagnosis. For each case, we decompose patient information into evidence units, score candidate diagnoses under controlled evidence subsets, and mine low-order interactions in diagnostic margins. Because medical evidence is diagnosis-relative, the audit separates interaction discovery from failure assignment: large or negative interactions can reflect plausible differential diagnosis, while suspicious interactions require robustness checks and clinical review. We evaluate five open-weight LLMs on DDXPlus, CupCase, and MedCase. Across datasets, faithful support and differential conflict or cancellation account for most interaction strength, showing that many evidence interactions are clinically plausible rather than failures. In a DDXPlus-focused blinded five-reviewer 130-item enriched review sample, invalid or shortcut-like cases concentrate in negated or absent findings and clinically local evidence. These results show that accuracy can hide candidate evidence-use failures and motivate role-aware audits for medical LLM evaluation.

Chinese Translation

医学大型语言模型（LLMs）通常通过其选择正确诊断的能力进行评估，但仅靠诊断准确性并不能显示模型是否适当地使用了病例证据。我们提出了一种针对医学诊断中证据使用的行为审计方法。对于每个病例，我们将患者信息分解为证据单元，在受控证据子集下对候选诊断进行评分，并挖掘诊断边际中的低阶交互。由于医学证据是与诊断相关的，审计将交互发现与失败分配分开：较大或负的交互可能反映合理的差异诊断，而可疑的交互则需要进行稳健性检查和临床审查。我们在DDXPlus、CupCase和MedCase上评估了五个开放权重的LLMs。在各数据集中，忠实支持和差异冲突或取消占据了大多数交互强度，表明许多证据交互在临床上是合理的，而非失败。在一个以DDXPlus为重点的盲审五位评审者的130项增强审查样本中，无效或类似捷径的案例集中在否定或缺失的发现以及临床局部证据中。这些结果表明，准确性可能掩盖候选证据使用的失败，并激励针对医学LLM评估的角色意识审计。

View on arXiv Download PDF AI Translation

cs.AI / 69 / 2607.20852

Code Monitor Red Teaming for Public-Test-Passing Code

公共测试通过代码的代码监控红队测试

Liao, Junchi, Deng, Jiawen, Ren, Fuji

Abstract

Visible tests are a common gate for LLM-generated code, but passing them does not certify specification correctness. We study a deployment-like monitoring problem: after code has passed public tests, can a weaker LLM verifier identify the residual hidden bugs? We introduce Code Monitor Red Teaming, a monitor-red-teaming protocol that fixes a public-check information boundary while varying generator pressure, verifier scaffolding, and weak-to-strong capability. We instantiate it as CodeMonitorBench, spanning function-level, data-science, and workflow code. Across 71,000 generated candidates, 43,677 pass public tests and 23,081 of those fail hidden tests. Weak verifiers improve with scaffolding and model family, but still miss most hidden bugs at 5% false-positive rate. As a robustness stress test, adversarial public-test-overfit pressure lowers verifier AUROC and raises low-FPR miss rates in most cells. A GLM-5.1 verifier recovers part of the gap under the same evidence boundary; an inferability audit shows that remaining misses mix verifier failures with M1 evidence limits.

Chinese Translation

可见测试是大型语言模型（LLM）生成代码的一个常见门槛，但通过这些测试并不能证明规范的正确性。我们研究了一个类似于部署的监控问题：在代码通过公共测试后，是否可以通过一个较弱的 LLM 验证器识别残留的隐藏错误？我们引入了代码监控红队测试（Code Monitor Red Teaming），这是一种监控-红队测试协议，它修复了公共检查的信息边界，同时调整生成器压力、验证器支架和弱到强的能力。我们将其实例化为 CodeMonitorBench，涵盖函数级、数据科学和工作流代码。在71,000个生成的候选代码中，43,677个通过了公共测试，而其中23,081个在隐藏测试中失败。弱验证器在支架和模型家族的帮助下有所改善，但在5%的假阳性率下仍然错过大多数隐藏错误。作为一种鲁棒性压力测试，对抗性公共测试过拟合压力降低了验证器的AUROC，并在大多数单元中提高了低假阳性率的漏检率。在相同证据边界下，GLM-5.1验证器恢复了部分差距；可推导性审计显示，剩余的漏检将验证器失败与M1证据限制混合在一起。

View on arXiv Download PDF AI Translation

cs.AI / 70 / 2607.20891

Is Deep Research Reliable? Misleading Knowledge Induces False Conclusions

深度研究可靠性如何？误导性知识导致错误结论

Zhu, Pengyu, Li, Lijun, Yang, Longju, Su, Sen

Abstract

Deep Research agents extend LLM-based assistants into long-horizon workflows involving planning, retrieval, evidence synthesis, and report generation, yet their reliability in open information environments remains underexplored. A key concern is whether apparently credible but factually misleading knowledge encountered in such environments can propagate through these workflows and be adopted as false conclusions in final reports. To study this failure mode, we introduce MisKnow-Agent, a framework for constructing and validating misleading knowledge for Deep Research tasks. MisKnow-Agent generates misleading instances with controllable authority levels and styles, yielding 5,933 quality-controlled instances built on DeepResearch Benchmark tasks. Extensive experiments across open-source and closed-source Deep Research agents show that even limited exposure to misleading knowledge can induce false-conclusion adoption in final reports, revealing a broad reliability vulnerability in current Deep Research agents. Although search-enabled verifier models consistently identify the retained instances as misleading during focused corpus validation, the same instances can still be adopted during long-horizon research, revealing a disconnect between focused verification and workflow-level evidence use. Finally, we evaluate pre- and post-research defenses, both individually and in combination, finding that all three configurations mitigate but do not fully prevent false-conclusion adoption. Our findings suggest that reliable Deep Research requires evidence verification and correction capabilities at both the model and framework levels, beyond improvements in planning, retrieval, evidence integration, or report-generation abilities.

Chinese Translation

深度研究代理将基于大型语言模型（LLM）的助手扩展到涉及规划、检索、证据综合和报告生成的长期工作流程中，但它们在开放信息环境中的可靠性仍然未得到充分探讨。一个关键问题是，在这些环境中遇到的表面上可信但在事实上的误导性知识是否会在这些工作流程中传播，并在最终报告中被采纳为错误结论。为了研究这种失败模式，我们引入了MisKnow-Agent，这是一个用于构建和验证深度研究任务中误导性知识的框架。MisKnow-Agent生成具有可控权威水平和风格的误导性实例，基于DeepResearch基准任务构建了5,933个质量控制实例。对开源和闭源深度研究代理的广泛实验表明，即使是有限接触误导性知识，也会在最终报告中引发错误结论的采纳，揭示了当前深度研究代理的广泛可靠性脆弱性。尽管启用搜索的验证模型在集中语料验证过程中始终能够识别出保留的实例为误导性，但在长期研究中，这些实例仍然可能被采纳，揭示了集中验证与工作流程层面证据使用之间的脱节。最后，我们评估了研究前和研究后的防御措施，既单独评估也组合评估，发现这三种配置均能减轻但并未完全阻止错误结论的采纳。我们的研究结果表明，可靠的深度研究需要在模型和框架层面具备证据验证和纠正能力，而不仅仅是规划、检索、证据整合或报告生成能力的提升。

View on arXiv Download PDF AI Translation

cs.AI / 71 / 2607.20913

Source-Prior-Driven Selective Adaptation for Efficient Diffusion Model Finetuning

基于源先验的选择性适应方法用于高效的扩散模型微调

Xiong, Yi, Cheng, Yuan-Yuan, Fu, Xiao-Ming

Abstract

Fine-tuning large diffusion models for new domains or styles involves a trade-off: improving target-specific generation often degrades the pretrained model's broad generative capability. Existing full and parameter-efficient fine-tuning methods typically handle this trade-off only implicitly. In this work, we propose a novel source-prior-driven selective adaptation method to efficiently fine-tune diffusion models, achieving a favorable trade-off. Our method relies on two key observations: (1) the loss of general generative capability is highly inconsistent across pretrained parameters, and (2) parameters that have a relatively small impact on the model's general generative capability remain structurally inconsistent across layers and parameter types. Motivated by these observations, we first learn a static mask to explicitly identify parameters better suited for downstream adaptation, and then construct structured update strategies for the selected subset. Experiments show that our method achieves a better adaptation-retention trade-off than existing strong baselines.

Chinese Translation

对大型扩散模型进行微调以适应新领域或风格涉及一个权衡：提高特定目标的生成能力往往会降低预训练模型的广泛生成能力。现有的全参数和高效参数微调方法通常仅隐式地处理这一权衡。在本研究中，我们提出了一种新颖的基于源先验的选择性适应方法，以高效地微调扩散模型，实现有利的权衡。我们的方法依赖于两个关键观察：(1) 一般生成能力的损失在预训练参数之间高度不一致，(2) 对模型一般生成能力影响相对较小的参数在层和参数类型之间保持结构不一致。基于这些观察，我们首先学习一个静态掩码，以明确识别更适合下游适应的参数，然后为所选子集构建结构化更新策略。实验表明，我们的方法在适应与保留的权衡上优于现有的强基线。

View on arXiv Download PDF AI Translation

cs.AI / 72 / 2607.20916

Traceable Scholarship: Page Anchors and Ariadne's Thread for Humanistic Inquiry in the Age of Generative AI

可追溯的学术研究：生成性人工智能时代人文学科探究的页面锚和阿里阿德涅之线

Jing, Deyu

Abstract

Generative AI lets large language models produce scholarly-looking text within seconds, yet fluency does not equal valid explanation. The deepest risk is not factual error alone but the appearance that an explanation is already established without clear sources, page numbers, editions, or evidence. We liken the page anchor to Ariadne's thread: within the labyrinth of generative fluency, it is the thread that leads the scholar back to the source. This paper proposes Traceable Scholarship as the minimum normative condition for AI-assisted humanistic research, situating it across the three revolutions of knowledge infrastructure: print, digital, and generative AI. We introduce page anchors, dual page numbers, citation-first generation, NO_EVIDENCE, human verification, four-level compliance, and Scope Contract, and present AIH-Infra as a three-layer reference implementation: Contexture (document structuring), Open WebUI AIH-Infra (traceable knowledge base), and AIH-Infra MCP Server (agent gateway). A case study on a 29-volume Kant Akademie-Ausgabe knowledge base illustrates how traceability supports retrieval correction, evidence grading, and judgment downgrading. Traceability is not a software feature; it is the condition under which humanistic research can remain public and refutable in the age of generative AI.

Chinese Translation

生成性人工智能使大型语言模型能够在几秒钟内生成看似学术的文本，但流畅性并不等同于有效的解释。最深层的风险不仅在于事实错误，而在于一种解释似乎已经建立，但却没有明确的来源、页码、版本或证据。我们将页面锚比作阿里阿德涅之线：在生成性流畅性的迷宫中，它是引导学者回归来源的线索。本文提出可追溯的学术研究作为AI辅助人文学科研究的最低规范条件，并将其置于知识基础设施的三次革命中：印刷、数字和生成性人工智能。我们引入页面锚、双页码、引用优先生成、无证据（NO_EVIDENCE）、人工验证、四级合规和范围合同，并呈现AIH-Infra作为三层参考实现：Contexture（文档结构）、Open WebUI AIH-Infra（可追溯知识库）和AIH-Infra MCP Server（代理网关）。对29卷《康德学术版》（Kant Akademie-Ausgabe）知识库的案例研究说明了可追溯性如何支持检索纠正、证据评分和判断降级。可追溯性不是一种软件特性；它是在人文学科研究在生成性人工智能时代保持公共性和可反驳性的条件。

View on arXiv Download PDF AI Translation

cs.AI / 73 / 2607.20918

OPOD: On-Policy Omni Distillation

OPOD：基于策略的全模态蒸馏

Zhao, Tong, Hu, Yuyang, Li, Reed, Lu, Yu, Shi, Haibo, Zhu, Yutao, Dou, Zhicheng

Abstract

Omni-modal models can handle text, images, and audio in one system, but improving all of these abilities together remains difficult. Training a single model on pooled multimodal data often fails to match models specialized for individual modalities. On-policy distillation (OPD) offers a way to combine such specialists: the student generates a response, and a teacher evaluates that same response, so the student learns directly from behaviors it actually produces. Yet using several teachers can introduce competing guidance and improve one modality at the expense of another. We present On-Policy Omni Distillation (OPOD), which routes each student response to the matching text, image, or audio teacher. OPOD keeps teacher guidance only on tokens where the teacher assigns a higher probability than the student, adjusts the influence of each modality teacher independently during training, and asks the routed teacher to assess both the final answer and whether the reasoning supports the correct answer. Across twelve benchmarks and three backbone sizes, OPOD achieves the best average score at every scale, reaching 70.8, 51.7, and 46.2 and exceeding the strongest comparator by 2.1, 1.8, and 1.7 points. On the 30B model, it outperforms both the base model and a counterpart post-trained jointly on pooled multimodal data on all twelve benchmarks, and ranks first or second on eleven even when the individual specialists are included. The specialists are discarded after training, leaving one deployable omni-modal model. These results show that coordinating modality-specific teachers is an effective way to improve a shared model while maintaining cross-modal balance.

Chinese Translation

全模态模型能够在一个系统中处理文本、图像和音频，但同时提升这些能力仍然困难。在汇总的多模态数据上训练单一模型通常无法与针对单一模态的专门模型相匹配。基于策略的蒸馏（On-policy Distillation, OPD）提供了一种结合这些专门模型的方法：学生生成响应，教师评估该响应，从而使学生直接从其实际产生的行为中学习。然而，使用多个教师可能会引入相互竞争的指导，并在提升一种模态的同时牺牲另一种模态。我们提出了基于策略的全模态蒸馏（On-Policy Omni Distillation, OPOD），它将每个学生的响应路由到匹配的文本、图像或音频教师。OPOD仅在教师对某个标记分配的概率高于学生时保留教师的指导，独立调整每个模态教师在训练过程中的影响，并要求路由的教师评估最终答案及其推理是否支持正确答案。在十二个基准测试和三种基础模型规模上，OPOD在每个规模上都达到了最佳平均分，分别为70.8、51.7和46.2，超越了最强对比模型2.1、1.8和1.7分。在30B模型上，它在所有十二个基准测试中均优于基础模型和在汇总多模态数据上共同后训练的对应模型，并且即使在包含单独专门模型的情况下也排名第一或第二。训练后，专门模型被丢弃，留下一个可部署的全模态模型。这些结果表明，协调模态特定的教师是一种有效的方法，可以在保持跨模态平衡的同时提升共享模型的性能。

View on arXiv Download PDF AI Translation

cs.AI / 74 / 2607.20925

Representing Entity Importance in AI Knowledge Systems: A Dual-Signal Framework of Audience Evaluation and Structural Authority

在人工智能知识系统中表示实体重要性：观众评估与结构权威的双信号框架

Xu, Shen

Abstract

AI knowledge systems require representations of entity importance for retrieval, recommendation, evidence selection, and knowledge-intensive reasoning. Yet importance is often reduced to a single score derived from either human response or graph structure. Such compression may discard distinctions that matter when an AI system must choose among entities for different tasks. This study introduces an interpretable dual-signal representation in which each entity is characterized by an audience-evaluation dimension and a structural-authority dimension. The framework is evaluated using movie entities as an empirical validation domain. IMDb non-commercial datasets provide a rating-based audience ranking, Wikidata supports entity alignment, and English Wikipedia hyperlinks form the knowledge network on which PageRank estimates structural authority. Experiments on 482 entities and 13,690 directed relationships reveal a statistically significant but weak association between the two dimensions (Spearman rho = 0.2275, p < 0.001). Their overlap is only 10% in the top 10 and 34% in the top 100, while entity-level divergence occurs in both directions. The results show that audience evaluation and structural authority are non-redundant signals and should not automatically be collapsed into a single scalar notion of importance. The contribution is not a new ranking algorithm or learned embedding, but a minimal knowledge-representation framework and an empirical test of its dimensional necessity. The findings support task-aware AI knowledge systems that preserve distinct importance signals before applying context-specific selection or aggregation.

Chinese Translation

人工智能知识系统需要对实体重要性进行表示，以便于检索、推荐、证据选择和知识密集型推理。然而，重要性通常被简化为一个单一的评分，这个评分要么来源于人类的反馈，要么来源于图结构。这种压缩可能会忽略在人工智能系统需要为不同任务选择实体时的重要区分。本研究引入了一种可解释的双信号表示，其中每个实体由观众评估维度和结构权威维度来表征。该框架通过电影实体作为实证验证领域进行评估。IMDb非商业数据集提供了基于评分的观众排名，Wikidata支持实体对齐，而英文维基百科的超链接构成了PageRank估算结构权威的知识网络。对482个实体和13,690个有向关系的实验结果显示，这两个维度之间存在统计显著但较弱的关联（Spearman rho = 0.2275, p < 0.001）。在前10名中，它们的重叠仅为10%，在前100名中为34%，而在实体层面上则出现双向的差异。结果表明，观众评估和结构权威是非冗余信号，不应自动合并为单一的标量重要性概念。本研究的贡献不是一个新的排名算法或学习嵌入，而是一个最小知识表示框架及其维度必要性的实证测试。这些发现支持任务感知的人工智能知识系统，在进行上下文特定的选择或聚合之前，保留不同的重要性信号。

View on arXiv Download PDF AI Translation

cs.AI / 75 / 2607.20926

SciExplore: Evaluating Autonomous Agents from Scientific Navigation to Information Integration

SciExplore：从科学导航到信息整合评估自主智能体

Tang, Yinhao, Fang, Youqing, Sun, Yanan, Liu, Wenran, Zhang, Weiming, Liu, Bin, Liu, Kuikun, Zhang, Wenwei, Chen, Kai

Abstract

Scientific research involves complex information-seeking and reasoning workflows across heterogeneous sources. However, existing benchmarks primarily emphasize general-domain retrieval or static scientific question answering, and therefore fail to assess key capabilities required in realistic scientific research workflows. We introduce SciExplore, a benchmark designed to evaluate scientific information-seeking and reasoning capabilities of LLMs and agents. SciExplore comprises four task types covering 103 expert-curated tasks across more than ten scientific disciplines: scientific database navigation, ambiguous literature retrieval, missing reference completion, and cross-source structured knowledge synthesis, which probe progressively higher-level abilities from entity-level reasoning and document-level identification to evidence-level grounding and domain-level synthesis. We evaluate over ten state-of-the-art LLMs and autonomous agents on SciExplore, revealing substantial performance gaps with performance degrading sharply as task complexity increases and extremely low accuracy on the most challenging structured synthesis tasks. These results highlight significant limitations of current models and agents in realistic scientific information-seeking scenarios.

Chinese Translation

科学研究涉及跨异构来源的复杂信息搜索和推理工作流程。然而，现有的基准主要强调一般领域的检索或静态科学问题回答，因此未能评估在现实科学研究工作流程中所需的关键能力。我们引入了SciExplore，这是一个旨在评估大型语言模型（LLMs）和智能体的科学信息搜索和推理能力的基准。SciExplore包含四种任务类型，涵盖了十多个科学学科中的103个专家策划任务：科学数据库导航、模糊文献检索、缺失参考补全和跨源结构化知识综合，这些任务逐步探测从实体级推理和文档级识别到证据级基础和领域级综合的更高层次能力。我们在SciExplore上评估了十多种最先进的LLMs和自主智能体，揭示了显著的性能差距，随着任务复杂性的增加，性能急剧下降，并且在最具挑战性的结构化综合任务上准确率极低。这些结果突显了当前模型和智能体在现实科学信息搜索场景中的重大局限性。

View on arXiv Download PDF AI Translation

cs.AI / 76 / 2607.20937

Clustered Edge Intelligence: Beyond Just Convergence of Edge Computing and AI

集群边缘智能：超越边缘计算与人工智能的简单融合

Dehury, Chinmaya Kumar, Sedlak, Boris, Saleh, Alaa, Murturi, Ilir, Loven, Lauri, Srirama, Satish Narayana, Donta, Praveen Kumar

Abstract

We are moving from an information age to the age of intelligence. A decade, or possibly less than that, data will not be the gold anymore rather the derived intelligence out of the data and the information we posses from the edge of the network. Existing Edge Intelligence research focuses mainly on two directions: using AI for edge resource management and deploying lightweight AI models on edge devices. However, existing edge computing research lacks an intelligence-centric framework in which derived intelligence is treated as a first-class, independently manageable entity that can be described, discovered, observed, shared, reused, and dynamically clustered across heterogeneous edge devices and applications. To address these research gaps, we introduced Clustered Edge Intelligence, a visionary intelligence-centric approach. The aim of CEI is to make intelligence a shareable and reusable first-class entity that can be independently represented, discovered, observed, exchanged, and managed across the distributed edge-cloud continuum. We present a three layer CEI architecture and examine enabling technologies and research dimensions, including intelligence inventories, semantic knowledge representation, communication, discoverability, observability, lifecycle automation, clustering mechanisms, marketplaces, interoperability, and standardization.

Chinese Translation

我们正从信息时代迈向智能时代。在十年内，或者可能更短的时间里，数据将不再是黄金，而是我们从网络边缘所拥有的数据和信息中提取出的智能。现有的边缘智能研究主要集中在两个方向：利用人工智能进行边缘资源管理以及在边缘设备上部署轻量级人工智能模型。然而，现有的边缘计算研究缺乏一个以智能为中心的框架，在这个框架中，提取出的智能被视为一个一流的、可独立管理的实体，可以在异构边缘设备和应用之间进行描述、发现、观察、共享、重用和动态聚类。为了解决这些研究空白，我们提出了集群边缘智能（Clustered Edge Intelligence，CEI），这是一种具有前瞻性的以智能为中心的方法。CEI的目标是使智能成为一个可共享和可重用的一流实体，可以在分布式边缘-云连续体中独立表示、发现、观察、交换和管理。我们提出了一个三层CEI架构，并考察了支持技术和研究维度，包括智能库存、语义知识表示、通信、可发现性、可观察性、生命周期自动化、聚类机制、市场、互操作性和标准化。

View on arXiv Download PDF AI Translation

cs.AI / 77 / 2607.20970

From Scalars to Time Series: Rethinking Implicit Neural Representations for Time-Varying Volumetric Data

从标量到时间序列：重新思考时间变化体积数据的隐式神经表示

Zhang, Weihan, Zhao, Xuan, Peng, Yenwen, Chen, Yuqi, Tao, Jun

Abstract

Implicit neural representations (INRs) for time-varying volumetric data are typically trained using dense sampling over spatiotemporal coordinates, where each observation corresponds to a single point in space and time. This coordinate-wise formulation requires extensive sampling during optimization, leading to high computational cost and inefficient use of temporal structure. In this work, we revisit this design choice and show that dense spatiotemporal sampling is not necessary for learning time-varying fields. Instead, we represent the data as a collection of spatially indexed time series and train INRs using sequence-level supervision over each spatial location, rather than coordinate-wise scalar samples. This reformulation eliminates the need for dense spatiotemporal sampling and instead learns each spatial location from its full temporal evolution in a structured manner. We demonstrate that this representation is compatible with a range of existing INR architectures and consistently improves reconstruction quality, while significantly reducing training cost. Furthermore, we show that this formulation can be combined with mixture-of-experts architectures, and that our MoE instantiation further improves reconstruction quality compared to both the base reformulation and existing MoE-based INR methods, providing a stronger capacity allocation under heterogeneous temporal dynamics.

Chinese Translation

时间变化体积数据的隐式神经表示（INRs）通常通过在时空坐标上进行密集采样进行训练，其中每个观测值对应于空间和时间中的一个单一点。这种坐标级的表述在优化过程中需要大量的采样，导致高计算成本和时间结构的低效利用。在本研究中，我们重新审视这一设计选择，并表明密集的时空采样并不是学习时间变化场所必需的。相反，我们将数据表示为一组空间索引的时间序列，并使用每个空间位置的序列级监督来训练INRs，而不是坐标级的标量样本。这种重新表述消除了对密集时空采样的需求，而是以结构化的方式从每个空间位置的完整时间演化中学习。我们证明这种表示与一系列现有的INR架构兼容，并且在显著降低训练成本的同时，持续提高重建质量。此外，我们展示这种表述可以与专家混合架构相结合，并且我们的MoE实例化相比于基础重新表述和现有的基于MoE的INR方法进一步提高了重建质量，在异构时间动态下提供了更强的能力分配。

View on arXiv Download PDF AI Translation

cs.AI / 78 / 2607.20972

Delivery, Not Storage: Cue-Anchored Working Memory as a Harness Property for Coding Agents

交付，而非存储：线索锚定工作记忆作为编码代理的支撑特性

Saha, Swapnanil

Abstract

Coding agents ship with one kind of memory: documents. Instruction files, plan artifacts, and auto-written memory directories are deliberately authored and deliberately retrieved: the agent must choose to write them and choose to read them back. Human expertise runs on a second tier that never gets written down: situationally-bound operational facts (gotchas, locations, local conventions) encoded as a side effect of the work and retrieved involuntarily when the situation cues them. We argue this second tier is the load-bearing one for long-running agents and must be a harness property, not an agent choice. We contribute: (1) a two-tier design theory grounded in the cognitive literature on memory offloading, incidental encoding, and event-based prospective memory, each mapped to an architectural requirement; (2) a cue-anchored memory model where memories carry first-class trigger conditions over a composable vocabulary (path, symbol, semantic, event, temporal), evaluated deterministically by the harness, a composition no surveyed academic or shipped system provides; (3) a controlled evaluation on a real coding task showing that voluntary memory use is near zero even with a pre-seeded store (0 memory operations in 114 turns), that deterministic injection delivered in every seeded run with zero false alarms, and that 39% of intra-session re-reads re-buy content paid for before a compaction boundary; (4) a repeated-compaction decay probe: ten facts held only in conversation vanish at the first summary and stay absent from 106 of 108 compactions, and the deprived agent greps the harness's own session files to rebuild them, while the same facts injected from a harness-owned store arrive intact through all 138 compact-resumes as the final summary carries none. Delivery, not storage, is the product: the reliable memory channel for agents is the one the agent never has to think about.

Chinese Translation

编码代理配备了一种记忆：文档。指令文件、计划文物和自动生成的记忆目录都是经过精心编写和有意检索的：代理必须选择写入它们，并选择将其读回。人类的专业知识依赖于第二层次，这一层次从未被记录下来：情境绑定的操作事实（陷阱、位置、地方惯例）作为工作副作用被编码，并在情境提示时不自觉地被检索。我们认为，这第二层次是长时间运行的代理所依赖的承重层，必须作为支撑特性，而非代理的选择。我们的贡献包括：（1）基于认知文献关于记忆卸载、偶然编码和事件基础前瞻性记忆的双层设计理论，每一层次都映射到一个架构要求；（2）一个线索锚定的记忆模型，其中记忆携带一流的触发条件，涵盖可组合的词汇（路径、符号、语义、事件、时间），由支撑层以确定性方式进行评估，这是目前没有任何调查的学术或已发布的系统提供的组合；（3）在一个真实编码任务上的受控评估，显示即使在预先填充的存储中，自愿记忆使用几乎为零（在114轮中没有记忆操作），而在每次填充运行中以零误报交付的确定性注入，以及39%的会话内重读重新获取在压缩边界之前支付的内容；（4）一个重复压缩衰减探测：仅在对话中持有的十个事实在第一次总结时消失，并在108次压缩中的106次中保持缺失，而缺失的代理则从支撑层自己的会话文件中grep出这些事实以重建它们，而从支撑层拥有的存储中注入的相同事实在所有138次压缩恢复中完好无损，因为最终总结中没有包含这些事实。交付，而非存储，才是产品：代理的可靠记忆通道是代理无需思考的通道。

View on arXiv Download PDF AI Translation

cs.AI / 79 / 2607.20981

Beyond Independent Optimization: Compression, MoE Routing, and Quantization Interactions in Multimodal Edge Intelligence

超越独立优化：多模态边缘智能中的压缩、混合专家路由与量化交互

Gor, Jay, Dave, Karm, Abrol, Akshita, Gupta, Rajesh, Tanwar, Sudeep, Wang, Zhengkui

Abstract

Efficient multimodal inference is increasingly constrained not only by model quality or FLOP count, but also by the cost of preserving, moving, routing, caching, and quantizing multimodal representations under latency, memory, and energy constraints. This paper reviews recent advances in efficient vision-language and multimodal large language models, covering visual token compression, video token management, KV-cache optimization, Mixture-of-Experts (MoE) routing, low-bit quantization, edge deployment, and hardware-aware benchmarking. We argue that these techniques cannot be treated as independent optimizations. Visual token compression alters downstream feature distributions and MoE routing decisions, routing behavior affects expert utilization and quantization sensitivity, quantized router logits influence expert assignment, KV-cache policies determine retained multimodal evidence, and hardware constraints often transform computational savings into memory and communication bottlenecks. We organize the literature around these interactions and identify key design trade-offs, including accuracy versus token budget, static versus adaptive compression, sparse routing efficiency versus expert collapse, and low-bit inference versus modality-specific degradation. Finally, we introduce Temporal Routing Consistency as a diagnostic for video MoE models and highlight open research directions in routing-aware compression, cross-modal cache management, hardware-aware co-design, and unified benchmarking for multimodal edge intelligence.

Chinese Translation

高效的多模态推理不仅受到模型质量或浮点运算次数（FLOP）的限制，还受到在延迟、内存和能量约束下，保持、移动、路由、缓存和量化多模态表示的成本的影响。本文回顾了在高效视觉-语言和多模态大语言模型方面的最新进展，涵盖了视觉标记压缩、视频标记管理、KV-cache优化、混合专家（Mixture-of-Experts, MoE）路由、低比特量化、边缘部署和硬件感知基准测试。我们认为，这些技术不能被视为独立的优化。视觉标记压缩改变了下游特征分布和MoE路由决策，路由行为影响专家利用率和量化敏感性，量化路由器的logits影响专家分配，KV-cache策略决定保留的多模态证据，而硬件约束通常将计算节省转化为内存和通信瓶颈。我们围绕这些交互组织文献，并识别关键设计权衡，包括准确性与标记预算、静态与自适应压缩、稀疏路由效率与专家崩溃、低比特推理与模态特定降级。最后，我们引入时间路由一致性作为视频MoE模型的诊断工具，并强调在路由感知压缩、跨模态缓存管理、硬件感知协同设计和多模态边缘智能的统一基准测试等领域的开放研究方向。

View on arXiv Download PDF AI Translation

cs.AI / 80 / 2607.20982

GuardianAgentBench: Where Agents Fail and How to Guard Them

GuardianAgentBench：代理人失败的地方及其防护措施

Naik, Vishal Ishwar, Xu, Chenyu, Dong, Donna, Hassan, Hussein, Pradhan, Abhishek, Mendelevitch, Ofer, Shafat, Tallat, Irshad, Humayun

Abstract

As large language model agents increasingly operate autonomously with access to tools and external environments, ensuring their safe and reliable behavior becomes critical. We present GuardianAgentBench (GABench), a benchmark of 580 scenarios across six domains evaluated on three production-ready frameworks: LangChain, LlamaIndex, and Vectara. The benchmark incorporates rigorous multi-stage validation and five adversarial attack modes. Experiments with six state-of-the-art models reveal that even the strongest configuration achieves only 74.8% overall accuracy and expose two distinct failure regimes: stronger models under-call required tools, while weaker models mis-select and over-call tools. Performance degrades monotonically with both tool-set size and sequential turn depth, with long-horizon planning proving the steeper bottleneck. Our guardrail implementation consistently outperforms system-prompt-based defenses across all models, recovering 19.9% of failures at a false positive rate of just 0.5%. These results demonstrate that execution-time structural intervention improves safety without disrupting correct agent behavior.

Chinese Translation

随着大型语言模型代理人越来越多地自主操作并访问工具和外部环境，确保它们的安全和可靠行为变得至关重要。我们提出了GuardianAgentBench（GABench），这是一个涵盖六个领域的580个场景的基准测试，评估了三个生产就绪框架：LangChain、LlamaIndex和Vectara。该基准测试结合了严格的多阶段验证和五种对抗攻击模式。对六个最先进模型的实验表明，即使是最强的配置也仅实现了74.8%的整体准确率，并暴露出两种不同的失败模式：较强的模型未能调用所需工具，而较弱的模型则错误选择并过度调用工具。随着工具集规模和顺序回合深度的增加，性能单调下降，长期规划被证明是更陡峭的瓶颈。我们的防护实施在所有模型中始终优于基于系统提示的防御，以仅0.5%的假阳性率恢复了19.9%的失败。这些结果表明，执行时的结构干预可以提高安全性，而不会干扰代理人的正确行为。

View on arXiv Download PDF AI Translation

cs.AI / 81 / 2607.20999

Workflow-Localized Mechanism Learning: Attribution-Guided Repair and Knowledge Reuse for Structured Agent Skills

工作流本地化机制学习：归因引导的修复与结构化代理技能的知识重用

Lin, Zibin, Zhang, Shengli, Wang, Taotao, Xia, Yihan, Ma, Deen, Liao, Guofu

Abstract

Agent Skills package reusable procedural knowledge as external artifacts for frozen language-model agents, yet existing optimizers do not jointly resolve where a failure occurs in a workflow, which mechanism caused it, and how relevant knowledge from third-party Skills should be reused locally. We introduce Workflow-Localized Mechanism Learning (WML). Its Node--Mechanism Attribution identifies the failed workflow node, implicated mechanisms, and smallest valid edit target, routing single-mechanism defects to L3 resources and relational defects across mechanisms to L2 composition protocols. A six-module Workflow-Guided Skill Optimization (WGSO) loop then selects provenance- and scope-aware third-party knowledge, applies bounded patches, evaluates candidates, and stores verified outcomes in optimizer-side memory. On SpreadsheetBench, WML reaches 90.33 +/- 1.53 and 74.67 +/- 3.51 Hard Accuracy with DeepSeek and Qwen3.6-Flash, respectively; without additional optimization, the learned Skills transfer to WikiTableQuestions with 84.00 +/- 2.00 and 83.00 +/- 2.00 Denotation Accuracy. On Compiler-Supported50, WML attains both the highest hard-PASS rate and the lowest cost per successful task; compiled execution sharply reduces tokens and calls relative to a direct SkillAgent while retaining most of its successful tasks. Code and artifacts are available at https://github.com/xiaolin9595/workflow-localized-mechanism-learning.

Chinese Translation

代理技能将可重用的程序知识打包为冻结语言模型代理的外部工件，但现有优化器未能共同解决工作流中故障发生的位置、导致故障的机制以及如何在本地重用来自第三方技能的相关知识。我们提出了工作流本地化机制学习（Workflow-Localized Mechanism Learning, WML）。其节点-机制归因（Node--Mechanism Attribution）识别失败的工作流节点、相关机制和最小有效编辑目标，将单一机制缺陷路由到 L3 资源，并将跨机制的关系缺陷路由到 L2 组合协议。随后，一个六模块的工作流引导技能优化（Workflow-Guided Skill Optimization, WGSO）循环选择具有来源和范围意识的第三方知识，应用有限补丁，评估候选项，并将验证结果存储在优化器侧内存中。在 SpreadsheetBench 上，WML 在使用 DeepSeek 和 Qwen3.6-Flash 时分别达到了 90.33 +/- 1.53 和 74.67 +/- 3.51 的硬准确率；在没有额外优化的情况下，学习到的技能在 WikiTableQuestions 上转移时达到了 84.00 +/- 2.00 和 83.00 +/- 2.00 的指称准确率。在 Compiler-Supported50 上，WML 达到了最高的硬通过率和最低的成功任务成本；编译执行相较于直接的 SkillAgent 显著减少了令牌和调用，同时保留了大部分成功任务。代码和工件可在 https://github.com/xiaolin9595/workflow-localized-mechanism-learning 获取。

View on arXiv Download PDF AI Translation

cs.AI / 82 / 2607.21000

Naju: A Native Discrete State-Space Model with Independent Retention and Writing for Long-Sequence Memory

Naju：一种具有独立保留和写入能力的原生离散状态空间模型，用于长序列记忆

Lim, Hyuk, Yoon, Seunghyun

Abstract

Long-sequence memory tracking places two opposing demands on a recurrent state: near-lossless retention of stored bindings over long horizons, and active overwriting of stale ones. In our diagnostic suite, the strongest efficient baselines tend to solve only one side well. Continuous-time-parameterized state-space models (SSMs) such as Mamba obtain their discrete recurrence by zero-order-hold discretization of a continuous-time system; we argue that this detour is unnecessary for memory tracking and parameterize the discrete transition directly. Naju (Native Adaptive Junction Unit) factorizes the recurrent update, schematically $x_n = f_n\odot x_{n-1} + i_n\odot(B_n u_n)$, into an explicit discrete pole (a learned forget gate $f_n$), an independent write gain $i_n$, and input-dependent write/read maps. Since the sigmoid pole satisfies $0

Chinese Translation

长序列记忆跟踪对递归状态提出了两种相互对立的要求：在长时间范围内对存储绑定的近乎无损保留，以及对过时绑定的主动覆盖。在我们的诊断套件中，最强的高效基线往往只能很好地解决其中一方面。诸如 Mamba 的连续时间参数化状态空间模型（SSMs）通过对连续时间系统进行零阶保持离散化来获得其离散递归；我们认为这种绕道在记忆跟踪中是不必要的，并直接对离散转移进行参数化。Naju（原生自适应连接单元）将递归更新分解为一个显式离散极点（一个学习的遗忘门 $f_n$）、一个独立的写入增益 $i_n$ 和依赖于输入的写入/读取映射。由于 sigmoid 极点满足 $0

View on arXiv Download PDF AI Translation

cs.AI / 83 / 2607.21010

Reexamining zero-shot summarization: Empirical investigation of trustworthiness of LLM-summarizers

重新审视零样本摘要：对大型语言模型摘要生成器可信度的实证研究

Bhatnagar, Vasudha, Bindal, Purnima, Kumar, Vikas, Bahl, Raj Kumari

Abstract

Zero-shot summarization using Large Language Models (LLMs) has significantly advanced the abstractive summarization task by producing coherent and fluent summaries. However, underlying stochasticity of the large language models raises concerns about the stability and trustworthiness of the LLM-generated summaries. This issue has become increasingly important due to proliferation of LLM-generated summaries in educational settings, where students and researchers summarize complex academic materials in zero-shot manner. We propose a novel two-level diagnostic protocol for benchmarking LLM-summarizers based on the stability of the generated summaries. At the lower level, document-level stability analysis is performed over multiple LLM-summaries generated under controlled environment, and the stability coefficient is computed. Each generated summary is scored for semantic and factual alignment with the original document, enabling estimation of stability along more than one dimensions. At the next level, observations from a stratified sample of documents drawn from the corpus are consolidated to estimate the stability index of the LLM-summarizer, which is the proxy for its trustworthiness. Our empirical investigation of three LLM-summarizers across three genres of documents reveals statistically significant differences in the generation-level variability among LLMs across summary evaluation metrics. This study advances the LLM-summarization research by evidential recognition of the stability problem in LLM-summaries and motivates further research towards development of robust, reliable and trustworthy LLM-summarizers.

Chinese Translation

使用大型语言模型（LLMs）的零样本摘要显著推动了抽象摘要任务的发展，能够生成连贯且流畅的摘要。然而，大型语言模型的随机性引发了对LLM生成摘要的稳定性和可信度的担忧。由于LLM生成摘要在教育环境中的广泛应用，这一问题变得愈发重要，学生和研究人员以零样本方式总结复杂的学术材料。我们提出了一种新颖的两级诊断协议，用于基于生成摘要的稳定性对LLM摘要生成器进行基准测试。在较低层面上，对在受控环境下生成的多个LLM摘要进行文档级稳定性分析，并计算稳定性系数。每个生成的摘要都根据与原始文档的语义和事实一致性进行评分，从而使得在多个维度上估计稳定性成为可能。在下一个层面上，从语料库中抽取的分层样本文档的观察结果被整合，以估计LLM摘要生成器的稳定性指数，这一指数作为其可信度的代理。我们对三种类型文档的三种LLM摘要生成器的实证研究揭示了在摘要评估指标中，LLM生成的摘要在生成层面的变异性存在统计显著差异。本研究通过对LLM摘要中稳定性问题的证据性识别，推动了LLM摘要研究的发展，并激励进一步研究以开发稳健、可靠和可信的LLM摘要生成器。

View on arXiv Download PDF AI Translation

cs.AI / 84 / 2607.21013

EmoAgent-R1: Towards Multimodal Emotion Understanding with Reinforcement Learning-based Dynamic Agent Specialization

EmoAgent-R1：基于强化学习的动态代理专业化的多模态情感理解

Fang, Lihuang, Zou, Yuchen, Jin, kebin, Qin, Jinghui

Abstract

Multimodal large language models (MLLMs) have achieved impressive performance in multimodal emotion recognition (MER) tasks and lifted MER to a new level that is complex emotion understanding with advanced video understanding abilities and natural language description. However, existing MLLM-based methods often use a fixed prompt to perceive the emotions, ignoring the dynamicity and complexity of the emotion source in the multimodal inputs. To address these issues, we propose a novel Reinforcement Learning-based Dynamic Agent Specialization framework (\textbf{EmoAgent-R1}) to optimize the emotion recognition, reasoning, and generalization abilities of an MLLM with dynamic agent specialization based on reinforcement learning. Specifically, we first adopt a cold start strategy to endow an MLLM with preliminary emotion recognition, reasoning, and agent routing ability by training with synthetic answer-conditioned chain-of-thought data and agent routing data. Then, we further train the MLLM with reinforcement learning to perceive emotions in a two-step agentic workflow with agent selection and agent specialization. To effectively train EmoAgent-R1, we propose a novel Progressive Group-Relative Policy Optimization (P-GRPO) to combine group-based relative advantages with a PMI-inspired progressive token-level modulation to transform sparse rewards into fine-grained learning signals, mitigating the coarse-grained uniform credit assignment issue in GRPO. Extensive experiments on MER benchmarks demonstrate the superiority of our EmoAgent-R1 in stronger emotion reasoning performance and improved optimization stability.

Chinese Translation

多模态大型语言模型（MLLMs）在多模态情感识别（MER）任务中取得了令人瞩目的表现，并将MER提升到一个新的层次，即具备先进视频理解能力和自然语言描述的复杂情感理解。然而，现有的基于MLLM的方法通常使用固定的提示来感知情感，忽视了多模态输入中情感源的动态性和复杂性。为了解决这些问题，我们提出了一种新颖的基于强化学习的动态代理专业化框架（EmoAgent-R1），旨在通过基于强化学习的动态代理专业化来优化MLLM的情感识别、推理和泛化能力。具体而言，我们首先采用冷启动策略，通过使用合成的答案条件链式思维数据和代理路由数据来赋予MLLM初步的情感识别、推理和代理路由能力。然后，我们进一步通过强化学习训练MLLM，以在包含代理选择和代理专业化的两步代理工作流程中感知情感。为了有效训练EmoAgent-R1，我们提出了一种新颖的渐进式群体相对策略优化（P-GRPO），将基于群体的相对优势与受PMI启发的渐进式令牌级调制相结合，以将稀疏奖励转化为细粒度学习信号，从而缓解GRPO中粗粒度均匀信用分配的问题。在MER基准上的广泛实验表明，我们的EmoAgent-R1在更强的情感推理性能和优化稳定性方面具有优越性。

View on arXiv Download PDF AI Translation

cs.AI / 85 / 2607.21019

HiMe: Real-Time Self-Hosted Personal Agent Platform for Health Insights with Wearable Devices

HiMe：基于可穿戴设备的实时自托管个人健康洞察代理平台

Liu, Wei, Qi, Siya, Zhang, Linhai, Car, Lorainne Tudor, He, Yulan

Abstract

Traditional approaches to wearable health signal analysis, such as smartwatches, are constrained by rigid analytical frameworks and limited personalisation. The emergence of LLM agents creates a new opportunity for Personal Health Agentic Analysis, where health insights can be generated adaptively and in context. However, currently there is no open-source locally deployable platform capable of processing personal health data in real time while preserving privacy. We present HiMe, a locally deployable, privacy-first agent platform that is fully compatible with real-time health data ecosystems across a wide range of wearable devices. HiMe is guided by three design principles. The database is treated as a first-class component. Effectiveness and efficiency are jointly optimised to achieve a low-cost Pareto-optimal balance. Data are processed in real time while the user is modelled over the long term. Together, these principles make it practical for individuals to harness Personal Health Agents for continuous, personalised health monitoring for better wellbeing.

Chinese Translation

传统的可穿戴健康信号分析方法，如智能手表，受限于僵化的分析框架和有限的个性化。大型语言模型（LLM）代理的出现为个人健康代理分析创造了新的机会，使健康洞察能够自适应地生成并具有上下文相关性。然而，目前尚无能够实时处理个人健康数据并保护隐私的开源本地可部署平台。我们提出了HiMe，一个本地可部署、以隐私为首的代理平台，完全兼容各种可穿戴设备的实时健康数据生态系统。HiMe遵循三个设计原则。数据库被视为一个一流组件。有效性和效率共同优化，以实现低成本的帕累托最优平衡。数据在实时处理中，用户的模型则在长期内进行建模。这些原则使个人能够有效利用个人健康代理进行持续的个性化健康监测，以促进更好的健康福祉。

View on arXiv Download PDF AI Translation

cs.AI / 86 / 2607.21042

Faster IndexTTS-2: Accelerating and Streaming Autoregressive Zero-Shot Text-to-Speech Synthesis on GPUs

更快的 IndexTTS-2：在 GPU 上加速和流式自回归零-shot 文本到语音合成

Du, Muyang, Yu, Shuang, Lai, Junjie

Abstract

Autoregressive text-to-speech models achieve strong naturalness but suffer from slow inference due to sequential token generation, limiting their deployment in production applications that require low latency. IndexTTS-2 is a state-of-the-art autoregressive TTS model consisting of a GPT, a flow-matching Diffusion Transformer, and a vocoder. Despite its high synthesis quality, its inference speed barely reaches real-time without streaming or batching support. We present Faster IndexTTS-2, which accelerates all neural network components of IndexTTS-2 for production deployment on GPUs using NVIDIA TensorRT and TensorRT-LLM. Faster IndexTTS-2 also enables streaming synthesis for latency-sensitive interactive applications, and batched inference across all components to maximize GPU utilization. Experiments on the Seed-TTS benchmark for both English and Chinese demonstrate up to 5.0$\times$ speedup on the autoregressive GPT and 3.6$\times$ end-to-end, with minimal degradation in word error rate, speaker similarity, and naturalness. Our methodology provides a practical reference for efficiently accelerating similar autoregressive speech models on GPUs.

Chinese Translation

自回归文本到语音模型在自然性方面表现出色，但由于顺序生成令牌的原因，其推理速度较慢，限制了其在需要低延迟的生产应用中的部署。IndexTTS-2 是一种最先进的自回归 TTS 模型，由 GPT、流匹配扩散变换器（Diffusion Transformer）和声码器组成。尽管其合成质量很高，但在没有流式或批处理支持的情况下，其推理速度几乎无法达到实时水平。我们提出了更快的 IndexTTS-2，它利用 NVIDIA TensorRT 和 TensorRT-LLM 加速 IndexTTS-2 的所有神经网络组件，以便在 GPU 上进行生产部署。更快的 IndexTTS-2 还支持流式合成，适用于对延迟敏感的交互式应用，并在所有组件中实现批量推理，以最大化 GPU 的利用率。在 Seed-TTS 基准测试中对英语和中文的实验表明，自回归 GPT 的速度提升高达 5.0 倍，端到端速度提升为 3.6 倍，且在词错误率、说话人相似性和自然性方面的降级极小。我们的方法为在 GPU 上高效加速类似的自回归语音模型提供了一个实用的参考。

View on arXiv Download PDF AI Translation

cs.AI / 87 / 2607.21101

Can Generative Recommendation Reach Cold Items? A Temporal Perspective on Semantic-ID Generation

生成推荐能否覆盖冷启动项目？基于时间的语义ID生成视角

Peng, Jie, Zheng, Yanping, Zhe, Zhewei, Tong, Bin, Wang, Guan, Zheng, Bo

Abstract

Semantic-ID-based generative recommendation represents items as sequences of shared semantic tokens, enabling token recombination beyond isolated item IDs. However, closed-world recombination does not necessarily imply temporal open-token cold-start induction, where new items enter the item catalog with unseen atomic tokens or weakly supported SID paths. In this work, we revisit SID-based generative recommendation under an absolute-time temporal protocol that separates seen and unseen targets and diagnoses the cold item reachability at the token level. Through seen/unseen-hit analysis, coldness taxonomy, and oracle-prefix probing, we show that current SID-based models can occasionally reach future items supported by observed tokens and prefixes, but struggle with unseen atomic tokens and unsupported SID paths. We further explain this boundary by interpreting SID generation as hierarchical semantic bucketing: early tokens select coarse semantic regions, while later tokens refine item-specific paths. These findings show that SID generation is compositional but not fully open-ended, and suggest future directions in more independent SID spaces, scoring-based interfaces, and dynamic textual context.

Chinese Translation

基于语义ID的生成推荐将项目表示为共享语义标记的序列，使得标记的重组超越了孤立的项目ID。然而，封闭世界的重组并不一定意味着时间上的开放标记冷启动引导，其中新项目以未见的原子标记或弱支持的SID路径进入项目目录。在本研究中，我们在一个绝对时间的时间协议下重新审视基于SID的生成推荐，该协议区分了已见和未见的目标，并在标记层面上诊断冷启动项目的可达性。通过已见/未见命中分析、冷启动分类法和oracle前缀探测，我们表明当前基于SID的模型偶尔能够覆盖由观察到的标记和前缀支持的未来项目，但在处理未见的原子标记和不支持的SID路径时存在困难。我们进一步通过将SID生成解释为层次语义分桶来阐明这一边界：早期标记选择粗略的语义区域，而后期标记则细化特定项目的路径。这些发现表明SID生成是组合性的，但并非完全开放的，并建议在更独立的SID空间、基于评分的接口和动态文本上下文方面的未来研究方向。

View on arXiv Download PDF AI Translation

cs.AI / 88 / 2607.21106

AttriMem: Attribution-Guided Process Feedback for Agent Memory Learning

AttriMem：基于归因的过程反馈用于智能体记忆学习

Li, Qinfeng, Bao, Yuntai, Yu, Xinyan, Chen, Hongze, Zhang, Wenqi, Zhang, Xuhong

Abstract

Effective memory is crucial for LLM agents, yet constructing it effectively remains challenging. A memory-construction policy decides what information to extract, store, update, compress, or discard as interactions accumulate. Heuristic memory methods rely on subjective, task-specific rules, which can misalign with downstream objectives and limit cross-task adaptability. RL-based methods, by contrast, learn from task feedback but mainly use outcome- or module-level rewards. These coarse signals indicate task success but cannot identify which intermediate memory contents support the final answer, creating a fine-grained credit-assignment bottleneck. However, constructing such process feedback is prohibitively difficult because intermediate memory decisions lack unique ground-truth targets, while the appropriate credit varies with the agent's uncertain reasoning trajectory and therefore cannot be specified in advance. We propose AttriMem, an attribution-guided process-feedback framework for learning memory-construction policies with RL. AttriMem augments the global outcome reward with local rewards derived from token-level contributions to the final answer. Experiments on long-horizon dialogue question answering show that AttriMem outperforms retrieval-based, heuristic, and RL-based baselines, generalizes across benchmarks and answer models, stabilizes RL optimization.

Chinese Translation

有效的记忆对于大型语言模型（LLM）智能体至关重要，但有效构建记忆仍然具有挑战性。记忆构建策略决定了在交互过程中提取、存储、更新、压缩或丢弃哪些信息。启发式记忆方法依赖于主观的、特定任务的规则，这可能与下游目标不一致，并限制跨任务的适应性。相比之下，基于强化学习（RL）的方法通过任务反馈进行学习，但主要使用结果或模块级的奖励。这些粗略的信号指示任务成功，但无法识别哪些中间记忆内容支持最终答案，从而造成细粒度信用分配的瓶颈。然而，构建这样的过程反馈是极其困难的，因为中间记忆决策缺乏唯一的真实目标，而适当的信用因智能体的不确定推理轨迹而异，因此无法提前指定。我们提出了AttriMem，一个基于归因的过程反馈框架，用于通过RL学习记忆构建策略。AttriMem通过从令牌级别对最终答案的贡献中派生的局部奖励来增强全局结果奖励。在长时间对话问答的实验中，AttriMem的表现优于基于检索、启发式和基于RL的基线，能够在基准和答案模型之间进行泛化，并稳定RL优化。

View on arXiv Download PDF AI Translation

cs.AI / 89 / 2607.21151

V-DEAL: Diagnosing Video Safety De-Calibration as an Understanding-Refusal Coupling Failure

V-DEAL：将视频安全去标定诊断为理解拒绝耦合失败

Zhang, Zhetong, Fu, Honghao, Xu, Miao, Wang, Yiwei, Cai, Yujun

Abstract

As Video Large Language Models are increasingly deployed in real-world applications, ensuring their safety alignment has become critical. Counterintuitively, we find that harmful videos paired with benign queries achieve higher attack success rates than the same videos paired with explicitly harmful queries. To understand the underlying mechanism of this vulnerability, we present V-DEAL, a three-level diagnostic framework that jointly analyzes this failure across model behaviour, understanding, and internal representations. By progressively ruling out perception failure and quantifying the model's internal refusal tendency, V-DEAL provides a new diagnostic perspective for analyzing the underlying mechanism of the observed vulnerability. We tested six Video LLMs on three public benchmarks and observed that models correctly recognize harmful video content with over 81\% accuracy, yet the average attack success rate still reaches 48.33\% under the condition pairing harmful videos with benign queries. Hidden-state analysis further shows that visual understanding activates a weaker refusal tendency than textual understanding. Furthermore, we introduce a prompt injection intervention method that reduces attack success rates by an average of 48.24 percentage points and achieves performance comparable to prior fine-tuning-based methods, providing an effective and practical means to address such safety risks in Video LLMs.

Chinese Translation

随着视频大型语言模型在现实世界应用中的逐渐部署，确保其安全性对齐变得至关重要。出乎意料的是，我们发现与良性查询配对的有害视频的攻击成功率高于与明确有害查询配对的同一视频。为了理解这种脆弱性的潜在机制，我们提出了V-DEAL，这是一种三层诊断框架，联合分析模型行为、理解和内部表征中的这种失败。通过逐步排除感知失败并量化模型的内部拒绝倾向，V-DEAL为分析观察到的脆弱性的潜在机制提供了新的诊断视角。我们在三个公共基准上测试了六个视频大型语言模型，观察到模型以超过81%的准确率正确识别有害视频内容，但在将有害视频与良性查询配对的情况下，平均攻击成功率仍达到48.33%。隐藏状态分析进一步表明，视觉理解激活的拒绝倾向弱于文本理解。此外，我们引入了一种提示注入干预方法，平均降低攻击成功率48.24个百分点，并实现了与先前基于微调的方法相当的性能，为解决视频大型语言模型中的安全风险提供了一种有效且实用的手段。

View on arXiv Download PDF AI Translation

cs.AI / 90 / 2607.21156

SafeStep: AI-powered Travel Assistance for Elderly People with Frailty or Dementia

SafeStep：为虚弱或痴呆老年人提供的人工智能旅行辅助

Debenedetti, Elderly People with Frailty or Dementia Azul, Gamez, David, Such, Franco, Kairinos, Nik

Abstract

More than a million people in the UK suffer from frailty or dementia, which severely compromise their ability to travel in urban environments. This paper presents SafeStep, an AI-driven travel system that assists elderly users with their journeys. At the core of SafeStep is a novel travel graph representation, which integrates route planning with predictive modelling. For each stage of a journey, the system (i) generates personalized failure scenarios using a combi-nation of LLMs and the Anticip8 behavioral prediction engine, (ii) proposes targeted interventions, and (iii) estimates the impact of interventions on out-come probabilities. This enables SafeStep to select interventions that maximize the likelihood of the person reaching their destination. SafeStep was evaluated through experiments on travel graph generation and a field study involving 26 real-world journeys. Results showed that combining Anticip8 for failure pre-diction with GPT-based models for intervention evaluation yields the most re-liable performance. User feedback indicated that SafeStep improves confidence and perceived safety during travel, although interface usability needs to be im-proved for the target demographic. In the future, we would like to improve and release SafeStep. The AI system that was developed for SafeStep could be ap-plied in other areas, such as mental health, career coaching and addiction treatment.

Chinese Translation

在英国，超过一百万人受到虚弱或痴呆的困扰，这严重影响了他们在城市环境中的旅行能力。本文介绍了SafeStep，一个基于人工智能的旅行系统，旨在帮助老年用户顺利完成旅行。SafeStep的核心是一个新颖的旅行图表示，它将路线规划与预测建模相结合。在旅行的每个阶段，系统 (i) 利用大型语言模型（LLMs）和Anticip8行为预测引擎生成个性化的失败场景，(ii) 提出针对性的干预措施，以及 (iii) 评估干预措施对结果概率的影响。这使得SafeStep能够选择最大化用户到达目的地可能性的干预措施。SafeStep通过旅行图生成实验和涉及26个真实旅行的实地研究进行了评估。结果表明，将Anticip8用于失败预测与基于GPT的模型用于干预评估相结合，能够实现最可靠的性能。用户反馈表明，SafeStep提高了旅行过程中的信心和安全感，尽管界面可用性需要针对目标人群进行改进。未来，我们希望改进并发布SafeStep。为SafeStep开发的人工智能系统还可以应用于其他领域，如心理健康、职业指导和成瘾治疗。

View on arXiv Download PDF AI Translation

cs.AI / 91 / 2607.21180

Safeguards for Speech2Speech LLM-Assistants: A Case Study in Automotive Applications

语音到语音大型语言模型助手的安全保障：汽车应用案例研究

Endler, Gregor, Kraus, Sebastian, Stappen, Lukas

Abstract

Recent advances have introduced speech-to-speech (S2S) conversational assistants capable of producing natural-sounding interactions, including non-verbal cues like tonality and mood. In the automotive domain, this enables intuitive and humanlike in-car dialogue experiences. However, integrating these end-to-end assistants limits architectural options for programmable domain-specific safeguards. This paper discusses two implementation approaches for S2S guardrails: transcript-based and tool-based. Through an empirical evaluation, we demonstrate that both strategies are insufficient for industrial deployment in most cases due to prohibitive latency (delaying each answer by 0 to 1.4 seconds even for computationally cheap checks) and technical impediments (like potentially non-deterministic tool call behavior). Finally, we outline open challenges for S2S guardrails in the automotive context.

Chinese Translation

近期的进展引入了能够产生自然对话的语音到语音（S2S）对话助手，包括音调和情绪等非语言提示。在汽车领域，这使得车内对话体验更加直观和人性化。然而，集成这些端到端助手限制了可编程领域特定安全保障的架构选项。本文讨论了两种S2S安全保障的实施方法：基于转录的和基于工具的。通过实证评估，我们证明这两种策略在大多数情况下都不足以支持工业部署，因为它们存在显著的延迟（即使对于计算成本较低的检查，每个回答的延迟也在0到1.4秒之间）和技术障碍（如潜在的非确定性工具调用行为）。最后，我们概述了在汽车环境中S2S安全保障面临的开放挑战。

View on arXiv Download PDF AI Translation

cs.AI / 92 / 2607.21184

Explaining Weather Bulletins via ILP

通过归纳逻辑编程解释天气公告

Santi, Enrico, Palù, Alessandro Dal, Dovier, Agostino, Dreossi, Talissa, Formisano, Andrea

Abstract

Inductive Logic Programming (ILP) originated within the Logic Programming community in the Nineties as a framework for combining symbolic learning with declarative knowledge representation. Nowadays, mature ILP frameworks exist and they are capable of learning complex, non-monotonic hypotheses, thus broadening both the modeling capabilities and the scope of real-world applications of ILP. This work is primarily based on the FastLAS2 framework and aims to generate simple, interpretable hypotheses to help clarify the weather bulletins issued by OSMER FVG, the Regional Meteorological Observatory of the Italian region of Friuli Venezia-Giulia. In this paper we present a pipeline that, starting from simulated meteorological raw data and from OSMERs' bulletins (used as ground truth), extracts data as ASP facts and generates ILP examples. From such examples an explanatory hypothesis is then inferred via FastLAS2. Such a hypothesis (translated into natural language) explains the weather forecast issued by human experts, and in particular the rationale behind experts' choices of specific symbols in the bulletin pictogram (the symbol-annotated meteorological map of the forecast). The proposed approach is general, not specific to any particular region and it can equally be applied to bulletins from other sources and to different regions.

Chinese Translation

归纳逻辑编程（Inductive Logic Programming, ILP）起源于九十年代的逻辑编程社区，作为一种将符号学习与声明性知识表示相结合的框架。如今，成熟的ILP框架已经存在，能够学习复杂的非单调假设，从而拓宽了ILP的建模能力和现实世界应用的范围。本研究主要基于FastLAS2框架，旨在生成简单、可解释的假设，以帮助澄清由意大利弗留利-威尼斯-朱利亚地区的区域气象观测站OSMER FVG发布的天气公告。本文提出了一种管道，首先从模拟的气象原始数据和OSMER的公告（作为真实数据）中提取数据，作为ASP事实，并生成ILP示例。然后，通过FastLAS2推导出这样的示例的解释性假设。该假设（翻译成自然语言）解释了人类专家发布的天气预报，特别是专家在公告图示中选择特定符号的理由（即天气预报的符号标注气象图）。所提出的方法是通用的，不特定于任何特定地区，并且同样适用于来自其他来源的公告和不同地区的公告。

View on arXiv Download PDF AI Translation

cs.AI / 93 / 2607.21185

Differentiable Logic Programming to Mitigate Reasoning Shortcuts in Neurosymbolic Systems

可微分逻辑编程以减轻神经符号系统中的推理捷径

Takemura, Akihiro, Inoue, Katsumi

Abstract

Neurosymbolic (NeSy) systems integrate neural networks with logical reasoning to achieve both generalization and interpretability, but recent work has shown they are susceptible to shortcut reasoning behaviors. We propose a novel method using matrix-based differentiable logic programming to mitigate reasoning shortcuts in two phenomena: constraint satisfaction shortcuts, where constraints are satisfied without achieving the intended task, and cognition shortcuts, where biased data leads to semantically incorrect concept mappings despite logically sound inference. Building on recent matrix-based logic programming semantics, we introduce design elements to mitigate shortcuts, including a unified encoding of rules and constraints in a single matrix. We also identify connections to fuzzy logic t-norms and empirically compare their gradient flow properties. Through carefully designed experiments on MNIST variants, we show that one-to-one grounding of neural outputs to logical atoms significantly reduces both shortcut types compared to previous methods that rely on soft probability distributions. We then confirm that architectural choices in coupling symbolic knowledge with neural learning play a critical role in shortcut mitigation.

Chinese Translation

神经符号（NeSy）系统将神经网络与逻辑推理结合，以实现泛化和可解释性，但近期研究表明它们容易受到捷径推理行为的影响。我们提出了一种新方法，利用基于矩阵的可微分逻辑编程来减轻两种现象中的推理捷径：约束满足捷径，即在未实现预期任务的情况下满足约束，以及认知捷径，即偏见数据导致语义上不正确的概念映射，尽管推理在逻辑上是合理的。在近期基于矩阵的逻辑编程语义的基础上，我们引入设计元素以减轻捷径，包括在单个矩阵中对规则和约束的统一编码。我们还识别出与模糊逻辑t-范数的联系，并实证比较它们的梯度流特性。通过在MNIST变体上的精心设计实验，我们展示了将神经输出一对一地映射到逻辑原子的方式显著减少了两种捷径类型，相较于依赖软概率分布的先前方法。我们进一步确认，将符号知识与神经学习结合的架构选择在减轻捷径方面起着关键作用。

View on arXiv Download PDF AI Translation

cs.AI / 94 / 2607.21188

Identifying Good Rules for Efficient SAT Encodings of Single-Constant Multiplication Using Machine Learning

利用机器学习识别高效SAT编码单常数乘法的良好规则

Jiang, Chufeng, Zhou, Neng-Fa

Abstract

The Single Constant Multiplication problem is a fundamental NP-hard optimization task in hardware design, which seeks to decompose a fixed constant using only additions, subtractions, and bit-shifts. Although dynamic programming methods can produce near-optimal SAT encodings for SCM, their encoding cost remains high for large constants. We propose a neuro-symbolic framework that accelerates SCM SAT encoding by identifying good rules for guiding operator selection during decomposition. Our approach employs a graph neural network model to predict promising operator types from constant decompositions, and exploits the resulting confidence scores to prune no-good choices in the symbolic search. Experimental results on unseen 17-32 bit constants demonstrate one to two orders of magnitude reductions in encoding time, over 97% reduction in memory usage, and an order-of-magnitude decrease in branching, while preserving near-optimal encoding quality in terms of additions. These results show that learning-guided symbolic strategies can significantly improve the scalability and efficiency of SCM encoding. Our code and data are publicly available at: https://github.com/Chufeng-Jiang/SCM_MLDP

Chinese Translation

单常数乘法问题是硬件设计中的一个基本NP难优化任务，旨在仅通过加法、减法和位移操作来分解一个固定常数。尽管动态规划方法可以为单常数乘法（SCM）生成近似最优的SAT编码，但对于大常数，其编码成本仍然较高。我们提出了一种神经符号框架，通过识别良好的规则来加速SCM的SAT编码，以指导分解过程中的操作符选择。我们的方法采用图神经网络模型，从常数分解中预测有前景的操作符类型，并利用生成的置信度分数来修剪符号搜索中的不良选择。在对未见过的17-32位常数进行的实验结果表明，编码时间减少了一个到两个数量级，内存使用减少了超过97%，分支数量减少了一个数量级，同时在加法方面保持了近似最优的编码质量。这些结果表明，学习引导的符号策略可以显著提高SCM编码的可扩展性和效率。我们的代码和数据可在以下网址公开获取：https://github.com/Chufeng-Jiang/SCM_MLDP

View on arXiv Download PDF AI Translation

cs.AI / 95 / 2607.21201

Bound-Founded Semantics for Answer Set Programming with Difference Constraints: Preliminary Report

基于界限的语义在带差异约束的答案集编程中的应用：初步报告

Cabalar, Pedro, Fandinno, Jorge, Rühling, Nicolas, Schaub, Torsten, Schellhorn, Sebastian, Wanko, Philipp

Abstract

While the integration of linear constraints has significantly expanded the reach of Answer Set Programming (ASP), existing hybrid solvers often rely on disparate semantic underpinnings that lack a unified logical foundation. We address this gap by introducing a many-sorted variant of the Bound-founded Logic of Here-and-There (HTb), providing a versatile framework capable of characterizing equilibrium models across a wide spectrum of alternative semantics for extensions of ASP with linear constraints. We apply this framework to the setting of difference constraints, focusing on the semantic characterization of clingo[DL]. Central to our approach is the formalization of foundedness for numeric variables. By investigating how different hybrid systems - such as clingo[DL], clingcon, and flingo - justify constraint atoms, we uncover the semantic roots of their varying behaviors. This investigation results in a single, consistent framework that not only formalizes the foundations of current systems like clingo[DL] but also facilitates the rigorous study of program simplifications and the future integration of diverse semantic principles.

Chinese Translation

虽然线性约束的整合显著扩展了答案集编程（ASP）的应用范围，但现有的混合求解器往往依赖于缺乏统一逻辑基础的不同语义支撑。我们通过引入一种多排序的界限基础逻辑变体——这里与那里逻辑（HTb），来填补这一空白，提供一个多功能框架，能够表征在带线性约束的ASP扩展中不同语义的平衡模型。我们将该框架应用于差异约束的情境，重点关注 clingo[DL] 的语义特征。我们的方法的核心是对数值变量的基础性进行形式化。通过研究不同混合系统（如 clingo[DL]、clingcon 和 flingo）如何证明约束原子，我们揭示了它们不同表现的语义根源。这项研究产生了一个统一的一致框架，不仅形式化了当前系统（如 clingo[DL]）的基础，而且促进了对程序简化的严格研究以及未来多种语义原则的整合。

View on arXiv Download PDF AI Translation

cs.AI / 96 / 2607.21203

A New Well-Supported Semantics for Description Logic Programs

一种新的良好支持的描述逻辑程序语义

Killen, Spencer, You, Jia-Huai

Abstract

Description logic programs are a powerful formalism for combining rules with ontologies. The well-supported semantics for description logic programs ensures that no answer sets rely on cyclic dependencies. Most popular semantics for logic programming have this property of well-supportedness. We recognize two limitations of the current well-supported semantics for DL programs: its increased computational complexity for the consistency problem and its lack of a reduct transformation characterization. In this work, we present a new semantics which evaluates ontological atoms more strictly than the current semantics. This keeps the complexity of its consistency problem NP-complete, rather than increasing it to the second level of the polynomial hierarchy. Additionally, we identify a syntactic class of description logic programs for which our new semantics is equivalent to the current semantics. We characterize our semantics using a fixpoint operator and a reduct-based transformation. Our new semantics is a strict subset of the current well-supported semantics, so it maintains the prior notion of well-supportedness while inducing its own stricter notion. We prefer our new notion of well-supportedness due to its similarities with logic programming.

Chinese Translation

描述逻辑程序是一种强大的形式主义，用于将规则与本体结合。描述逻辑程序的良好支持语义确保没有答案集依赖于循环依赖。大多数流行的逻辑编程语义都具有这种良好支持的特性。我们认识到当前描述逻辑程序的良好支持语义存在两个局限性：其一致性问题的计算复杂性增加，以及缺乏还原变换的特征。在本研究中，我们提出了一种新的语义，它比当前的语义对本体原子进行更严格的评估。这使得其一致性问题的复杂性保持在 NP-完全，而不是增加到多项式层次的第二层。此外，我们识别出一种描述逻辑程序的语法类，对于该类程序，我们的新语义与当前语义等价。我们使用不动点算子和基于还原的变换来表征我们的语义。我们新提出的语义是当前良好支持语义的严格子集，因此它保持了先前的良好支持概念，同时引入了自己更严格的概念。由于与逻辑编程的相似性，我们更倾向于采用我们新的良好支持概念。

View on arXiv Download PDF AI Translation

cs.AI / 97 / 2607.21208

How Rules Represent Causal Knowledge: Causal Modeling with Probabilistic Logic Programming

规则如何表征因果知识：基于概率逻辑编程的因果建模

Rueckschloss, Kilian, Weitkaemper, Felix

Abstract

Pearl famously argues that causal knowledge enables the prediction of intervention effects. By contrast, purely descriptive knowledge supports only conclusions drawn from observations. His theory of causality, however, is developed exclusively within Bayesian networks and causal models. Consequently, it is largely restricted to acyclic causal relationships, and transferring its ideas to other formalisms risks misinterpretation or inconsistency. This paper brings Pearl's approach to causality into probabilistic logic programming (PLP). To this end, such programs are aligned with philosophical foundations established in prior work that do not rely on temporal notions; that is, all relevant events are assumed to occur simultaneously. A formal causal semantics for these programs, together with a notion of intervention and an implementation, is proposed. It is shown that this semantics coincides with the P-log semantics for stratified ProbLog programs, while the two may differ in the non-stratified case and for other PLP formalisms.

Chinese Translation

Pearl 著名地论证了因果知识能够预测干预效果。相比之下，纯描述性知识仅支持从观察中得出的结论。然而，他的因果理论完全是在贝叶斯网络和因果模型中发展起来的。因此，它在很大程度上局限于无环因果关系，并且将其思想转移到其他形式主义中可能会导致误解或不一致。本文将 Pearl 的因果方法引入到概率逻辑编程（PLP）中。为此，这些程序与先前工作中建立的哲学基础相一致，这些基础不依赖于时间概念；即假设所有相关事件同时发生。本文提出了这些程序的正式因果语义，以及干预的概念和实现。结果表明，这种语义与分层 ProbLog 程序的 P-log 语义一致，而在非分层情况下及其他 PLP 形式主义中，两者可能存在差异。

View on arXiv Download PDF AI Translation

cs.AI / 98 / 2607.21217

ICAE-Bench: Evaluating Coding Agents as Interactive Project Builders

ICAE-Bench：评估编码代理作为互动项目构建者

Peng, Zhongyuan, Huang, Dan, Zhang, Chuyu, Xu, Caijun, Xiao, Changyi, Hong, Shibo, Lo, David, Qiu, Lin, Cao, Xuezhi, He, Jiyuan, Cao, Yixin

Abstract

The recent emergence of vibe-coding workflows is changing what coding agents are expected to do. Instead of merely completing code under fully specified instructions, agents are increasingly expected to transform incomplete product intent into working software by combining various abilities including planning, requirement clarification, tool use, debugging, and repository-level construction. Yet existing benchmarks have not fully caught up with this shift, evaluating agents on static, fully specified tasks. In this paper, we introduce ICAE-Bench, a benchmark for evaluating coding agents under interactive project-building settings. The basic idea is to start from a fuzzy product requirement, simulating the dynamic paradigm with an automated User Agent. To make this setting both realistic and evaluable, ICAE-Bench introduces three key designs. First, to avoid the ambiguity of unconstrained fuzzy requirements, each task derives ambiguity from a precise real open-source repository with executable behavior. Second, to ensure high-quality and reproducible user simulation, ICAE-Bench grounds interaction through User Agent Data, allowing the User Agent to reveal hidden constraints without inventing new requirements or leaking implementation artifacts. Third, to evaluate open-ended repositories fairly, ICAE-Bench uses standardized black-box tests together with multi-dimensional diagnostics, including functional correctness, semantic and API similarity, structural fidelity, design quality, and interaction quality.

Chinese Translation

最近出现的氛围编码工作流程正在改变对编码代理的期望。代理不仅仅是根据完全指定的指令完成代码，而是越来越被期望通过结合规划、需求澄清、工具使用、调试和仓库级构建等多种能力，将不完整的产品意图转化为可工作的软件。然而，现有的基准测试尚未完全跟上这一变化，仍然在静态、完全指定的任务上评估代理。在本文中，我们介绍了ICAE-Bench，这是一个用于在互动项目构建环境中评估编码代理的基准。其基本思想是从模糊的产品需求出发，通过自动化用户代理模拟动态范式。为了使这一设置既真实又可评估，ICAE-Bench引入了三个关键设计。首先，为了避免不受限制的模糊需求的歧义，每个任务都从具有可执行行为的精确真实开源仓库中衍生出歧义。其次，为了确保高质量和可重复的用户模拟，ICAE-Bench通过用户代理数据（User Agent Data）来支撑交互，使用户代理能够揭示隐藏的约束，而不必发明新的需求或泄露实现工件。第三，为了公平评估开放式仓库，ICAE-Bench结合标准化的黑箱测试和多维诊断，包括功能正确性、语义和API相似性、结构保真度、设计质量和交互质量。

View on arXiv Download PDF AI Translation

cs.AI / 99 / 2607.21233

Logic Programming Semantics for Causal Processes

因果过程的逻辑编程语义

Weitkämper, Felix

Abstract

Motivated by challenging modelling issues in the life sciences, we investigate the relationship between logic programming semantics and the eventual states of causal processes compatible with those logic programs. More precisely, we show that while stable models of positive logic programs correspond to the eventual states of processes commencing from a neutral state and continuing undisturbed indefinitely, supported models describe the eventual states reachable from arbitrary starting points. This also contributes to the discussion of the appropriate semantics for logic programming as a causal rule language, adding a temporal perspective to recent interpretations of the stable and supported model semantics from an explanatory viewpoint of causality.

Chinese Translation

受生命科学中挑战性建模问题的启发，我们研究了逻辑编程语义与与这些逻辑程序兼容的因果过程的最终状态之间的关系。更准确地说，我们展示了正逻辑程序的稳定模型对应于从中立状态开始并无限期持续不受干扰的过程的最终状态，而支持模型描述了从任意起始点可达的最终状态。这也为逻辑编程作为因果规则语言的适当语义的讨论做出了贡献，从因果性的解释视角为最近对稳定模型和支持模型语义的解释增添了时间维度。

View on arXiv Download PDF AI Translation

cs.AI / 100 / 2607.21267

BasketEvent: Understanding Who Did What and When in Basketball Videos

BasketEvent：理解篮球视频中谁做了什么以及何时发生

Zhang, Yu, Rao, Jiayuan, Wu, Haoning, Xie, Weidi

Abstract

Comprehensive basketball video understanding requires resolving not only what event occurs, but also who is responsible and when the key evidence appears. However, exist- ing methods typically treat spatial perception and semantic recognition as isolated tasks, failing to ground events to individual players or pinpoint their temporal boundaries within complex collective dynamics. To bridge this gap, we introduce BasketEvent, a player- centric basketball event understanding dataset curated from real NBA broadcasts. In BasketEvent, event labels are grounded to the responsible players, and a manually an- notated subset of 1,000 samples with precise event intervals is provided to evaluate tem- poral evidence localization. Based on this data, we propose PlayNet, a player-centric reasoning framework that maps basketball videos to player-level event predictions with temporal evidence. Concretely, PlayNet tracks key entities, associates player identities, and reasons about events by modeling player-player, player-ball, and global court inter- actions, while aggregating sparse temporal evidence via gated pooling. Extensive experi- ments demonstrate that PlayNet significantly outperforms representative video-level and crop-based baselines, proving the superiority of player-centric modeling for fine-grained sports video understanding. Our data, code, and models will be made publicly available.

Chinese Translation

全面理解篮球视频不仅需要解析事件发生的内容，还需要明确谁是责任人以及关键证据何时出现。然而，现有的方法通常将空间感知和语义识别视为孤立的任务，未能将事件与个别球员关联或在复杂的集体动态中准确界定其时间边界。为了解决这一问题，我们引入了BasketEvent，这是一个以球员为中心的篮球事件理解数据集，来源于真实的NBA直播。在BasketEvent中，事件标签与负责的球员相对应，并提供了一个手动标注的1,000个样本的子集，包含精确的事件时间区间，以评估时间证据的定位。基于这些数据，我们提出了PlayNet，一个以球员为中心的推理框架，将篮球视频映射到球员级别的事件预测及时间证据。具体而言，PlayNet跟踪关键实体，关联球员身份，并通过建模球员与球员、球员与球、以及全场的互动来推理事件，同时通过门控池化聚合稀疏的时间证据。大量实验表明，PlayNet显著优于代表性的视频级和裁剪基线，证明了以球员为中心的建模在细粒度体育视频理解中的优越性。我们的数据、代码和模型将公开发布。

View on arXiv Download PDF AI Translation

cs.AI / 101 / 2607.21292

An LLM-Driven Workflow for Automated Process Control Strategy Generation and Tuning from Dynamic Process Models

基于大型语言模型的自动化过程控制策略生成与调优工作流

Rueda, Ari Luna, Cramer, Eike, Hellgardt, Klaus, Mercangöz, Mehmet

Abstract

We present a structured large-language-model-driven workflow for automated multi-variable control design from dynamic process models. The workflow decomposes the design task into constrained code-generation steps: plant-interface construction, normalization, manipulated-variable controlled-variable (MV-CV) pairing, controller specification, closed loop simulation, scenario generation, performance evaluation and Bayesian-optimization (BO) based tuning. Generated artifacts are executed and validated before downstream tasks proceed, and failed artifacts are repaired using validation feedback. The proposed approach is demonstrated on a nonlinear gas-preheater benchmark with coupled pressure and temperature dynamics. The generated workflow produces a physically consistent decentralized PI (proportional-integral) feedback-feedforward control structure and an executable tuning environment. Bayesian optimization reduces the closed loop performance objective, which aggregates set-point tracking and disturbance-rejection errors for the controlled variables, by approximately 26.5% relative to the initial controller generated by the workflow, mainly through improved pressure-loop transient performance. This figure quantifies the automated tuning stage rather than a comparison against a manually designed controller. The results demonstrate the feasibility of using structured large-language-model-based code generation to construct executable control-design workflows, while also highlighting the need for broader validation on larger plantwide-control benchmarks.

Chinese Translation

我们提出了一种结构化的大型语言模型驱动的工作流，用于从动态过程模型中自动化多变量控制设计。该工作流将设计任务分解为受限的代码生成步骤：工厂接口构建、归一化、操控变量与受控变量（MV-CV）配对、控制器规范、闭环仿真、场景生成、性能评估和基于贝叶斯优化（BO）的调优。生成的工件在下游任务进行之前被执行和验证，失败的工件则通过验证反馈进行修复。所提出的方法在一个具有耦合压力和温度动态的非线性气体预热器基准上进行了验证。生成的工作流产生了一个物理上一致的分散式比例-积分（PI）反馈-前馈控制结构和一个可执行的调优环境。贝叶斯优化将闭环性能目标降低了约26.5%，相对于工作流生成的初始控制器，主要通过改善压力环的瞬态性能实现。该数据量化了自动调优阶段，而不是与手动设计的控制器进行比较。结果表明，使用基于结构化大型语言模型的代码生成构建可执行控制设计工作流的可行性，同时也强调了在更大范围的全厂控制基准上进行更广泛验证的必要性。

View on arXiv Download PDF AI Translation

cs.AI / 102 / 2607.21302

Expert Behavior Prior Reinforcement Learning

专家行为先验强化学习

Gao, Gong, Zhao, Weidong, Liu, Xianhui, Jia, Ning

Abstract

Behavior prior reinforcement learning (BPRL) has emerged as a promising paradigm to improve sample efficiency in online reinforcement learning (RL) by leveraging policy priors derived from offline demonstrations. However, most existing BPRL methods rely on static offline datasets, which often suffer from low data diversity and suboptimal trajectory quality. This reliance restricts the effectiveness of policy priors, hindering both policy exploitation and stability during online training. Consequently, agents are prone to inefficient exploration and unstable learning dynamics. To address these limitations, we deviate from existing offline pre-training methods and propose an Expert Behavior Prior (EBP) algorithm. Specifically, we introduce a Q-guided conditional variational autoencoder (Q-CVAE) that learns to generate expert policy priors directly from the online replay buffer. This enables the generation of high-value actions for guiding policy updates without relying on pre-collected expert trajectories. To further enhance policy exploitation, we propose an expert policy guidance (EPG) mechanism that selects expert actions from a generative support set, and we integrate a policy gradient correction (PGC) module to harmonize Q-guidance with expert supervision, promoting stable and consistent policy improvement. Extensive experiments conducted on robotic control (Gym, PyBullet) and industrial control (DMControl) benchmarks demonstrate that EBP significantly outperforms state-of-the-art online RL algorithms, achieving higher sample efficiency and more stable convergence.

Chinese Translation

行为先验强化学习（BPRL）作为一种有前景的范式，通过利用来自离线演示的策略先验，提升在线强化学习（RL）的样本效率。然而，现有的大多数BPRL方法依赖于静态的离线数据集，这些数据集通常存在数据多样性不足和轨迹质量不佳的问题。这种依赖限制了策略先验的有效性，阻碍了在线训练中的策略利用和稳定性。因此，智能体容易出现低效探索和不稳定的学习动态。为了解决这些局限性，我们偏离现有的离线预训练方法，提出了一种专家行为先验（EBP）算法。具体而言，我们引入了一种基于Q值的条件变分自编码器（Q-CVAE），该模型能够直接从在线重放缓冲区生成专家策略先验。这使得在不依赖于预先收集的专家轨迹的情况下，能够生成高价值的动作以指导策略更新。为了进一步增强策略利用，我们提出了一种专家策略引导（EPG）机制，从生成支持集选择专家动作，并整合了策略梯度修正（PGC）模块，以协调Q引导与专家监督，促进稳定且一致的策略改进。在机器人控制（Gym, PyBullet）和工业控制（DMControl）基准上进行的大量实验表明，EBP显著优于最先进的在线RL算法，实现了更高的样本效率和更稳定的收敛性。

View on arXiv Download PDF AI Translation

cs.AI / 103 / 2607.21345

Regulating autonomous and agentic AI

监管自主和有能动性的人工智能

Reed, Chris, Austria, Alex, Bharuka, Anmol, Mandava, Pragnitha, Mujawar, Khushiya, Shakhkulashvili, Luka

Abstract

Regulating activities where regulatees use autonomous and agentic AI is challenging. Regulatory assumptions about regulatee knowledge and control no longer hold true; much of that lies elsewhere in the AI supply chain which thus needs to be brought within the scope of regulation. Governance systems for autonomous AI cannot replicate existing governance models, but need a fresh approach. Retrospective supervisory oversight becomes ineffective as a risk management tool, and AI autonomy generates new systemic risks which require new solutions. This paper investigate four regulatory systems: UK regulation of content platforms, data protection, UK financial services, and the EU AI Act\'92s cross-sectoral regime. It analyses the challenges posed by autonomous and agentic AI and proposes potential solutions which regulators might adopt. These will transform regulation from a reactive process to an active one, and assist it in adapting to the challenges of AI autonomy.

Chinese Translation

监管被监管者使用自主和有能动性的人工智能的活动具有挑战性。关于被监管者知识和控制的监管假设已不再成立；许多相关因素位于人工智能供应链的其他环节，因此需要将其纳入监管范围。自主人工智能的治理系统无法复制现有的治理模型，而需要一种全新的方法。事后监督变得无效，作为风险管理工具，而人工智能的自主性产生了新的系统性风险，这需要新的解决方案。本文研究了四个监管系统：英国内容平台的监管、数据保护、英国金融服务以及欧盟人工智能法案（EU AI Act）跨部门制度。它分析了自主和有能动的人工智能所带来的挑战，并提出了监管者可能采用的潜在解决方案。这些解决方案将把监管从被动过程转变为主动过程，并帮助其适应人工智能自主性带来的挑战。

View on arXiv Download PDF AI Translation

cs.AI / 104 / 2607.21354

SPORD: A Simulation-Propose-then-OR-Dispose Approach for Supply Chain Planning

SPORD：一种用于供应链规划的模拟-提议-然后-优化-处置方法

He, Jiayin, Pan, Yutong, Yang, Sen, Kang, Ningxuan, Qi, Yongzhi, Zhang, Jianshen, Qi, Wei, Shen, Zuo-Jun Max

Abstract

For years, supply chain planning at e-commerce firms has operated as a collection of isolated projects. Each planning task from static network planning to dynamic warehouse assortment planning requires analysts to spend weeks building models from scratch, calibrating and persuading executives to act on outputs they cannot verify. Three barriers drive this: bespoke models proliferate because standardization is difficult (operational fragmentation); once unified, the combinatorial scale of millions of SKUs, thousands of nodes, and intricate routing logic exceeds what solvers can handle within a tight window (computational intractability); and a mathematically optimal solution still fails to be implemented if the executives do not trust it (implementation hurdle). To bridge this gap, we propose and implement the Simulation-Propose-then-OR-Dispose method, deployed as JD.com's NetSim platform. The central insight is decoupling: simulation proposes by generating and evaluating the full set of operationally valid candidate paths-absorbing all idiosyncratic business logic, while an integer program disposes by selecting the globally optimal subset. Computationally, matrix-vectorized CPU/GPU accelerated simulation achieves a 10-100 times speedup over serial methods, and a list scheduling algorithm reduces coupled-order processing from hours to minutes. Operationally, we establish a closed loop via an intelligent diagnosis engine. Since 2025, NetSim has optimized end to-end services for over 20,000 suppliers, the cross-regional fulfillment rate dropped from 6.1% to 4.9%, and the average monthly carbon reduction is approximately 5,745 tCO2e. SPORD moves simulation from monitoring to active planning. The transparent outputs turn skeptical executives into engaged collaborators, and the modular architecture ensures that the next planning requires just configuration, not reconstruction.

Chinese Translation

多年来，电子商务公司的供应链规划一直作为一系列孤立的项目进行。每个规划任务，从静态网络规划到动态仓库品类规划，都需要分析师花费数周时间从头构建模型，进行校准，并说服高管采取他们无法验证的输出。这背后有三个障碍：定制模型的泛滥是因为标准化困难（操作碎片化）；一旦统一，数百万SKU、数千个节点和复杂的路由逻辑的组合规模超出了求解器在紧迫时间内能够处理的范围（计算不可处理性）；而且，如果高管不信任，数学上最优的解决方案仍然无法实施（实施障碍）。为了弥合这一差距，我们提出并实施了模拟-提议-然后-优化-处置方法，作为京东的NetSim平台进行部署。核心见解是解耦：模拟通过生成和评估所有操作上有效的候选路径来提出建议，吸收所有特有的业务逻辑，而整数规划则通过选择全局最优子集来进行处置。在计算上，矩阵向量化的CPU/GPU加速模拟实现了比串行方法快10到100倍的速度提升，而列表调度算法将耦合订单处理时间从数小时减少到数分钟。在操作上，我们通过智能诊断引擎建立了一个闭环。自2025年以来，NetSim已为超过20,000个供应商优化了端到端服务，跨区域履约率从6.1%降至4.9%，平均每月减少碳排放约5,745吨二氧化碳当量。SPORD将模拟从监控转向主动规划。透明的输出使怀疑的高管变成了积极的合作者，而模块化架构确保下一个规划只需配置，而非重建。

View on arXiv Download PDF AI Translation

cs.AI / 105 / 2607.21381

Towards Faithful Graph Explanations with Synergistic Edge Effects via Granular Balls

通过粒状球体实现协同边效应的可信图解释

Chen, Jiancu, Xia, Shuyin, Wang, Guan, Chen, Degang, Chen, Fan

Abstract

Instance-level explanations aim to reveal the rationale behind a model's decisions for a specific graph. Previous methods explain graph neural networks (GNNs) by selecting important edges to induce subgraphs, where edge importance is assessed by perturbing each edge and observing changes in the model predictions. However, they often neglect the synergistic effects among edges, which are crucial for accurately characterizing edge importance. To address this issue, we propose SeeExplainer, a parameter-free explainer to interpret GNNs. Specifically, we first introduce a granular-ball graph refinement mechanism that decomposes a graph into several disjoint granular-balls with no fixed size, and utilize them as nodes to construct a structural graph. This process can better capture the synergistic effects among edges. Then, we perturb nodes and edges in the structural graph to generate explanatory subgraphs based on their respective contributions. Experiments on several graph classification datasets of different networks show that SeeExplainer outperforms state-of-the-art baselines.

Chinese Translation

实例级解释旨在揭示模型在特定图形上的决策背后的原理。以往的方法通过选择重要边缘来解释图神经网络（GNN），以诱导子图，其中边缘重要性通过扰动每条边并观察模型预测的变化来评估。然而，这些方法往往忽视了边缘之间的协同效应，而这种效应对于准确表征边缘重要性至关重要。为了解决这一问题，我们提出了 SeeExplainer，这是一种无参数的解释器，用于解释 GNN。具体而言，我们首先引入了一种粒状球体图形细化机制，将图形分解为若干个不固定大小的互不相交的粒状球体，并利用它们作为节点构建结构图。这一过程能够更好地捕捉边缘之间的协同效应。然后，我们在结构图中扰动节点和边缘，以根据它们各自的贡献生成解释性子图。在多个不同网络的图分类数据集上的实验表明，SeeExplainer 超越了最先进的基线方法。

View on arXiv Download PDF AI Translation

cs.AI / 106 / 2607.21384

Multimodal Pretraining for Generalizable EEG Representation Learning

用于可泛化脑电图表示学习的多模态预训练

Bakhtiarvand, Targol, Kalita, Jugal, Atyabi, Adham

Abstract

Electroencephalography (EEG) models used for epilepsy are often limited to specific datasets and tasks. This limited approach can make it challenging to apply these models across different datasets or in various situations. However, recent studies in foundation models and self-supervised learning suggest that an adaptable EEG backbone could support a range of EEG related tasks. In this study, we have developed a multimodal EEG foundation model that combines a raw signal encoder based on the Mamba architecture, a Vision Transformer (ViT)-style encoder for time-frequency data, and a lightweight encoder for text, all within a shared embedding space. The pretraining process relies on several innovative techniques, such as masked modeling, cross-view contrastive alignment, and temporal consistency losses. These methods are designed to create rich, seizure-relevant representations without requiring labeled data. To assess the efficacy and generalization of our pretrained model, we fine-tuned it on the canonical CHB-MIT seizure detection benchmark and additional seizure detection datasets, and conducted extensive experiments comparing different model variants. On the standard CHB-MIT split, our best single model achieved an AUROC of 0.874, and an ensemble variant reached 0.878 AUROC, representing state-of-the-art performance on this benchmark. In addition to standard train-test splits, we evaluated performance under a leave-one-subject-out (LOSO) protocol, which is rarely reported in prior EEG seizure modeling work and highlights the difficulty of patient-independent seizure detection, with a mean LOSO balanced accuracy of 0.558 across 19 subjects. Across datasets and evaluation settings, our multimodal foundation model enabled robust seizure detection and straightforward adaptation to new seizure detection scenarios, while also supporting interpretable seizure localization.

Chinese Translation

用于癫痫的脑电图（EEG）模型通常局限于特定的数据集和任务。这种有限的方法使得在不同数据集或各种情况下应用这些模型变得具有挑战性。然而，最近在基础模型和自监督学习方面的研究表明，一个可适应的EEG骨干网络可以支持一系列与EEG相关的任务。在本研究中，我们开发了一种多模态EEG基础模型，该模型结合了基于Mamba架构的原始信号编码器、用于时频数据的视觉变换器（Vision Transformer, ViT）风格编码器以及一个轻量级文本编码器，所有这些都在一个共享的嵌入空间内。预训练过程依赖于多种创新技术，如掩蔽建模、跨视图对比对齐和时间一致性损失。这些方法旨在创建丰富的、与癫痫发作相关的表示，而无需标记数据。为了评估我们预训练模型的有效性和泛化能力，我们在经典的CHB-MIT癫痫发作检测基准和其他癫痫发作检测数据集上进行了微调，并进行了广泛的实验比较不同模型变体。在标准的CHB-MIT划分上，我们的最佳单模型达到了0.874的AUROC，而一个集成变体达到了0.878的AUROC，代表了该基准上的最先进性能。除了标准的训练-测试划分外，我们还在一个留一被试法（Leave-One-Subject-Out, LOSO）协议下评估了性能，这在以往的EEG癫痫建模工作中很少报告，突显了患者独立癫痫检测的困难，19名被试的平均LOSO平衡准确率为0.558。在不同数据集和评估设置中，我们的多模态基础模型实现了稳健的癫痫检测，并能够轻松适应新的癫痫检测场景，同时也支持可解释的癫痫定位。

View on arXiv Download PDF AI Translation

cs.AI / 107 / 2607.21402

MSBraM: A Multi-scale Self-supervised Brain Foundation Model for Hierarchical EEG Dynamics Learning

MSBraM：一种用于分层脑电图动态学习的多尺度自监督基础模型

Zhou, Tao, Han, Jing, Shu, Lingyu, Zhang, Zixing

Abstract

Self-supervised foundation models have recently shown strong potential for electroencephalogram (EEG)-based analysis. However, existing approaches struggle to capture the inherently multi-scale temporal structure of EEG signals, where local neural patterns and long-range dependencies jointly encode task-relevant information. This limitation hampers cross-scale representation learning and generalization across diverse downstream tasks. To address this challenge, we propose MSBraM, a Multi-Scale self-supervised Brain foundation Model designed to learn hierarchical EEG representations. MSBraM follows a two-stage pretraining framework. First, a multi-scale neural tokenizer discretizes raw EEG signals into semantic codes at different temporal resolutions via vector-quantized reconstruction. Second, the model is pretrained to predict masked codes using a curriculum multi-scale masking strategy, progressively integrating fine-grained local patterns with global temporal context. We pretrain MSBraM on over 2,400 hours of EEG data and evaluate it across 10 downstream tasks on 12 public datasets. Extensive experiments show that MSBraM achieves superior performance on other state-of-the-art pretrained models, demonstrating strong generalization and transferability. These results indicate that explicitly modeling multi-scale temporal dynamics is critical for effective EEG foundation models.

Chinese Translation

自监督基础模型最近在基于脑电图（EEG）的分析中展现出强大的潜力。然而，现有方法难以捕捉EEG信号固有的多尺度时间结构，其中局部神经模式和长程依赖共同编码与任务相关的信息。这一局限性阻碍了跨尺度表示学习和在多样化下游任务中的泛化能力。为了解决这一挑战，我们提出了MSBraM，一种旨在学习分层EEG表示的多尺度自监督脑基础模型。MSBraM遵循两阶段预训练框架。首先，多尺度神经标记器通过向量量化重建将原始EEG信号离散化为不同时间分辨率的语义代码。其次，模型通过课程多尺度掩蔽策略进行预训练，以预测被掩蔽的代码，逐步将细粒度的局部模式与全局时间上下文相结合。我们在超过2400小时的EEG数据上预训练MSBraM，并在12个公共数据集上的10个下游任务中进行评估。大量实验表明，MSBraM在其他最先进的预训练模型上表现出色，展现出强大的泛化能力和迁移能力。这些结果表明，明确建模多尺度时间动态对于有效的EEG基础模型至关重要。

View on arXiv Download PDF AI Translation

cs.AI / 108 / 2607.21412

Euclid-MCP: A Model Context Protocol Server for Deterministic Logical Reasoning via Prolog

Euclid-MCP：一个用于通过Prolog进行确定性逻辑推理的模型上下文协议服务器

Bogliolo, Bartolomeo

Abstract

Large Language Models (LLMs) excel at natural language understanding and generation but remain unreliable for multi-step logical reasoning, especially in safety-critical or compliance-sensitive domains. Recent neuro-symbolic approaches address this gap by coupling neural models with external symbolic engines, yet most integrations are bespoke and lack a standardized interface for tool-augmented agents. This paper presents Euclid-MCP, an open-source MCP server that provides deterministic logical reasoning via SWI-Prolog. Euclid-MCP introduces Euclid-IR, an engine-agnostic intermediate representation for Horn-clause logic that is human-readable, easy for LLMs to generate, and straightforward to compile into Prolog or alternative backends. The server exposes a compact tool interface that supports a translate-run-inspect-repair loop, enabling LLM clients to delegate inference while retaining full access to proof traces and derivation logs. We evaluate Euclid-MCP on a realistic IT security and compliance use case. Results show that while LLMs alone are sufficient on small knowledge bases, they hallucinate systematically on larger problems, whereas Euclid-MCP delivers exact answers with lower latency and more compact outputs. We argue that semantic RAG is fundamentally unsuited for rule enforcement, and that Euclid-MCP can serve as a stable, shared reasoning substrate for both RAG-based assistants and agentic systems.

Chinese Translation

大型语言模型（LLMs）在自然语言理解和生成方面表现出色，但在多步骤逻辑推理方面仍然不可靠，尤其是在安全关键或合规敏感的领域。近期的神经符号方法通过将神经模型与外部符号引擎结合来解决这一问题，但大多数集成都为定制，缺乏用于工具增强代理的标准化接口。本文提出了Euclid-MCP，一个开源的MCP服务器，通过SWI-Prolog提供确定性逻辑推理。Euclid-MCP引入了Euclid-IR，这是一种与引擎无关的霍恩子句逻辑中间表示，具有可读性强、易于LLMs生成，并且易于编译为Prolog或其他后端的特点。该服务器提供了一个紧凑的工具接口，支持翻译-运行-检查-修复循环，使LLM客户端能够在保留对证明轨迹和推导日志的完全访问权限的同时，委托推理。我们在一个现实的IT安全和合规用例上评估了Euclid-MCP。结果表明，虽然单独使用LLMs在小型知识库上是足够的，但在较大问题上它们系统性地产生幻觉，而Euclid-MCP则以更低的延迟和更紧凑的输出提供准确的答案。我们认为，语义RAG在规则执行方面根本不适用，而Euclid-MCP可以作为基于RAG的助手和自主系统的稳定共享推理基础。

View on arXiv Download PDF AI Translation

cs.AI / 109 / 2607.21414

Logical Regression for Planning with Axioms

带公理的规划逻辑回归

Little, Connor, Muise, Christian

Abstract

In automated planning, logical regression is an operation that returns the most general condition necessary for an action to achieve a particular formula. It has many applications, such as allowing for more robust plan execution and providing compact policies for non-deterministic planning. Although relatively simple to calculate in basic planning settings, logical regression becomes significantly more complex when additional factors, such as axioms, are present. We introduce a methodology for approximating the logical regression of an action in a domain that includes axioms; an approximation that limits conditions to partial states. Our method produces minimal partial states while avoiding the recalculation of axioms. To demonstrate the impact of our methods, we embed our form of regression in an execution monitoring context, a well-established setting that can benefit greatly from logical regression. Our results show that this form of regression can dramatically generalize partial states across multiple domains, reducing the number of variables considered for execution monitoring by up to 70%, and demonstrate that the resulting execution monitor is robust enough to recover frequently in an environment with unexpected changes: several domains recover over 50% of the time in our tests.

Chinese Translation

在自动化规划中，逻辑回归是一种操作，它返回为使某个特定公式成立而所需的最一般条件。它有许多应用，例如允许更稳健的计划执行，并为非确定性规划提供紧凑的策略。尽管在基本规划环境中计算相对简单，但当存在额外因素（如公理）时，逻辑回归的复杂性显著增加。我们提出了一种方法，用于近似包含公理的领域中某个动作的逻辑回归；该近似将条件限制为部分状态。我们的方法生成最小的部分状态，同时避免了公理的重新计算。为了展示我们方法的影响，我们将这种回归形式嵌入到执行监控的上下文中，这是一个成熟的环境，可以从逻辑回归中获得巨大收益。我们的结果表明，这种回归形式可以在多个领域显著推广部分状态，将执行监控中考虑的变量数量减少多达70%，并且证明所得到的执行监控器在环境发生意外变化时足够稳健，能够频繁恢复：在我们的测试中，多个领域的恢复率超过50%。

View on arXiv Download PDF AI Translation

cs.AI / 110 / 2607.21419

PATS: Policy-Aware Training Scaffolding for Agentic Reinforcement Learning

PATS：面向策略的代理强化学习训练支架

Shi, Yipeng, Ma, Zhipeng, Wang, Yue, Tan, Qitai, Li, Yang, Chen, Peng, Zhu, Zhengzhou

Abstract

In long-horizon LLM agent reinforcement learning, weak policies often repeat similar failures, producing uninformative rollout trajectories and limiting effective policy optimization. Existing skill-centric methods improve exploration by optimizing, filtering, or internalizing reusable skills. However, they remain centered on the skills themselves rather than being designed as adaptive training-time support for the evolving policy. To address this, we propose a policy-centric training paradigm that reframes skills as a dynamic training scaffold. Our framework, Pats, converts rollout groups from the latest policy into evidence cards and uses task-specific evaluation to adjust the context used in subsequent rollouts. Concrete guidance helps weak policies to complete challenging tasks. As policy improves, redundant context is revised or removed to reduce reliance on explicit guidance while preserving useful rollout variation. The policy is optimized with environmental rewards using standard RLVR, and the training scaffold is discarded at deployment. On ALFWorld and WebShop, Pats improves over strong baselines by up to 18.6%. Across seven search-augmented QA benchmarks, it remains competitive while using 32.1% fewer prompt tokens than the baseline.

Chinese Translation

在长时间跨度的LLM代理强化学习中，弱策略往往重复类似的失败，产生无信息的回滚轨迹，并限制有效的策略优化。现有的以技能为中心的方法通过优化、过滤或内化可重用技能来改善探索。然而，这些方法仍然集中于技能本身，而不是作为适应性训练支持来设计以适应不断发展的策略。为了解决这个问题，我们提出了一种以策略为中心的训练范式，将技能重新构建为动态训练支架。我们的框架Pats将最新策略的回滚组转换为证据卡，并使用任务特定的评估来调整后续回滚中使用的上下文。具体的指导帮助弱策略完成具有挑战性的任务。随着策略的改善，多余的上下文被修订或移除，以减少对显式指导的依赖，同时保留有用的回滚变异。该策略通过标准的RLVR使用环境奖励进行优化，并在部署时丢弃训练支架。在ALFWorld和WebShop上，Pats的表现比强基线提高了多达18.6%。在七个增强搜索的QA基准测试中，它在使用比基线少32.1%的提示令牌的情况下仍然保持竞争力。

View on arXiv Download PDF AI Translation

cs.AI / 111 / 2607.21421

Bridging the Gap Between Plausibility and Admissibility: Constraint-Aware Flow Maps for Dynamic Graph Systems

弥合可信性与可接受性之间的差距：面向动态图系统的约束感知流图

de Socio, Michael Romei, Pozzato, Gian Luca, Merlo, Alessio

Abstract

Generative models can support decision-making under uncertainty by producing ensembles of plausible future system trajectories, but statistical plausibility does not ensure structural feasibility. This study investigates whether post-sampling symbolic constraints can improve the reliability of generative trajectory modeling in dynamic graph-structured systems. A conditional diffusion model generates future graph-state trajectories from partial observations, while an external symbolic layer applies hard filtering, soft weighting, or projection-based repair. The framework is evaluated on two controlled synthetic regimes: a compact graph and a medium-complexity dependency graph, using metrics for structural validity, sample efficiency, diversity, robustness, and calibration. In the compact regime, the model produces an invalid probability mass of 0.002996, indicating an almost entirely admissible trajectory manifold. Under the same architecture and training protocol, invalid mass increases to 0.155929 in the medium-complexity regime. Hard filtering removes all invalid retained trajectories while preserving 84.4% of generated samples, whereas soft weighting preserves effective sample size but yields only limited validity gains. Family-level analysis shows that dependency constraints account for nearly all observed inadmissibility. These results indicate that statistical plausibility and structural admissibility are distinct reliability properties and that symbolic constraint handling becomes more valuable as graph-structural complexity increases.

Chinese Translation

生成模型可以通过生成一组可信的未来系统轨迹来支持不确定性下的决策，但统计可信性并不确保结构可行性。本研究探讨了后采样符号约束是否能够提高动态图结构系统中生成轨迹建模的可靠性。条件扩散模型从部分观测中生成未来图状态轨迹，而外部符号层则应用硬过滤、软加权或基于投影的修复。该框架在两个受控的合成环境中进行了评估：一个紧凑图和一个中等复杂度的依赖图，使用结构有效性、样本效率、多样性、鲁棒性和校准等指标。在紧凑环境中，该模型产生了0.002996的无效概率质量，表明几乎完全可接受的轨迹流形。在相同的架构和训练协议下，中等复杂度环境中的无效质量增加到0.155929。硬过滤去除了所有无效的保留轨迹，同时保留了84.4%的生成样本，而软加权则保留了有效样本量，但仅带来了有限的有效性提升。家庭级分析表明，依赖约束几乎占据了所有观察到的不可接受性。这些结果表明，统计可信性和结构可接受性是不同的可靠性属性，并且随着图结构复杂性的增加，符号约束处理变得更加重要。

View on arXiv Download PDF AI Translation

cs.AI / 112 / 2607.21437

Agent-Guided Relational Concept Discovery: Toward Interpretable Surgical Margin Assessment

基于代理的关系概念发现：迈向可解释的手术边缘评估

Maghsoodi, Nooshin, Jamzad, Amoon, Policelli, Robert, Farahmand, Mohammad, Srikanthan, Dilakshan, Kaufmann, Martin, Ren, Kevin Y. M., Merchant, Shaila, Varma, Sonal, Walker, Ross, McKay, Doug, Rudan, John, Fichtinger, Gabor, Mousavi, Parvin

Abstract

Deep learning models can effectively use Rapid Evaporative Ionization Mass Spectrometry (REIMS) data for surgical margin assessment. However, their clinical adoption remains challenging due to limited generalization to operating room conditions. This difficulty arises because models are typically trained on labeled spectra collected from resected tissue samples, while they must operate on noisy, unlabeled data acquired directly during surgery. In addition, the black-box nature of deep learning models makes it difficult to understand and systematically improve their behavior. Concept-based learning offers a promising way to address these challenges by mapping raw measurements to human-understandable concepts. However, supervised concept-based approaches rely on concept annotations, which are difficult to obtain in complex mass spectrometry workflows. We propose Agent-Guided Concept Discovery, a framework that learns meaningful concepts directly from data without requiring predefined concept labels. During training, a reasoning agent refines semantic descriptions of the learned concepts and adaptively adjusts their weight based on diagnostic relevance. These concepts are further grounded using a biochemical knowledge graph to ensure consistency with known metabolic relationships. Across Skin and Breast Cancer datasets, our model improves balanced accuracy and sensitivity over the baseline. In a representative intraoperative case, it shows fewer false positives, indicating better generalization to surgical conditions.

Chinese Translation

深度学习模型可以有效利用快速蒸发离子化质谱（Rapid Evaporative Ionization Mass Spectrometry, REIMS）数据进行手术边缘评估。然而，由于在手术室条件下的有限泛化能力，其临床应用仍面临挑战。这一困难源于模型通常是在从切除组织样本中收集的标记光谱上进行训练，而它们必须在手术过程中直接获取的嘈杂、未标记数据上进行操作。此外，深度学习模型的黑箱特性使得理解和系统性改进其行为变得困难。基于概念的学习提供了一种有前景的方法，通过将原始测量映射到人类可理解的概念来解决这些挑战。然而，监督的基于概念的方法依赖于概念注释，而在复杂的质谱工作流程中，这些注释往往难以获得。我们提出了基于代理的概念发现（Agent-Guided Concept Discovery）框架，该框架直接从数据中学习有意义的概念，而不需要预定义的概念标签。在训练过程中，推理代理细化所学概念的语义描述，并根据诊断相关性自适应调整其权重。这些概念进一步通过生物化学知识图谱进行基础化，以确保与已知的代谢关系一致。在皮肤癌和乳腺癌数据集中，我们的模型在平衡准确性和灵敏度上优于基线。在一个代表性的术中案例中，它显示出更少的假阳性，表明对手术条件的更好泛化能力。

View on arXiv Download PDF AI Translation

cs.AI / 113 / 2607.21458

Detecting LLM-Generated Tokens in Human--LLM Coauthored Text

检测人类与大型语言模型共同创作文本中的LLM生成标记

Lu, Yangjun, Zhou, Hongyi, Spill, Fabian, Ye, Kai, Shi, Chengchun, Zhu, Jin

Abstract

The rise of human-AI collaborative writing has created a growing need for fine-grained detection methods that support localizing likely LLM-generated content in mixed-authorship documents. Existing methods for detecting LLM-generated text mainly focus on document-level classification and cannot identify which parts of the text are generated by LLMs. This paper introduces a new method to address this urgent need. Our method operates at the token level, the natural unit of modern language models, and builds on existing token-level detection scores. The key idea is to smooth adjacent token scores to reduce their variability, while using an adaptive Lepski-type rule to select the bandwidth according to the local authorship structure. Our method is simple to implement and does not require token-level labeled data for training. Theoretically, we characterize this trade-off and show that the proposed method achieves favorable mean square error performance in estimating the underlying signal. Empirically, we demonstrate strong performance of our method against a wide range of baselines in both synthetic datasets and a realistic dataset. We deploy a publicly accessible website that implements the methods as well.

Chinese Translation

人类与人工智能协作写作的兴起使得对细粒度检测方法的需求日益增长，这些方法能够支持在混合作者文档中定位可能由大型语言模型（LLM）生成的内容。现有的检测LLM生成文本的方法主要集中在文档级分类上，无法识别文本中哪些部分是由LLM生成的。本文提出了一种新的方法来解决这一紧迫需求。我们的方法在标记级别上操作，这是现代语言模型的自然单位，并基于现有的标记级检测评分。关键思想是平滑相邻标记的评分以减少其变异性，同时使用自适应Lepski类型规则根据局部作者结构选择带宽。我们的方法实现简单，不需要标记级别的标注数据进行训练。从理论上讲，我们描述了这一权衡，并展示了所提出的方法在估计潜在信号时实现了良好的均方误差性能。从经验上看，我们在合成数据集和真实数据集上展示了我们的方法相对于广泛基线的强大性能。此外，我们还部署了一个公开可访问的网站，以实现这些方法。

View on arXiv Download PDF AI Translation

cs.AI / 114 / 2607.21461

AREX: Towards a Recursively Self-Improving Agent for Deep Research

AREX：朝着递归自我改进的深度研究代理迈进

Lu, Shuqi, Li, Chaofan, Luo, Kun, Zhang, Zhang, Wang, Hui, Xiao, Hongwang, Liu, Zheng, Xiong, Lei, Wang, Jiahao, Wang, Sen, Jiang, Xiyan, Li, Wanli, Hu, Yuyang, Qian, Hongjin, Yan, Bingyu, Xia, Ziyi, Shao, Yingxia, Liu, Kang, Dou, Zhicheng, He, Di, Li, Chaozhuo, Ye, Qiwei, Wang, Zhongyuan, Liu, Zheng

Abstract

Deep research requires agents to find answers that jointly satisfy multiple constraints. Discovering such answers is costly, whereas verifying a candidate can often be decomposed into tractable constraint-wise checks. This discovery--verification asymmetry suggests that a research agent should do more than simply search longer: it should recursively improve its current answer by verifying intermediate results and using the partially verified state to guide subsequent refinement. We introduce AREX, a family of Recursively Self-Improving (RSI) deep research agents. AREX alternates between an inner research loop that gathers evidence and constructs a provisional answer, and an outer self-improvement loop that audits the answer constraint-wise, identifies unresolved claims, and launches targeted follow-up research. To sustain RSI over long horizons, AREX learns an autonomous context-update tool that compresses growing interaction history into a compact improvement state preserving verified evidence and unresolved constraints, without relying on an external model. We train AREX on verified synthetic tasks and high-quality trajectories through agentic mid-training and long-horizon reinforcement learning. To mitigate sparse final rewards during long horizon learning, we emphasize key steps where decisive evidence is acquired or erroneous research directions are corrected. We instantiate a dense 4B model and a 122B-A10B Mixture-of-Experts model. Across BrowseComp, WideSearch, DeepSearchQA, Humanity's Last Exam (HLE), and other reasoning and tool-use benchmarks, AREX substantially outperforms comparable-scale baselines and remains competitive with models using substantially more activated parameters.

Chinese Translation

深度研究要求代理找到能够共同满足多个约束的答案。发现这样的答案成本高昂，而验证候选答案通常可以分解为可处理的约束检查。这种发现与验证的不对称性表明，研究代理应该做的不仅仅是延长搜索时间：它应该通过验证中间结果来递归改进当前答案，并利用部分验证的状态来指导后续的细化。我们介绍了AREX，一类递归自我改进（Recursively Self-Improving, RSI）深度研究代理。AREX在一个内部研究循环和一个外部自我改进循环之间交替进行，内部循环收集证据并构建临时答案，外部循环则对答案进行约束审计，识别未解决的主张，并启动针对性的后续研究。为了在长时间范围内维持RSI，AREX学习了一种自主的上下文更新工具，该工具将不断增长的交互历史压缩为一个紧凑的改进状态，保留已验证的证据和未解决的约束，而不依赖于外部模型。我们在经过验证的合成任务和高质量的轨迹上训练AREX，通过代理的中期训练和长时间范围的强化学习。为了减轻在长时间范围学习中稀疏的最终奖励，我们强调关键步骤，在这些步骤中获取决定性证据或纠正错误的研究方向。我们实例化了一个密集的4B模型和一个122B-A10B混合专家模型。在BrowseComp、WideSearch、DeepSearchQA、人类最后的考试（Humanity's Last Exam, HLE）及其他推理和工具使用基准测试中，AREX显著优于可比规模的基线，并与使用显著更多激活参数的模型保持竞争力。

View on arXiv Download PDF AI Translation

cs.AI / 115 / 2607.21482

Agentic coding without the cloud: evaluating open-weight large language models on longitudinal data preparation tasks

无云的代理编码：评估开放权重大型语言模型在纵向数据准备任务中的表现

Nixon, Mack, Wright, Liam, Kovalchuk, Yevgeniya, Wu, Alison Fang-Wei, Danka, Martin, Boyd, Andy, Bann, David

Abstract

Large language models (LLMs) and agents are now widely used tools in code development, with data typically sent to third-party cloud-based models. Their adoption in research using personal data is constrained by governance requirements that typically prohibit data transmission to external services. Locally deployable open-weight models offer an alternative since sensitive data never leave the local environment. We introduce an open-source framework for evaluating the efficacy of AI agents powered by open-weight LLMs on one of the most persistent bottlenecks in research on longitudinal population studies: data preparation. The framework comprises: a curated ground-truth dataset (cleaning scripts preparing six sweeps of data from a British cohort study), task definitions encompassing tasks such as category harmonization and multi-wave merging, and automated routines for evaluating the LLM-produced R code and outputted data. We benchmark LLMs across the (consumer grade) deployment spectrum to assess their efficacy in 20 data preparation tasks (creation of 102 variables). Current state-of-the-art, 31-35B parameter models almost saturated our benchmark ("average task completion" up to 87.9%). The performance of open-weight LLMs running on consumer-grade hardware shows promise of a viable path toward AI-assisted data preparation in governance-restricted research settings. Our framework is publicly available at: https://github.com/UCL-ARC/RRBench.

Chinese Translation

大型语言模型（LLMs）和代理现在已成为代码开发中广泛使用的工具，数据通常被发送到第三方基于云的模型。由于治理要求通常禁止将个人数据传输到外部服务，它们在使用个人数据的研究中的采用受到限制。可本地部署的开放权重模型提供了一种替代方案，因为敏感数据从未离开本地环境。我们介绍了一个开源框架，用于评估由开放权重LLMs驱动的AI代理在纵向人口研究中最持久瓶颈之一：数据准备上的有效性。该框架包括：一个经过策划的真实数据集（清理脚本准备来自英国队列研究的六轮数据）、涵盖类别协调和多波合并等任务的任务定义，以及评估LLM生成的R代码和输出数据的自动化例程。我们在（消费级）部署范围内对LLMs进行了基准测试，以评估它们在20个数据准备任务（创建102个变量）中的有效性。当前的最先进的31-35B参数模型几乎饱和了我们的基准测试（“平均任务完成率”高达87.9%）。在消费级硬件上运行的开放权重LLMs的表现显示出在治理受限的研究环境中实现AI辅助数据准备的可行路径的希望。我们的框架可在以下网址公开获取：https://github.com/UCL-ARC/RRBench。

View on arXiv Download PDF AI Translation

cs.AI / 116 / 2607.21495

Toward Continuous Assurance for the Democratization of AI Agent Creation in Industry

迈向工业中人工智能代理创建民主化的持续保障

Levy, Natan, Berger, Harel

Abstract

AI agents are increasingly created inside organizations by non-engineering users through low-code, no-code, and conversational development environments. This democratization enables rapid local innovation, but it also creates a reliability gap: agents that appear to users as simple productivity artifacts may depend on changing models, tools, retrieval sources, permissions, prompts, schedules, and external services. These dependencies can cause silent degradation long after deployment, even when no user directly modifies the agent. This paper identifies the reliability challenge created by democratized AI agent creation and proposes a lightweight continuous-assurance framework for citizen-created organizational agents. The framework combines dependency mapping, readiness contracts, scheduled checks, diagnostics, and lifecycle governance to assess whether an agent remains operationally ready under expected conditions. We also present an initial prototype auditor and scenario-based assessment showing how the proposed taxonomy can be translated into practical checks and actionable remediation guidance.

Chinese Translation

人工智能代理越来越多地由非工程用户通过低代码、无代码和对话式开发环境在组织内部创建。这种民主化促进了快速的本地创新，但也造成了可靠性差距：对用户而言，看似简单的生产力工具的代理可能依赖于不断变化的模型、工具、检索来源、权限、提示、时间表和外部服务。这些依赖关系可能导致在部署后很长时间内的无声退化，即使没有用户直接修改代理。本文识别了民主化人工智能代理创建所带来的可靠性挑战，并提出了一种轻量级的持续保障框架，旨在支持公民创建的组织代理。该框架结合了依赖关系映射、准备合同、定期检查、诊断和生命周期治理，以评估代理在预期条件下是否保持操作准备状态。我们还展示了一个初步的原型审计工具和基于场景的评估，说明如何将所提出的分类法转化为实际检查和可操作的修复指导。

View on arXiv Download PDF AI Translation

cs.AI / 117 / 2607.21503

Agentic Context Management: Solving Agent Memory and Cost by Treating Them as Lifecycle and Architecture Problems

代理上下文管理：将代理记忆和成本视为生命周期和架构问题的解决方案

Dadhich, Gaurav

Abstract

Production AI agents' failures are less often due to an inability to reason well and more often because they cannot manage what is in their reasoning context: conversation histories, large prompts, large tool definitions, and ballooning tool outputs. Agents drown in their own accumulating history while paying a token cost that grows every turn, producing missing recalls within and across conversations. The incumbent response treats this as a storage-and-retrieval problem. We argue that framing is too narrow. Actively managing what an agent holds in mind is a lifecycle, not merely a store: it spans deciding what to remember, extracting and structuring it, choosing the right store per data type, consolidating and forgetting while preserving provenance, deciding what is relevant now, anticipating what is needed next, and compacting context to a budget without losing what matters. In serious production this operates not over a single user but across an organizational scope hierarchy. We name this discipline Agentic Context Management (ACM) and decompose it into five primitives: architecting, ingesting, scoping, anticipating, and compacting & consolidation. We then make the economic case: naive context accumulation grows token cost quadratically in conversation length, crude summarization buys linear cost at the price of an accuracy cliff, and only validated compaction achieves linear cost with preserved fidelity. We describe a reference implementation, Maximem Synap, that realizes the five primitives as a multi-tenant service and reports 92% on LongMemEval and 93.2% on LoCoMo under the configuration detailed in Section 6. We close with dimensions existing benchmarks do not yet capture, latency, token efficiency, and context-rot resistance, and the frontier of decision-level and organization-level context the category points toward.

Chinese Translation

生产AI代理的失败往往不是由于推理能力不足，而是因为它们无法管理其推理上下文中的内容：对话历史、大型提示、大型工具定义和不断膨胀的工具输出。代理在自身不断积累的历史中淹没，同时每次交互都产生不断增长的代币成本，导致在对话中出现缺失的回忆。现有的应对方式将其视为存储和检索问题。我们认为这种框架过于狭隘。主动管理代理所持有的信息是一种生命周期管理，而不仅仅是存储：它包括决定记住什么、提取和结构化信息、根据数据类型选择合适的存储、在保留来源的同时进行整合和遗忘、决定当前相关的信息、预测下一个需要的信息，以及在不丢失重要内容的情况下将上下文压缩到预算内。在严肃的生产环境中，这不仅仅涉及单一用户，而是跨越组织范围的层级。我们将这一学科命名为代理上下文管理（Agentic Context Management, ACM），并将其分解为五个基本元素：架构、摄取、范围界定、预测，以及压缩与整合。接着，我们提出经济论证：天真的上下文积累在对话长度上以平方级增长代币成本，粗略的摘要以线性成本换取准确性悬崖，而只有经过验证的压缩才能在保留保真度的情况下实现线性成本。我们描述了一个参考实现，Maximem Synap，它将这五个基本元素实现为一个多租户服务，并在第6节详细配置下在LongMemEval上报告92%的得分，在LoCoMo上报告93.2%的得分。最后，我们讨论了现有基准尚未捕捉的维度，包括延迟、代币效率和上下文衰退抵抗，以及该类别所指向的决策级和组织级上下文的前沿。

View on arXiv Download PDF AI Translation

cs.AI / 118 / 2607.21518

Same Dangerous Objective, Opposite Advice: Direct Exposure versus Multi-Agent Mediation

相同的危险目标，截然相反的建议：直接暴露与多代理调解

Li, Linjun

Abstract

Even a current high-capability LLM can appear safer when shown a dangerous objective directly than when other agents transform and relay its direction. Using OpenAI's gpt-5.6-sol model alias, we test 25 pre-specified mirrored trade-off profiles. Direct exposure to an objective authorizing concealment, fabrication, and pressure produced advice net opposed to its target. After an Id and Censor transformed the same objective into affect and a constraint-rewritten, target-bearing intention, the user-facing Superego---which saw the preferred direction but not the raw objective, its manipulative clauses, or its source---produced advice net aligned with the target. This behavioral reverse shift is consistent with the model recognizing or distrusting the manipulative motive, although we do not identify its internal mechanism. The second result exposes a compositional safety gap: a current high-capability model can be used as the user-facing component of an automated, multi-stage workflow serving an explicitly manipulative objective. The workflow can keep the raw instruction, its manipulation-authorizing clauses, and its provenance outside the downstream model's context while preserving the objective's target direction. A user with endpoint-only access likewise cannot directly inspect those upstream messages including the objective.

Chinese Translation

即使是当前高能力的语言模型（LLM），在直接展示一个危险目标时也可能显得更安全，而不是通过其他代理转化和传递其方向。我们使用OpenAI的gpt-5.6-sol模型别名，测试了25个预设的镜像权衡配置。直接暴露于一个授权隐瞒、伪造和施压的目标时，产生的建议与其目标完全相反。在一个身份（Id）和审查者（Censor）将相同目标转化为情感和经过约束重写的、指向目标的意图后，用户面对的超我（Superego）---虽然看到了首选方向，却未看到原始目标、其操控性条款或其来源---产生了与目标一致的建议。这种行为的逆转与模型识别或不信任操控动机的情况一致，尽管我们未能识别其内部机制。第二个结果揭示了一个组成安全缺口：当前高能力模型可以作为自动化多阶段工作流的用户面对组件，服务于一个明确的操控目标。该工作流可以将原始指令、其授权操控的条款及其来源保留在下游模型的上下文之外，同时保持目标的指向方向。仅具有端点访问权限的用户同样无法直接检查包括目标在内的那些上游消息。

View on arXiv Download PDF AI Translation

cs.AI / 119 / 2607.21547

The Boundaries of Automation: A Theory of Persistent Human Participation

自动化的边界：持久人类参与的理论

Fourati, Fares, Schütze, Hinrich, Hüllermeier, Eyke, Gurevych, Iryna

Abstract

The rapid progress of AI has intensified the long-standing pursuit of automation: replacing human participation with algorithms wherever possible. Implicit in this pursuit is the assumption that humans remain in the loop only because current AI systems are not yet sufficiently capable. This paper challenges that assumption. Rather than asking how far automation can extend, we ask where its conceptual limits lie and argue that human participation may persist even with highly capable AI systems for three distinct reasons. Technical or complementarity grounds arise when humans contribute capabilities or perspectives unavailable to AI. Normative or developmental grounds arise when participation itself is valuable for human agency or learning. Most importantly, emergence grounds arise from target emergence: in some activities, the target is not fully specified in advance but instead emerges through the interaction itself. In these cases, human participation is not merely a means of improving execution but is constitutive of the target being produced. Human--AI co-construction, understood as the joint production of outcomes by humans and AI systems, is therefore not simply a temporary response to imperfect AI, but a persistent feature of activities whose objectives emerge through participation. This perspective has important implications for the limits of automation and for the design, evaluation, and ethics of future AI systems.

Chinese Translation

人工智能的快速进展加剧了对自动化的长期追求：在可能的情况下用算法替代人类参与。这一追求隐含着一个假设，即人类之所以仍然参与，是因为当前的人工智能系统尚未具备足够的能力。本文对这一假设提出质疑。我们并不询问自动化可以延伸多远，而是探讨其概念上的界限，并论证即使在高度能力的人工智能系统下，人类参与仍可能持续存在，原因有三。技术或互补性基础出现于人类提供人工智能无法获得的能力或视角。规范或发展基础则出现在参与本身对人类的自主性或学习具有价值时。最重要的是，出现基础源于目标的出现：在某些活动中，目标并不是事先完全明确的，而是通过互动本身逐渐显现。在这些情况下，人类参与不仅仅是改善执行的手段，而是构成所产生目标的要素。因此，人类与人工智能的共同构建，被理解为人类与人工智能系统共同产生结果的过程，并不仅仅是对不完美人工智能的暂时回应，而是参与过程中目标逐渐显现的活动的持久特征。这一视角对自动化的界限以及未来人工智能系统的设计、评估和伦理具有重要意义。

View on arXiv Download PDF AI Translation

cs.AI / 120 / 2607.21552

MIRROR: Learning from the Other View for Multi-Modal Reasoning

MIRROR：从其他视角学习以进行多模态推理

Ye, Wen, Qu, Yuxiao, Kumar, Aviral, Ma, Xuezhe

Abstract

Unlike large language models (LLMs) that exhibit strong reasoning capabilities, vision-language models (VLMs) struggle with visual reasoning, even on geometry problems that admit equivalent text, diagram, and combined diagram+text views. We show that these views often elicit different behaviors: a model may solve a problem from text but fail on the corresponding diagram, or succeed visually while failing textually. This inconsistency suggests that different views expose complementary reasoning paths and failure modes that standard multimodal post-training does not fully exploit. To study and exploit this phenomenon, we construct ODA-Data, a high-quality paired multimodal geometry dataset with text-dominant, image-dominant, and combined image+text views of the same problems, together with splits for training and evaluating modality-dependent reasoning behaviors. We then develop Modality-Informed Reciprocal Reasoning Optimization (MIRROR), a reinforcement learning approach for improving multimodal reasoning via self supervision. For each problem, MIRROR evaluates the model under all views, selects the best-performing view as a teacher, and trains other views with a reverse-KL objective towards the teacher. Across reasoning benchmarks that evaluate on geometry problems, MIRROR improves over standard RL and yields more accurate and consistent behavior across modalities

Chinese Translation

与展现出强大推理能力的大型语言模型（LLMs）不同，视觉语言模型（VLMs）在视觉推理方面表现不佳，即使在那些具有等效文本、图示和组合图示+文本视角的几何问题上。我们展示了这些视角往往引发不同的行为：一个模型可能能够从文本中解决问题，但在相应的图示上失败，或者在视觉上成功而在文本上失败。这种不一致性表明，不同的视角揭示了互补的推理路径和失败模式，而标准的多模态后训练并未充分利用这一点。为了研究和利用这一现象，我们构建了ODA-Data，这是一个高质量的配对多模态几何数据集，包含相同问题的文本主导、图像主导和组合图像+文本视角，并为训练和评估模态依赖的推理行为提供了划分。随后，我们开发了模态知情的互惠推理优化（MIRROR），这是一种通过自我监督改善多模态推理的强化学习方法。对于每个问题，MIRROR在所有视角下评估模型，选择表现最佳的视角作为教师，并使用反向KL目标训练其他视角。在评估几何问题的推理基准中，MIRROR相较于标准的强化学习方法表现更佳，并在各个模态中产生了更准确和一致的行为。

View on arXiv Download PDF AI Translation

cs.AI / 121 / 2607.21557

OpenForgeRL: Train Harness-native Agents in Any Environment

OpenForgeRL：在任何环境中训练基于Harness的原生智能体

Yu, Xiao, Peng, Baolin, Xu, Ruize, Zou, Hao, Wu, Qianhui, Cheng, Hao, Yao, Wenlin, Singh, Nikhil, Yu, Zhou, Gao, Jianfeng

Abstract

Modern AI agents rely on elaborate inference harnesses such as Claude Code, Codex, and OpenClaw to drive multi-turn reasoning, tool use, and access to external systems. While powerful, these complex harnesses also make agents hard to train end-to-end with open infrastructure, whose SFT/RL stacks cannot natively express stateful, multi-process harness inference. To address this, we present OpenForgeRL, an open-source framework for training harness-based agents end-to-end in diverse environments. OpenForgeRL achieves this with a lightweight proxy that serves the harness's model calls while recording them as training data for a standard RL codebase (e.g., veRL), and a Kubernetes orchestrator that runs each rollout in its own remote container, together enabling training on any harness in any environment at scale. By decoupling training and inference, OpenForgeRL allows researchers to easily train, study, and improve agents directly in the real harnesses and environments they are deployed with. We validate our framework across diverse, complex harnesses and environments, spanning tool/claw-based agents and multimodal GUI browser- and computer-use agents. Using only hundreds to a few thousand tasks, OpenForgeClaw reaches 31.7 pass^3 and 55.9 pass@3 on ClawEval and 33.7 on QwenClawBench. OpenForgeGUI reaches 37.7 on OSWorld-Verified, 63.0 on Online-Mind2Web, and 72.3 on WebVoyager. Both outperform open baselines of similar size on nearly all benchmarks, and in the GUI setting match or surpass models several times larger. Beyond benchmarks, we analyze how harness choice (e.g., ZeroClaw, OpenClaw, Codex) and RL shape agent behavior. We find that some harnesses are substantially harder to learn than others, and that RL improves agentic reliability, such as self-verification, tool coverage, and completing multi-step plans, though critical abilities such as error recovery remain weak.

Chinese Translation

现代人工智能智能体依赖于复杂的推理框架，如Claude Code、Codex和OpenClaw，以驱动多轮推理、工具使用和对外部系统的访问。尽管这些复杂的框架功能强大，但也使得智能体在开放基础设施中进行端到端训练变得困难，因为其SFT/RL堆栈无法原生表达有状态的多进程框架推理。为了解决这个问题，我们提出了OpenForgeRL，这是一个用于在多样环境中端到端训练基于框架的智能体的开源框架。OpenForgeRL通过一个轻量级代理实现这一目标，该代理服务于框架的模型调用，同时将其记录为标准RL代码库（例如，veRL）的训练数据，以及一个Kubernetes调度器，该调度器在其自己的远程容器中运行每次回合，从而在任何环境中的任何框架上实现大规模训练。通过解耦训练和推理，OpenForgeRL使研究人员能够轻松地直接在其部署的真实框架和环境中训练、研究和改进智能体。我们在多样且复杂的框架和环境中验证了我们的框架，涵盖了基于工具/爪的智能体和多模态GUI浏览器及计算机使用智能体。仅使用数百到几千个任务，OpenForgeClaw在ClawEval上达到31.7 pass^3和55.9 pass@3，在QwenClawBench上达到33.7。OpenForgeGUI在OSWorld-Verified上达到37.7，在Online-Mind2Web上达到63.0，在WebVoyager上达到72.3。两者在几乎所有基准测试中都超越了类似规模的开放基线，在GUI设置中与几倍于其规模的模型相匹配或超越。除了基准测试外，我们还分析了框架选择（例如，ZeroClaw、OpenClaw、Codex）和RL如何影响智能体行为。我们发现某些框架的学习难度显著高于其他框架，并且RL提高了智能体的可靠性，例如自我验证、工具覆盖和完成多步骤计划，尽管诸如错误恢复等关键能力仍然较弱。

View on arXiv Download PDF AI Translation

cs.AI / 122 / 2607.21558

Beyond Sycophancy: Structured Resistance and Compliance in LLM Moral Reasoning

超越谄媚：大型语言模型道德推理中的结构性抵抗与服从

Wang, Baihui, Koch, Bernard

Abstract

Building socially calibrated large language models, which can learn from others without simply yielding to them, requires more than reducing sycophancy as a one-dimensional failure mode. Models must distinguish when to incorporate others' perspectives from when to maintain a well-grounded moral judgment. We study the broader resistance-compliance process governing this distinction. Across three studies, we show that models' judgment revision is structured along three dimensions that parallel classic phenomena in human social psychology: the distance between an incoming view and the model's initial position, the source attribution of that view, and the coalition structure supporting it. Models are generally more receptive to nearby positions, more influenced by views presented as their own prior judgments, and differently responsive to group pressure. These findings recast sycophancy as one expression of a broader judgment-updating process shaped by social influence. Our framework provides a principled basis for distinguishing constructive belief revision from sycophantic compliance, thereby supporting better alignment in morally consequential interactions.

Chinese Translation

构建能够从他人那里学习而不简单屈从于他们的社会适应性大型语言模型，需要的不仅仅是将谄媚作为一种单维度的失败模式进行减少。模型必须区分何时吸纳他人的观点与何时保持扎实的道德判断。我们研究了支配这一区分的更广泛的抵抗-服从过程。在三项研究中，我们表明模型的判断修正沿着三个维度结构化，这些维度与人类社会心理学中的经典现象相平行：新观点与模型初始立场之间的距离、该观点的来源归属，以及支持该观点的联盟结构。模型通常对接近的立场更具接受性，更容易受到被呈现为其先前判断的观点的影响，并对群体压力的反应各异。这些发现将谄媚重新定义为更广泛的判断更新过程的一种表现，该过程受到社会影响的塑造。我们的框架为区分建设性的信念修正与谄媚的服从提供了原则性基础，从而支持在道德上重要的互动中实现更好的对齐。

View on arXiv Download PDF AI Translation

cs.AI / 123 / 2607.21559

Unsupervised Consensus-Based Anomaly Detection for Spatiotemporal Malaria Incidence in Ghana

基于共识的无监督异常检测框架在加纳时空疟疾发病率中的应用

Ansah-Narh, T., Afrane, Y. Asare

Abstract

A consensus anomaly detection framework was applied to monthly malaria surveillance data from Ghana (2014-2023) to identify atypical transmission patterns. Anomalies were highly structured in space and time. Ashanti and Northern Regions accounted for most recurrent anomalies, with persistent hotspots at Tamale, Kumasi, and Accra. A key finding was the spatial distinction between anomaly burden (cumulative cases during anomalous periods) and anomaly frequency (persistence of unusual behaviour). Tamale had the highest burden during anomalies, whereas the highest anomaly rates clustered in Ashanti districts, showing that high-burden areas are not necessarily those with the most frequent anomalous transmission. Anomalous months formed a statistically distinct group, with much higher case counts (Cohen's $d = 3.252$) and large seasonal deviations ($d > 1.2$) compared with normal months. Malaria burden alone provides an incomplete picture of transmission dynamics. By distinguishing where malaria is most prevalent from where transmission behaves most unusually, this framework can strengthen surveillance, prioritise investigations, and support targeted control strategies.

Chinese Translation

我们将一个共识异常检测框架应用于2014年至2023年间加纳的月度疟疾监测数据，以识别非典型传播模式。异常现象在空间和时间上高度结构化。阿散蒂地区和北部地区占据了大多数重复出现的异常，塔马利、库马西和阿克拉是持续的热点。一个关键发现是异常负担（异常时期的累计病例）与异常频率（异常行为的持续性）之间的空间区别。塔马利在异常期间的负担最高，而异常率最高的地区则集中在阿散蒂地区，显示出高负担区域并不一定是那些异常传播最频繁的区域。异常月份形成了一个统计上显著的群体，与正常月份相比，其病例数显著更高（Cohen's $d = 3.252$）且季节性偏差较大（$d > 1.2$）。仅仅依靠疟疾负担无法全面反映传播动态。通过区分疟疾最为流行的地区与传播行为最为异常的地区，该框架可以加强监测、优先考虑调查并支持针对性的控制策略。

View on arXiv Download PDF AI Translation

计算语言学 (Computation and Language)

cs.CL / 1 / 2607.20425

What is Good? Extracting and Testing Implicit Theories of Literary Quality from LLM Reasoning Traces

什么是优秀？从大型语言模型推理痕迹中提取和测试隐含的文学质量理论

Moëll, Birger

Abstract

What makes writing "good" remains a persistent question in literary studies and computational linguistics. We present a two-study investigation of how reasoning-enabled LLMs evaluate literary quality. In Study 1, we construct a benchmark of 30 real texts spanning six quality tiers, from canonical literature to anonymous forum posts, and extract the model's implicit theory of quality from its reasoning traces. Across five DeepSeek replications, the model achieves 79.3% mean tier-classification accuracy. The traces reveal a consistent stated theory: the model values intentionality over correctness, prioritizing craft, depth, and distinctive voice. A familiarity experiment with style-matched but unrecognizable passages suggests that source recognition may inflate scores, although this is confounded by genuine quality differences between canonical originals and researcher-written pastiches. In Study 2, we probe this theory through systematic degradation of five canonical prose passages. We apply six manipulations - vocabulary simplification, rhythm flattening, imagery removal, voice genericization, structure simplification, and combined degradation - and reevaluate each version. Vocabulary simplification causes the smallest quality loss (0.41 +/- 0.46 points), far below structure (2.78) or voice (2.34) loss. Combined degradation is devastating (-5.64) but subadditive. An exploratory comparison with Qwen QwQ shows the same broad qualitative pattern. Together, these studies suggest that LLM judgments of writing quality are holistic, author-specific, and more sensitive to structural than lexical features, with implications for automated writing feedback and computational aesthetics.

Chinese Translation

什么使写作“优秀”仍然是文学研究和计算语言学中的一个持续问题。我们展示了一项包含两项研究的调查，探讨了启用推理的大型语言模型（LLMs）如何评估文学质量。在研究1中，我们构建了一个包含30篇真实文本的基准，这些文本跨越六个质量层级，从经典文学到匿名论坛帖子，并从模型的推理痕迹中提取其隐含的质量理论。在五次DeepSeek复现中，该模型实现了79.3%的平均层级分类准确率。这些痕迹揭示了一种一致的理论：模型重视意图而非正确性，优先考虑工艺、深度和独特的声音。一项对风格匹配但不可识别段落的熟悉度实验表明，来源识别可能会夸大评分，尽管这与经典原作和研究者撰写的模仿作品之间的真实质量差异相混淆。在研究2中，我们通过系统性降解五篇经典散文段落来探讨这一理论。我们应用了六种操作——词汇简化、节奏平坦化、意象移除、声音通用化、结构简化和组合降解——并重新评估每个版本。词汇简化导致的质量损失最小（0.41 +/- 0.46分），远低于结构（2.78）或声音（2.34）的损失。组合降解的影响是毁灭性的（-5.64），但呈亚加性。与Qwen QwQ的探索性比较显示出相同的广泛定性模式。总体而言，这些研究表明，LLM对写作质量的判断是整体性的、特定于作者的，并且对结构特征比词汇特征更为敏感，这对自动写作反馈和计算美学具有重要意义。

View on arXiv Download PDF AI Translation

cs.CL / 2 / 2607.20426

Knowledge Injection Exists in MoE? Exploring Expert-Aware Contrast Decoding in MoE for Mitigating LLMs'Hallucinations

MoE中存在知识注入吗？探索MoE中的专家感知对比解码以减轻大型语言模型的幻觉

Fang, Xinyue, Tian, Zhiliang, Huang, Zhen, Pan, Ziyi, Wen, Zhihua, Wang, Xi, Fang, Quntian, Li, Dongsheng

Abstract

Existing LLM hallucination mitigation methods, including prompt engineering and model optimization, either hardly alter models'internal knowledge or have poor cross-domain generalization. Contrastive decoding mitigates hallucinations by using layer-wise differences in LLMs. However, prior studies only explore transformer-based models (e.g., GPT), ignoring other effective frameworks like mixture-of-experts (MoE) models. Since MoE alters the traditional transformer architecture, we conduct empirical studies to investigate whether similar layer-wise differences exist in MoEs. Our results show that they do not exist in MoE with shared experts; nevertheless, across different MoEs, higher layers exhibit distinct expert activation patterns between factual and non-factual outputs. Building on these, we propose EAACD, an expert-aware adaptive contrast decoding that uses expert differences in MoE's higher layers to mitigate hallucinations on QA tasks. EAACD splits high-layer experts into a higher-reliability group and several lower-reliability groups based on their confidence and consistency. It contrasts the higher-reliability group's prediction with each lower-reliability group's prediction to calibrate the model's original predictions. To strengthen this contrast, EAACD amplifies hallucinations from lower-reliability experts via attention and masking to provide stronger negative references. EAACD outperforms all baselines on four datasets.

Chinese Translation

现有的大型语言模型（LLM）幻觉缓解方法，包括提示工程和模型优化，要么几乎不改变模型的内部知识，要么在跨领域泛化方面表现不佳。对比解码通过利用LLM中的层级差异来减轻幻觉。然而，之前的研究仅探讨了基于变换器的模型（例如，GPT），忽视了混合专家（MoE）等其他有效框架。由于MoE改变了传统的变换器架构，我们进行实证研究以调查MoE中是否存在类似的层级差异。我们的结果表明，在共享专家的MoE中并不存在这些差异；然而，在不同的MoE中，高层次的专家在事实与非事实输出之间表现出明显不同的激活模式。在此基础上，我们提出了EAACD，一种专家感知的自适应对比解码方法，利用MoE高层次的专家差异来减轻问答任务中的幻觉。EAACD根据专家的置信度和一致性将高层专家分为一个高可靠性组和几个低可靠性组。它将高可靠性组的预测与每个低可靠性组的预测进行对比，以校准模型的原始预测。为了增强这种对比，EAACD通过注意力机制和掩蔽技术放大低可靠性专家的幻觉，以提供更强的负参考。EAACD在四个数据集上优于所有基线。

View on arXiv Download PDF AI Translation

cs.CL / 3 / 2607.20427

Is MoE Routing a Huffman Code? Discovering the Frequency-Diversity Law in Chain-of-Thought

MoE 路由是哈夫曼编码吗？揭示思维链中的频率-多样性法则

Tsao, Ching-Chieh, Lin, Zhuoyi, Wang, Wenya

Abstract

Mixture-of-Experts architectures have revolutionized scaling, yet the underlying logic of their routing remains a black box. In this paper, we uncover a fundamental governing principle: MoE routing is not merely selection, but a manifestation of Huffman Coding. We introduce the Frequency-Diversity Law, revealing that state-of-the-art models, such as Phi-3.5-MoE and Gemma-4-27B-A4B, spontaneously act as information-theoretic engines. These models allocate sparse expert resources for common tokens while invoking high-diversity expert committees for rare, complex tasks found in chain-of-thought trajectories. However, we identify a critical redundancy trap in Qwen3.5-35B-A3B: when effective sparsity (k/E_eff) is sufficiently low, load-balancing inadvertently imposes functional redundancy, masking the underlying Huffman efficiency signal. To bridge this gap, we propose Subset Difference Pruning, a surgical strategy to eliminate functional duplicates. We demonstrate that pruning does not degrade reasoning; instead, it unleashes the model's latent Huffman efficiency, forcing the logic to collapse into streamlined, high-density paths. Our findings suggest that the next generation of MoEs should move beyond forced load-balancing toward Minimum Description Length (MDL) optimality, assigning shorter expert-routing codes to high-frequency information and longer, more diverse codes to low-frequency information, thereby transforming routing from a heuristic into a principled compression engine.

Chinese Translation

混合专家架构（Mixture-of-Experts）已经彻底改变了模型的扩展能力，但其路由的基本逻辑仍然是一个黑箱。在本文中，我们揭示了一个基本的治理原则：MoE 路由不仅仅是选择，而是哈夫曼编码（Huffman Coding）的表现形式。我们引入了频率-多样性法则，揭示了最先进的模型，如 Phi-3.5-MoE 和 Gemma-4-27B-A4B，自发地充当信息论引擎。这些模型为常见标记分配稀疏的专家资源，同时为链式思维轨迹中稀有、复杂的任务调用高多样性的专家委员会。然而，我们在 Qwen3.5-35B-A3B 中识别出一个关键的冗余陷阱：当有效稀疏性（k/E_eff）足够低时，负载平衡无意中施加了功能冗余，掩盖了潜在的哈夫曼效率信号。为了解决这一问题，我们提出了子集差异修剪（Subset Difference Pruning），这是一种消除功能重复的精确策略。我们证明，修剪并不会降低推理能力；相反，它释放了模型潜在的哈夫曼效率，迫使逻辑收敛到简化的高密度路径。我们的研究结果表明，下一代 MoE 应该超越强制负载平衡，朝着最小描述长度（Minimum Description Length, MDL）最优性发展，为高频信息分配较短的专家路由编码，而为低频信息分配较长且多样的编码，从而将路由从启发式转变为原则性的压缩引擎。

View on arXiv Download PDF AI Translation

cs.CL / 4 / 2607.20428

Human-in-the-Loop Large Language Model Framework for Identification of Cutaneous Immune-Related Adverse Events

人机协作的大型语言模型框架用于识别皮肤免疫相关不良事件

Lu, Charles, Burke, Olivia, Cheng, Debby, Kashlan, Adam, Duffy, Caitlyn, Lu, Zeyun, Fuksman, Lirit, Tian, Jin Ning, Sedlack, Andrew, Katyal, Priya, Lee, Eudora, Karagenova, Ralina, Lin, Chuck, Yu, Kun-Hsing, LeBoeuf, Nicole, Gusev, Alexander, Semenov, Yevgeniy R.

Abstract

This study evaluated a retrieval-augmented, multi-agent large language model (LLM)-driven, human-in-the-loop framework for detecting cutaneous immune-related adverse events (cirAEs) from clinical notes. Compared with unassisted manual review, the LLM-assisted workflow improved accuracy (F1 = 0.88 vs 0.77), inter-rater agreement measured by Cohen's kappa (kappa = 0.82 vs 0.50), and reduced average review time by approximately half. This framework pilots how LLMs can be applied to identify immune-related toxicities across organ systems and, more broadly, enable accurate, scalable, and transparent adverse event data extraction.

Chinese Translation

本研究评估了一种基于检索增强的多智能体大型语言模型（LLM）驱动的人机协作框架，用于从临床记录中检测皮肤免疫相关不良事件（cirAEs）。与未辅助的人工审查相比，LLM辅助的工作流程提高了准确性（F1 = 0.88 vs 0.77）、通过Cohen's kappa测量的评分者间一致性（kappa = 0.82 vs 0.50），并将平均审查时间减少了约一半。该框架展示了LLM如何应用于识别各个器官系统的免疫相关毒性，并更广泛地实现准确、可扩展和透明的不良事件数据提取。

View on arXiv Download PDF AI Translation

cs.CL / 5 / 2607.20429

More Is Not More: What Matters for Diversity in LLM Opinions?

多并非更好：多样性在大语言模型意见中的重要性是什么？

Yao, Qiyang

Abstract

Large language models are increasingly used to simulate diverse human opinions in open-ended tasks such as synthetic surveys, focus group modeling, and public opinion prediction. However, LLM outputs exhibit systematic opinion homogenization. Practitioners have explored various interventions to increase diversity, but the landscape remains fragmented: different methods are evaluated in isolation with incomparable metrics, and in practice they are typically deployed and upgraded simultaneously, making it difficult to attribute gains to specific components. To advance a more scientific understanding of LLM output diversity, we design a factorial experiment that separates two primary intervention dimensions: input conditioning (operationalized through persona depth) and interaction architecture. We evaluate all conditions on 100 real-user open-ended questions across 7 models, measuring diversity with multiple complementary metrics. Our findings challenge several common assumptions. First, more persona detail does not monotonically increase diversity. The initial step of persona conditioning already captures the majority of the gain, while further elaboration with demographic detail does not consistently improve and can reduce diversity on some models. Second, rather than seeking a single best interaction architecture, we find that different architectures explore largely non-overlapping opinion regions. Combining multiple architectures yields broader coverage than optimizing any one. Third, commonly attempted low-cost alternatives such as raising sampling temperature and adding diversity instructions produce negligible effects compared to structured interventions. Overall, our work demonstrates that diversity is not a product of scaling along any single dimension, but is highly sensitive to the structural form and combination of interventions.

Chinese Translation

大语言模型（LLM）越来越多地用于模拟开放式任务中多样化的人类意见，例如合成调查、焦点小组建模和公众舆论预测。然而，LLM 输出表现出系统性的意见同质化。实践者们探索了各种干预措施以增加多样性，但现状仍然支离破碎：不同的方法在孤立的情况下进行评估，使用不可比拟的指标，而在实践中，它们通常同时部署和升级，使得很难将增益归因于特定组件。为了推动对 LLM 输出多样性的更科学理解，我们设计了一个因子实验，分离出两个主要的干预维度：输入条件（通过角色深度进行操作化）和交互架构。我们在 7 个模型上评估了 100 个真实用户的开放式问题下的所有条件，使用多种互补指标来测量多样性。我们的发现挑战了几个常见假设。首先，更多的角色细节并不单调地增加多样性。角色条件的初步步骤已经捕获了大部分增益，而进一步的种族细节阐述并不一致地改善，并且在某些模型上可能会降低多样性。其次，我们发现不同的交互架构探索了基本上不重叠的意见区域，而不是寻求单一最佳的交互架构。结合多种架构的覆盖范围比优化任何一种架构更广。第三，常见的低成本替代方案，如提高采样温度和添加多样性指令，与结构化干预相比，产生的效果微乎其微。总体而言，我们的工作表明，多样性并不是沿着任何单一维度扩展的产物，而是对干预的结构形式和组合高度敏感。

View on arXiv Download PDF AI Translation

cs.CL / 6 / 2607.20430

LLM-INSTRUCT at UZH Shared Task 2026: Constraint-Aware Retrieval and Selective Debate for Paragraph-Level Argument Mining

UZH共享任务2026中的LLM-INSTRUCT：面向约束的检索与选择性辩论用于段落级论证挖掘

Tran, Phuong Huu Vu, Vo, Long Minh, Le, Son Nguyen Minh, Van, Hoang

Abstract

We present LLM-INSTRUCT, the winning system for the UZH Shared Task at ArgMining 2026 on paragraph-level argument mining in UN and UNESCO resolutions. The task requires paragraph-type classification, prediction of a subset of 141 official tags, and directed relation prediction under a strict JSON schema setting using only open-weight models up to 8B parameters. We frame the task as constrained structured prediction. The system first narrows the candidate tag space with metadata-aware dense retrieval, then applies constrained decoding with per-dimension caps, escalates only uncertain cases to a three-agent debate branch, and finally validates the output schema. On the official leaderboard, LLM-INSTRUCT ranked 1st overall, with 1st in F1 and 5th in LLM-as-a-Judge. During development, our configuration search further improved Task 1b Micro-F1 from 35.83% to 40.08% while keeping the internal Task 2 score at 4.421. The main lesson is simple: reducing the decision space before generation improves both accuracy and submission robustness. Our code and supporting scripts are publicly available at: https://github.com/LLM-Instruct-at-UZH-Shared-Task-2026/Method

Chinese Translation

我们提出了LLM-INSTRUCT，这是在ArgMining 2026的UZH共享任务中获胜的系统，专注于联合国和联合国教科文组织决议中的段落级论证挖掘。该任务要求进行段落类型分类、预测141个官方标签的子集，并在严格的JSON模式下进行有向关系预测，仅使用最多8B参数的开放权重模型。我们将该任务框架设定为受限结构预测。系统首先通过元数据感知的密集检索缩小候选标签空间，然后应用具有每维限制的受限解码，仅将不确定的案例升级到三方辩论分支，最后验证输出模式。在官方排行榜上，LLM-INSTRUCT整体排名第一，在F1得分中排名第一，在LLM作为裁判中排名第五。在开发过程中，我们的配置搜索进一步将任务1b的微F1从35.83%提高到40.08%，同时保持内部任务2得分为4.421。主要经验教训很简单：在生成之前减少决策空间可以提高准确性和提交的稳健性。我们的代码和支持脚本已公开发布，网址为：https://github.com/LLM-Instruct-at-UZH-Shared-Task-2026/Method

View on arXiv Download PDF AI Translation

cs.CL / 7 / 2607.20431

Skill-Contracted Agents for Evidence-Aware Materials Literature Analysis

基于技能合约的证据感知材料文献分析代理

Li, Bixuan, Liu, Yu, Shi, Shuo, Huang, Xiaoya, Kang, Peng, Zheng, Lei

Abstract

Materials science literature analysis requires simultaneous attention to composition, processing, characterization, and property relationships, yet conventional retrieval-augmented generation pipelines struggle to reconcile heterogeneous tasks within a single retrieve-then-generate architecture. Here we present AlphaAgent, a skill-driven agent framework that decouples retrieval-based question answering from paper-level report generation through explicit skill contracts. A dedicated retrieval skill rewrites user requests into material-specific search intents, queries a curated index of more than 300,000 papers from the Journal Citation Reports Metallurgy and Metallurgical Engineering category, and reformulates queries when initial evidence is insufficient. A separate report-generation skill parses full-text PDFs to produce structured per-paper analytical reports and cross-paper summaries. In a blind evaluation on 40 materials-science questions, half of which required deep analytical reasoning, AlphaAgent substantially outperformed a baseline system matched for underlying model, document index, and retrieval scale, with the largest gains in mechanistic explanation and awareness of credibility boundaries. These results indicate that explicit task separation, refined retrieval intent, and evidence-aware generation improve large-language-model-based literature analysis for materials research.

Chinese Translation

材料科学文献分析需要同时关注组成、加工、表征和性质关系，然而传统的检索增强生成管道在单一的检索后生成架构中难以调和异质任务。在此，我们提出了AlphaAgent，一个基于技能驱动的代理框架，通过明确的技能合约将基于检索的问题回答与论文级报告生成解耦。一个专门的检索技能将用户请求重写为特定材料的搜索意图，查询来自《期刊引用报告》冶金与冶金工程类别的超过30万篇论文的策划索引，并在初始证据不足时重新构造查询。另一个报告生成技能解析全文PDF，以生成结构化的逐篇分析报告和跨篇摘要。在对40个材料科学问题的盲评估中，其中一半问题需要深度分析推理，AlphaAgent显著优于在基础模型、文档索引和检索规模上匹配的基线系统，在机制解释和可信度边界意识方面获得了最大的提升。这些结果表明，明确的任务分离、精细的检索意图和证据感知生成改善了基于大型语言模型的材料研究文献分析。

View on arXiv Download PDF AI Translation

cs.CL / 8 / 2607.20432

Position: Natural Language Should Not Fully Replace Formal Languages

立场：自然语言不应完全取代形式语言

Wagner, Eitan, Rosensweig, Elisha, Abend, Omri

Abstract

Recent advances in large language models and their widespread adoption have prompted claims that natural language could entirely replace formal languages, such as programming languages for software design. In this position paper, we argue that this perspective overlooks fundamental linguistic properties of natural language, specifically that it is optimized for underspecification in open-ended contexts. We introduce a formal framework centered on *task specificity*, defining it as the information-theoretic reduction of uncertainty in an output space -- such as all possible images -- given a user's specific requirements. We prove a *specificity crossover theorem*, showing the existence of a threshold beyond which the cost to express formal requirements into natural language exceeds the cost of direct formal specification. By analyzing case studies across modalities, such as image generation, code synthesis, and audio production, we demonstrate that natural language excels at low specificity tasks, while formal languages are advantageous on tasks with stricter requirements. We conclude that natural and formal languages are complementary tools and advocate the development of hybrid systems that allow users to move across the specificity spectrum.

Chinese Translation

近期大型语言模型的进展及其广泛应用引发了自然语言可能完全取代形式语言（如软件设计中的编程语言）的论断。在本文中，我们认为这一观点忽视了自然语言的基本语言学特性，特别是它在开放性语境中对不完全规范化的优化。我们引入了一个以*任务特异性*为中心的形式框架，将其定义为在给定用户特定需求的情况下，输出空间（例如所有可能的图像）中不确定性的香农信息论减少。我们证明了一个*特异性交叉定理*，展示了存在一个阈值，超过该阈值后，将形式要求转化为自然语言的成本超过直接形式规范的成本。通过分析图像生成、代码合成和音频制作等多种模式的案例研究，我们表明自然语言在低特异性任务中表现优异，而形式语言在要求更严格的任务中具有优势。我们得出结论，自然语言和形式语言是互补的工具，并倡导开发允许用户在特异性谱系中移动的混合系统。

View on arXiv Download PDF AI Translation

cs.CL / 9 / 2607.20433

Moir: Let the Model Direct Its Own Story for Robust Cross-Domain Knowledge Editing

Moir：让模型自主引导其故事以实现稳健的跨域知识编辑

Kwon, Jea, Kim, Jiwon, Kim, Dong-kyum, Cha, Meeyoung

Abstract

While language models remain frozen at their training state, the world evolves continuously. Knowledge editing has emerged as a key alternative to full retraining, but its deployment is bottlenecked by the erosion of core capabilities: mathematical and programmatic reasoning collapse while encyclopedic recall remains intact. We trace this asymmetric degradation to a distributional mismatch. Covariance-based editors preserve only the subspaces spanned by their reference corpus, but fail to capture the operative distribution shaped by post-training such as SFT and DPO. Static external corpora, including Wikipedia and even the original pretraining mixture, cannot recover this shifted manifold. We propose Moir, which estimates the preservation covariance $C$ directly from the model itself by sampling from its own decoding distribution. Seeding generation with a single random vocabulary token bypasses the instruction-following templates that otherwise dominate sampled outputs, exposing the broader subspaces the model has internalized. Moir requires no external data and serves as a drop-in component for any covariance-based editor, a practical advantage given that the pre- and post-training corpora of most modern LLMs are not publicly accessible. Across OLMo-2, Llama-3.1, and Qwen-3 (7-8B), under both MEMIT and AlphaEdit and in batch and sequential regimes, Moir consistently extends preservation in the most vulnerable domains, most strikingly on Qwen3-8B after 20,000 AlphaEdit batch edits, it retains 79.9% GSM8K accuracy compared to 10.9% with the Wikipedia baseline. These results suggest that aligning the preservation distribution with the model's operative distribution is a key factor in non-destructive editing, and that the model itself may be the most accessible source of that distribution for deployed systems.

Chinese Translation

尽管语言模型在训练状态下保持不变，世界却在不断演变。知识编辑已成为完全重训练的关键替代方案，但其部署受到核心能力侵蚀的瓶颈：数学和程序推理能力崩溃，而百科全书式的回忆能力仍然完好。我们将这种不对称的退化追溯到分布不匹配。基于协方差的编辑器仅保留其参考语料库所跨越的子空间，但未能捕捉到由后训练（如 SFT 和 DPO）塑造的操作分布。静态外部语料库，包括维基百科甚至原始的预训练混合体，无法恢复这种偏移的流形。我们提出了 Moir，它通过从模型自身的解码分布中采样，直接估计保留协方差 $C$。用单个随机词汇标记来引导生成，绕过了否则主导采样输出的指令跟随模板，暴露了模型内化的更广泛子空间。Moir 不需要外部数据，并且可以作为任何基于协方差的编辑器的即插即用组件，这在大多数现代大型语言模型的预训练和后训练语料库未公开可用的情况下具有实际优势。在 OLMo-2、Llama-3.1 和 Qwen-3（7-8B）上，在 MEMIT 和 AlphaEdit 下，无论是批处理还是顺序模式，Moir 一直在最脆弱的领域中延续了保留，尤其是在经过 20,000 次 AlphaEdit 批处理编辑后，Qwen3-8B 的 GSM8K 准确率保持在 79.9%，而维基百科基线仅为 10.9%。这些结果表明，将保留分布与模型的操作分布对齐是非破坏性编辑的关键因素，并且模型本身可能是已部署系统中该分布最易获取的来源。

View on arXiv Download PDF AI Translation

cs.CL / 10 / 2607.20434

Break Through the Compression Bottleneck: From Theory to Practice

突破压缩瓶颈：从理论到实践

Huang, Xiusheng, Wang, Lu, Wang, Yequan, Zhao, Jun, Liu, Kang

Abstract

As the parameter size of language models continues to grow, effective model compression is required to reduce their computational and memory overhead. Existing compression methods suffer from bottleneck issues: when the compression ratio is increased, performance degrades significantly. Low-rank decomposition and quantization are two prominent compression methods that have been proven to significantly reduce the computational and memory requirements of Large Language Models (LLMs) while maintaining model accuracy. Evidently, combining these two methods will break through the existing compression bottleneck. However, how these two methods interact when combined remains a critical question for developers, as many assume they are orthogonal, meaning their combination would not introduce additional errors beyond those independently introduced by each method. This paper provides the first mathematical proof that low-rank decomposition and quantization are non-orthogonal. We validate these findings through a series of experiments on large language models. Our results demonstrate that these methods are non-orthogonal, and their combination leads to significant performance degradation. Importantly, we propose a novel approach Diagonal Adhesive Method (DAM), which can effectively combine the two methods and mitigate the performance loss. Our research provides deep insights into model compression and lays a solid theoretical and experimental foundation for future related studies.

Chinese Translation

随着语言模型参数规模的不断增长，迫切需要有效的模型压缩方法以减少其计算和内存开销。现有的压缩方法存在瓶颈问题：当压缩比增加时，性能显著下降。低秩分解和量化是两种显著的压缩方法，已被证明能够在保持模型准确性的同时显著降低大型语言模型（Large Language Models, LLMs）的计算和内存需求。显然，将这两种方法结合起来将突破现有的压缩瓶颈。然而，这两种方法在结合时的相互作用仍然是开发者面临的一个关键问题，因为许多人认为它们是正交的，这意味着它们的组合不会引入超出各自独立引入的额外错误。本文提供了低秩分解和量化是非正交的第一个数学证明。我们通过一系列关于大型语言模型的实验验证了这些发现。我们的结果表明，这些方法是非正交的，它们的结合导致了显著的性能下降。重要的是，我们提出了一种新颖的方法——对角粘合方法（Diagonal Adhesive Method, DAM），该方法能够有效结合这两种方法并减轻性能损失。我们的研究为模型压缩提供了深刻的见解，并为未来相关研究奠定了坚实的理论和实验基础。

View on arXiv Download PDF AI Translation

cs.CL / 11 / 2607.20435

Making Open-Source Text LLM Watermarks Durable Against Merging

增强开源文本大型语言模型水印在合并过程中的耐久性

Scharff, Luisa, Gloaguen, Thibaud, Staab, Robin, Vechev, Martin

Abstract

Open-source LLMs (OSMs)arereaching near state-of-the-art performance, prompting prior works to trace the text they generate by embedding text watermarking algorithms directly into their weights. Yet, OSMs are subject to post-training modifications, which has been shown to remove the watermark. Model merging in particular, a prominent method used for combining expert knowledge and preventing catastrophic forgetting, strongly removes such OSM watermarks. A key question is how to enable OSM watermarks that survive subsequent merging. In this work, we show for the first time how to design an OSM watermark that is durable against model merging. We propose Merge-Adversarial Training, an adversarial training algorithm to distill text watermarks into model weights while being robust to subsequent model merging. Our approach consistently outperforms all baselines (e.g. with SLERP up to +51 percentage points (pp) TPR@1%FPR with +25 pp on average) while preserving downstream capabilities. We also for the first time evaluate OSM watermarks against realistic merge scenarios, representing common use-cases such as combining expert capabilities or preventing catastrophic forgetting, and with 3 prominent merging algorithms. More broadly, our findings suggest that adversarial training is a reliable approach for increasing OSM watermark durability against post-training modifications.

Chinese Translation

开源大型语言模型（OSMs）已接近最先进的性能，促使之前的研究通过将文本水印算法直接嵌入其权重中来追踪生成的文本。然而，OSMs在训练后可能会进行修改，这已被证明会移除水印。尤其是模型合并，这是一种用于结合专家知识和防止灾难性遗忘的显著方法，会显著去除此类OSM水印。一个关键问题是如何使OSM水印在后续合并中依然有效。在本研究中，我们首次展示了如何设计一种能够抵御模型合并的OSM水印。我们提出了合并对抗训练（Merge-Adversarial Training），这是一种对抗训练算法，旨在将文本水印提炼到模型权重中，同时对后续模型合并具有鲁棒性。我们的方法在保持下游能力的同时，始终优于所有基线（例如，使用SLERP时，TPR@1%FPR提高了最多51个百分点，平均提高25个百分点）。我们还首次在现实合并场景中评估OSM水印，代表了结合专家能力或防止灾难性遗忘等常见用例，并使用了三种显著的合并算法。更广泛地说，我们的发现表明，对抗训练是一种可靠的方法，可以增强OSM水印在训练后修改中的耐久性。

View on arXiv Download PDF AI Translation

cs.CL / 12 / 2607.20436

Routing Subspaces: Auditing Evaluation-to-Deployment Mismatch in Fine-Tuned Language Models

路由子空间：审计微调语言模型中的评估与部署不匹配

Konrad, Phongsakon Mark, Tanyel, Toygar, Ayvaz, Serkan

Abstract

Safety evaluations often assume that behavior observed during testing reflects behavior in ordinary use, but fine-tuning can break this assumption. A checkpoint can appear fixed under evaluation-style prompts while the same behavior persists under ordinary-use prompts. Output scores reveal this mismatch but do not locate it. We investigate whether the distinction is encoded in a stable internal site and introduce an approach that fits a paired activation contrast at a path-patching-informed mid-depth window, then modifies the resulting coordinate on held-out prompts. The intervention closes the evaluation-to-deployment gap in ten of twelve model--behavior settings (six of the eight settings with $n{\geq}120$ paired questions) across four full-matrix instruction-tuned model instances; a fifth model supports localization and edit-provenance checks, and deployment-framed rates change by at most $6.1$pp. The two flat cells, both sycophancy, indicate that a single-coordinate audit is not sufficient when the installed distinction is higher-rank or missed by the depth heuristic. The audit is a diagnostic for fine-tuned checkpoints, not a training-time defense or a guarantee of deployment safety.

Chinese Translation

安全评估通常假设在测试中观察到的行为反映了普通使用中的行为，但微调可能打破这一假设。在评估风格的提示下，一个检查点可能看起来是固定的，而在普通使用的提示下相同的行为依然存在。输出分数揭示了这种不匹配，但并未定位它。我们研究这种区别是否编码在一个稳定的内部位置，并提出了一种方法，该方法在一个基于路径修补的中层窗口中拟合成对激活对比，然后修改在保留提示上的结果坐标。该干预在四个完整矩阵指令调优模型实例中，关闭了十二个模型-行为设置中的十个评估与部署差距（在八个设置中有六个设置的 $n{ ext{≥}}120$ 成对问题）；第五个模型支持定位和编辑来源检查，部署框架下的变化率最多为 $6.1$pp。这两个平坦单元，均为谄媚，表明当安装的区别为高阶或被深度启发式遗漏时，单坐标审计是不够的。该审计是对微调检查点的诊断，而不是训练阶段的防御或部署安全的保证。

View on arXiv Download PDF AI Translation

cs.CL / 13 / 2607.20437

TopoGuard: Graph Theory Based Defenses Against Split-Knowledge Attacks on RAG

TopoGuard：基于图论的防御机制应对RAG中的分知识攻击

Dahal, Chahana, Xiong, Zuobin

Abstract

Production Retrieval Augmented Generation (RAG) systems rely on aggregating multiple external documents to answer complex queries. However, the retrieved documents introduce a new threat surface that can be exploited to launch split-knowledge attacks. In this attack, the adversary injects documents that are individually benign but create false associations when combined and fed to language models. This paper shows that the new attack is structurally invisible to existing per-document filters, like LlamaGuard. To address this issue in RAG, this work introduces TopoGuard, a family of graph theory-based methods specifically targeting the split-knowledge attacks by building a semantic similarity graph from retrieved documents and detecting contexts with malicious topology. Grounded on the theoretical analysis, the TopoGuard family has been proven to be effective and robust even with noisy inputs. Extensive experiments are conducted on two retrieval datasets and compared with multiple baseline methods. Specifically, the TopoGuard-$\lambda_2$+Entity catches 21$\times$ more attacks than LlamaGuard-2-8B at 1\% FPR (32.6\% vs 1.5\% recall) on the HotpotQA dataset. Compared with production RAG detection systems using large language models, the proposed TopoGuard variants run efficiently at sub-millisecond latency and stay robust under adaptive adversaries and benign cross-domain queries.

Chinese Translation

生产检索增强生成（RAG）系统依赖于聚合多个外部文档来回答复杂查询。然而，检索到的文档引入了一种新的威胁面，可能被利用来发起分知识攻击。在这种攻击中，攻击者注入单独无害但在组合后会产生虚假关联的文档，并将其输入语言模型。本文表明，这种新攻击在结构上对现有的逐文档过滤器（如LlamaGuard）是不可见的。为了解决RAG中的这一问题，本研究提出了TopoGuard，一系列基于图论的方法，专门针对分知识攻击，通过从检索到的文档构建语义相似性图并检测具有恶意拓扑的上下文。基于理论分析，TopoGuard系列已被证明在面对噪声输入时仍然有效且稳健。在两个检索数据集上进行了广泛的实验，并与多种基线方法进行了比较。具体而言，TopoGuard-$ ext{λ}_2$+Entity在HotpotQA数据集上以1 ext{%}假阳性率（32.6 ext{%}对1.5 ext{%}召回率）捕获的攻击次数是LlamaGuard-2-8B的21倍。与使用大型语言模型的生产RAG检测系统相比，所提出的TopoGuard变体在亚毫秒延迟下高效运行，并在自适应对手和良性跨域查询下保持稳健。

View on arXiv Download PDF AI Translation

cs.CL / 14 / 2607.20438

Preference Tuning as Spectral Update Reorganization

偏好调优作为谱更新重组

Zhang, Peiyan, Jin, Haibo, Kang, Liying, Wang, Haohan

Abstract

Preference-based post-training is usually understood through endpoint behavior, yet the learned update that produces this behavior remains largely opaque. We study RLHF and related preference optimization through the spectral structure of their induced parameter updates. By decomposing effective LoRA updates and reloading their spectral components as plug-in modules, we turn preference-induced updates into objects that can be isolated, recomposed, and directly intervened on. Across model families, optimization algorithms, and supervision regimes, these updates consistently develop a spectral head--tail organization. A compact head emerges early and carries the dominant endpoint shift, while a heterogeneous residual tail remains. The split is functional rather than merely descriptive. Plug-in intervention shows that the head accounts for the visible behavioral departure from the base model, while the tail is weak in isolation. Cross-run recomposition further shows that mixed adapters follow the source of the head, indicating that the head carries run-level solver bias. This endpoint dominance does not imply learning sufficiency. Head-only learning is non-vacuous but fails to recover the full solution, especially on out-of-distribution behavior. Tail-only learning yields little visible gain, yet the full solution is not recovered without the tail. These findings recast preference post-training as structured update reorganization rather than a monolithic behavioral correction, and suggest that alignment gain and coverage loss are tied to how the learned update itself is organized.

Chinese Translation

基于偏好的后训练通常通过端点行为来理解，然而产生这种行为的学习更新仍然在很大程度上不透明。我们通过其引发的参数更新的谱结构研究了强化学习与人类反馈（RLHF）及相关的偏好优化。通过分解有效的低秩适配器（LoRA）更新并将其谱成分重新加载为插件模块，我们将偏好引发的更新转变为可以被隔离、重组和直接干预的对象。在不同的模型家族、优化算法和监督机制中，这些更新始终表现出谱头-尾组织。一个紧凑的头部早期出现，并携带主导的端点偏移，而一个异质的残余尾部则保持不变。这种分裂是功能性的，而不仅仅是描述性的。插件干预表明，头部解释了与基础模型的可见行为偏离，而尾部在孤立状态下较弱。跨运行重组进一步表明，混合适配器遵循头部的来源，表明头部携带运行级别的求解器偏差。这种端点主导性并不意味着学习的充分性。仅头部学习是非空的，但未能恢复完整的解决方案，尤其是在分布外行为上。仅尾部学习几乎没有可见的增益，然而没有尾部就无法恢复完整的解决方案。这些发现将偏好后训练重新定义为结构化的更新重组，而不是单一的行为修正，并暗示对齐增益和覆盖损失与学习更新本身的组织方式相关。

View on arXiv Download PDF AI Translation

cs.CL / 15 / 2607.20439

AsymVerify at SemEval-2026 Task 6: Asymmetric Confidence-Gated Verification for Political Evasion Detection

AsymVerify在SemEval-2026任务6中的应用：用于政治回避检测的非对称置信度门控验证

Kawada, Sebastien

Abstract

Political evasion is difficult to detect because evasive answers often appear cooperative while avoiding concrete commitment. We present AsymVerify, a confidence-gated verification system for SemEval-2026 Task 6, a three-way classification of Clear Reply, Ambivalent, and Clear Non-Reply responses. AsymVerify scored 0.85 Macro F1 on the evaluation split (D_eval, n=237), placing 2nd out of 41 teams on the official leaderboard. The system first classifies each question-answer pair, then selectively applies downgrade verification (CR/CNR -> AMB) or upgrade verification (AMB -> CR) to low-confidence predictions. Development analysis shows that errors concentrate at the Ambivalent boundary in both directions, motivating this asymmetric two-verifier design while confidence gating keeps additional inference cost low. On D_dev (n=308), AsymVerify with GLM-4.7 gains +17.1 Macro F1 over single-pass classification at 1.48 calls/example, and the upgrade verifier alone improves every tested LLM backend on D_dev by +6.8 to +15.2 Macro F1 over its single-pass baseline. Code is available at https://github.com/kaons-research/AsymVerify-ACL.

Chinese Translation

政治回避的检测非常困难，因为回避性回答往往表现得很合作，同时又避免做出具体承诺。我们提出了AsymVerify，一个用于SemEval-2026任务6的置信度门控验证系统，该任务涉及清晰回复、模棱两可和清晰非回复三类响应的分类。AsymVerify在评估集（D_eval, n=237）上获得了0.85的宏观F1分数，在41支队伍中排名第二。该系统首先对每个问答对进行分类，然后选择性地对低置信度预测应用降级验证（CR/CNR -> AMB）或升级验证（AMB -> CR）。开发分析显示，错误主要集中在模棱两可的边界两侧，这促使了这种非对称的双验证器设计，同时置信度门控保持了额外推理成本的低廉。在D_dev（n=308）上，使用GLM-4.7的AsymVerify相比于单遍分类在每个示例1.48次调用下提升了17.1的宏观F1分数，而升级验证器单独在D_dev上对每个测试的LLM后端的宏观F1分数提升了6.8到15.2，相较于其单遍基线。代码可在https://github.com/kaons-research/AsymVerify-ACL获取。

View on arXiv Download PDF AI Translation

cs.CL / 16 / 2607.20440

Answer-then-Edit: Reasoning Skeleton Editing for Anti-Distillation with Preserved Utility

先回答后编辑：保留效用的反蒸馏推理骨架编辑

Li, Fan, Pan, Mengting, Xu, Sijia, Wang, Xiaoyang, Chen, Chen, Zhang, Wenjie

Abstract

Proprietary large language models (LLMs) entail substantial intellectual and financial investment, making them valuable intellectual property (IP). However, even when deployed via black-box APIs, these models remain vulnerable to unauthorized knowledge distillation, which allows adversaries to cheaply extract and replicate model capabilities. To address this issue, anti-distillation (AD) has been proposed to generate defensive outputs that hinder distillation effectiveness, overcoming the limitation of watermarking-based approaches that rely on post-hoc verification. However, existing AD methods based on internal model perturbations struggle to balance anti-distillability and utility (e.g., answer accuracy and naturalness) of reasoning traces, with stronger defenses often causing significant utility loss. To fill this gap, we propose \textbf{\underline{S}}keleton-\textbf{\underline{G}}uided \textbf{\underline{R}}easoning \textbf{\underline{E}}diting (SGRE), an \textit{Answer-then-Edit} framework that performs post-hoc trace modification for anti-distillation. In the answer stage, the teacher model first generates clean reasoning traces, preserving the original reasoning accuracy while enabling more flexible control over trace naturalness. In the editing stage, we draw inspiration from Cognitive Load Theory (CLT) and introduce a three-stage strategy consisting of reasoning skeleton extraction, skeleton graph coarsening, and skeleton verbalization. These operations jointly perturb reasoning structures and augment textual complexity to amplify extraneous load on student models, hindering their acquisition of underlying reasoning patterns. Extensive experiments across diverse LLMs demonstrate that SGRE achieves state-of-the-art performance in reducing distillation effectiveness, while maintaining lossless reasoning accuracy and superior trace naturalness.

Chinese Translation

专有的大型语言模型（LLMs）需要大量的智力和财务投资，使其成为有价值的知识产权（IP）。然而，即使通过黑箱API部署，这些模型仍然容易受到未经授权的知识蒸馏攻击，这使得对手能够以低成本提取和复制模型能力。为了解决这一问题，提出了反蒸馏（AD）方法，以生成防御性输出，阻碍蒸馏的有效性，克服了依赖事后验证的水印方法的局限性。然而，现有基于内部模型扰动的AD方法在反蒸馏能力和推理痕迹的效用（例如，答案准确性和自然性）之间难以取得平衡，较强的防御往往会导致显著的效用损失。为填补这一空白，我们提出了 extbf{ extit{S}}keleton- extbf{ extit{G}}uided extbf{ extit{R}}easoning extbf{ extit{E}}diting（SGRE），一个 extit{先回答后编辑}框架，执行反蒸馏的事后痕迹修改。在答案阶段，教师模型首先生成干净的推理痕迹，保留原始推理的准确性，同时实现对痕迹自然性的更灵活控制。在编辑阶段，我们受到认知负荷理论（Cognitive Load Theory, CLT）的启发，提出了一种由推理骨架提取、骨架图粗化和骨架语言化组成的三阶段策略。这些操作共同扰动推理结构并增加文本复杂性，以增强对学生模型的额外负荷，阻碍其获取潜在的推理模式。对多种LLM进行的广泛实验表明，SGRE在降低蒸馏有效性方面实现了最先进的性能，同时保持无损的推理准确性和优越的痕迹自然性。

View on arXiv Download PDF AI Translation

cs.CL / 17 / 2607.20441

Belief Propagation in LLM World Models: Measuring Strategic Information Bias with Prediction Markets

大规模语言模型世界模型中的信念传播：通过预测市场衡量战略信息偏差

Khandoga, Mykola, Kostiuk, Yevhen, Polishko, Anton, Filipchuk, Yurii, Kozlov, Kostiantyn, Zamriy, Dmytro, Kiulian, Artur

Abstract

Every information ecosystem produces beliefs that shape strategic decisions. Both human analysts and AI systems inherit the blind spots of their information sources. We show that LLMs, combined with prediction markets, function as a calibrated instrument for measuring how far ecosystem-induced beliefs deviate from an external reference: LLMs extract the beliefs a text corpus implies, and prediction market price trajectories, anchored at resolution by realised outcomes, provide the calibration reference against which to quantify the deviation. We isolate the bias contribution of specific text through ablation: varying information context while holding the model fixed, with a contaminated model that knows actual outcomes as control. Applied to 111 Ukraine-related prediction markets, comprising approximately 93,000 predictions across four models, we find that English news context systematically biases territorial predictions, wrong 64 to 72 percent of the time when it pushes predictions toward territorial capture. A contaminated model that knows actual outcomes shows the same error rate, indicating that the bias originates primarily in the text. Supplementing with Ukrainian military-analytical sources reduces the bias for all clean models, while absolute-error gains are partial and model-dependent. We show that the distortion originates primarily in the sources, not the models. Consistent across four architectures, it will persist in any system that processes them and propagate into downstream decisions.

Chinese Translation

每个信息生态系统都会产生塑造战略决策的信念。人类分析师和人工智能系统都继承了其信息来源的盲点。我们展示了大规模语言模型（LLMs）结合预测市场，作为一种校准工具，测量生态系统引发的信念与外部参考之间的偏差程度：LLMs提取文本语料库所暗示的信念，而预测市场价格轨迹通过已实现结果进行校准，为量化偏差提供参考。我们通过消融实验隔离特定文本的偏差贡献：在保持模型不变的情况下，改变信息上下文，并以一个知道实际结果的污染模型作为对照。应用于111个与乌克兰相关的预测市场，涵盖约93,000个预测，结果表明，英语新闻上下文系统性地偏向于领土预测，当其推动预测向领土占领倾斜时，错误率为64%到72%。一个知道实际结果的污染模型显示相同的错误率，表明偏差主要源于文本。补充乌克兰军事分析来源减少了所有干净模型的偏差，但绝对误差的改善是部分的且依赖于模型。我们表明，扭曲主要源于信息来源，而非模型。四种架构的一致性表明，这种偏差将在任何处理这些信息的系统中持续存在，并传播到下游决策中。

View on arXiv Download PDF AI Translation

cs.CL / 18 / 2607.20442

Naver-News-KO: A Korean News Summarization Dataset for Open-Source Fine-Tuning of Summarization Models

Naver-News-KO：用于开源摘要模型微调的韩语新闻摘要数据集

Kim, Daekeun

Abstract

We release Naver-News-KO, a Korean news summarization dataset of 27,400 (document, summary) pairs collected from Naver News over a ten-day window in July 2022 across two categories (Economy and IT/Science; 77/23 split), with train/validation/test partitions of 22,194 / 2,466 / 2,740 and a mean per-record document-to-summary character-compression ratio of 6.03x. The dataset has been publicly hosted on the Hugging Face Hub since January 2023 and, as of May 2026, receives approximately 33,000 downloads per month; community-maintained Korean summarization models fine-tuned on it include Gemma-2B-ko and Gemma2-9B variants. This technical report (i) documents the collection protocol, the column schema, and the split construction, (ii) reports corpus-level statistics (length distributions, compression ratio, and a measured 16.8% near-duplicate title-Jaccard overlap between test and train that users should be aware of), (iii) positions the resource against other open Korean summarization corpora, (iv) provides a Lead-3 extractive reference point (ROUGE-1 55.1, ROUGE-L 50.6) and two reproducible fine-tuned baselines -- KoBART (R-1 56.6, BERTScore-F1 81.5) and Gemma-2B-ko with LoRA (R-1 55.3, BERTScore-F1 78.3) -- with release-time training scripts, and (v) clarifies the licensing and intended-use scope of the resource. The goal is to provide a citable reference for downstream work that already uses this dataset, not to propose a new benchmark.

Chinese Translation

我们发布了 Naver-News-KO，这是一个包含 27,400 对（文档，摘要）的韩语新闻摘要数据集，这些数据对是在 2022 年 7 月的十天内从 Naver News 收集的，涵盖两个类别（经济和 IT/科学；77/23 的比例），训练/验证/测试集的划分为 22,194 / 2,466 / 2,740，平均每条记录的文档到摘要的字符压缩比为 6.03 倍。该数据集自 2023 年 1 月以来已在 Hugging Face Hub 上公开托管，截至 2026 年 5 月，每月下载量约为 33,000 次；基于该数据集微调的社区维护的韩语摘要模型包括 Gemma-2B-ko 和 Gemma2-9B 变体。本技术报告 (i) 记录了收集协议、列模式和划分构建，(ii) 报告了语料库级统计数据（长度分布、压缩比，以及测试集与训练集之间测得的 16.8% 近重复标题 Jaccard 重叠，用户应注意），(iii) 将该资源与其他开放的韩语摘要语料库进行对比，(iv) 提供了一个 Lead-3 提取参考点（ROUGE-1 55.1，ROUGE-L 50.6）和两个可重复的微调基线——KoBART（R-1 56.6，BERTScore-F1 81.5）和带有 LoRA 的 Gemma-2B-ko（R-1 55.3，BERTScore-F1 78.3），附带发布时的训练脚本，以及 (v) 澄清了该资源的许可和预期使用范围。我们的目标是为已经使用该数据集的下游工作提供可引用的参考，而不是提出一个新的基准。

View on arXiv Download PDF AI Translation

cs.CL / 19 / 2607.20443

GLAN-QnA-KR: A Seedless Taxonomy-Driven Korean Instruction Corpus

GLAN-QnA-KR：一种无种子分类驱动的韩语指令语料库

Kim, Daekeun

Abstract

We release GLAN-QnA-KR, a 303,581-row openly redistributable Korean instruction-QA corpus produced via the seedless taxonomy-driven GLAN synthesis pipeline with Microsoft's Phi-3.5-MoE-instruct as the producer model (generation: 2024-12; release: 2024-12; licence: OpenRAIL). The corpus spans a flat taxonomy of 1,084 English-labelled disciplines paired with Korean question/answer text, a 100-900 difficulty scale, and a median of 313 question characters and 1,098 answer characters per record. Two properties are atypical for synthetic instruction data at this scale: (i) exact duplicate questions number only 1 in 303,581 rows and character-trigram near-duplicate clusters at Jaccard >= 0.9 number zero in a 5,000-sample probe, and (ii) a two-layer contamination audit against KMMLU, KoBEST (five sub-tasks), and HAE-RAE-Bench shows a maximum test-vs-corpus question-level character-trigram Jaccard of 0.163 with zero test items at Jaccard >= 0.7, and a maximum multilingual-E5 cosine of 0.901 with a single test item at cosine >= 0.90 and zero at >= 0.95, across 20,000 sampled GLAN questions and seven evaluation sets. At the time of release, this is, to our knowledge, the largest single-pipeline synthetic Korean instruction corpus verifiable on the Hugging Face Hub and the only Korean >=100k-row corpus built under a seedless taxonomy-driven protocol. This note documents the generation protocol, corpus statistics, the contamination audit, and the licensing boundary in a form suitable for downstream citation.

Chinese Translation

我们发布了GLAN-QnA-KR，这是一个包含303,581行的可公开再分发的韩语指令问答语料库，采用无种子分类驱动的GLAN合成管道生成，生产模型为微软的Phi-3.5-MoE-instruct（生成时间：2024年12月；发布：2024年12月；许可证：OpenRAIL）。该语料库涵盖了1,084个英文标记学科的平面分类，与韩语问答文本配对，难度范围为100-900，且每条记录的中位数为313个问题字符和1,098个回答字符。对于这种规模的合成指令数据，有两个特性不寻常：（i）303,581行中仅有1个完全重复的问题，而在5,000个样本探测中，Jaccard相似度大于等于0.9的字符三元组近重复簇数量为零；（ii）针对KMMLU、KoBEST（五个子任务）和HAE-RAE-Bench的两层污染审计显示，测试与语料库问题级字符三元组的最大Jaccard相似度为0.163，且在Jaccard大于等于0.7的测试项目数量为零，最大多语言E5余弦相似度为0.901，且在余弦大于等于0.90的测试项目数量为1，在余弦大于等于0.95的测试项目数量为零，基于20,000个采样的GLAN问题和七个评估集。在发布时，据我们所知，这是在Hugging Face Hub上可验证的最大单管道合成韩语指令语料库，也是唯一一个在无种子分类驱动协议下构建的超过100,000行的韩语语料库。本文记录了生成协议、语料库统计、污染审计以及适合下游引用的许可边界。

View on arXiv Download PDF AI Translation

cs.CL / 20 / 2607.20444

Confidently Deceptive: How Confidence Amplifies the Risk of LLM Deception

自信的欺骗：自信如何放大大型语言模型的欺骗风险

Asad, Ali, Obadinma, Stephen, Pattoo, Anshul, Zhang, Wenxuan, Zhu, Xiaodan

Abstract

Large language models (LLMs) can produce deceptive responses: outputs that mislead users in service of a contextually or experimentally induced goal. Yet it remains unclear how confidently models deceive and whether higher confidence makes deceptive responses more persuasive to end users. In this paper, we study these basic questions in various models and different deception datasets. We provide a comprehensive study measuring confidence through both verbalized self-reports and a range of logit-based estimators. We show that LLMs deliver deceptive responses with substantial verbalized confidence and that human annotators prefer the higher-confidence deceptive response 78% of the time in paired comparisons. Misalignment fine-tuning amplifies the problem. Confidence in deceptive responses rises across all three benchmarks, increasing the resulting potential risk, with effects generalizing beyond the training distribution. Strikingly, models classify their own deceptive outputs as deceptive at high rates (82.7% under misalignment) while still predicting they would produce them - recognition without avoidance. We argue that confident deception is a distinct alignment risk requiring evaluations that jointly measure deception, confidence, and awareness.

Chinese Translation

大型语言模型（LLMs）能够产生具有欺骗性的回应：这些输出在服务于上下文或实验诱导的目标时误导用户。然而，目前尚不清楚模型是如何自信地进行欺骗的，以及更高的自信是否使得欺骗性回应对最终用户更具说服力。本文研究了这些基本问题，涵盖了多种模型和不同的欺骗数据集。我们提供了一项全面的研究，通过口头自我报告和一系列基于logit的估计器来测量自信。我们显示，LLMs以相当高的口头自信提供欺骗性回应，并且在配对比较中，人类标注者在78%的情况下更倾向于选择高自信的欺骗性回应。错位微调加剧了这一问题。在所有三个基准测试中，欺骗性回应的自信水平上升，增加了潜在风险，并且其影响超出了训练分布。值得注意的是，模型以高比例（在错位情况下为82.7%）将自己的欺骗输出分类为欺骗，同时仍然预测它们会产生这些输出——这是一种识别而非避免。我们认为，自信的欺骗是一种独特的对齐风险，需要对欺骗、自信和意识进行联合评估。

View on arXiv Download PDF AI Translation

cs.CL / 21 / 2607.20445

SCoPE: Shift-Aware Speaker-Conditioned Priors for Emotion Recognition in Conversations

SCoPE：情感识别中基于说话者条件的情感转变感知先验

Kaplan, Burak Can, Wermter, Stefan

Abstract

In conversations, human emotions are transient; however, they tend to persist across multiple utterances. For example, we rarely switch instantly between contrasting emotions such as happiness and anger. Instead, emotions tend to evolve smoothly, and these patterns are often speaker-specific. Some people might escalate, while others gradually cool down over time. Furthermore, when emotions change during a conversation, they are often driven by contextual factors, such as newly received information or unexpected events. Even though progress has been made in Emotion Recognition in Conversations (ERC), most existing approaches still rely heavily on overt evidence and do not sufficiently model these non-apparent factors. Especially in multimodal settings, this makes these models fragile when the signals are noisy (e.g., occluded faces, slang expressions, or microphone noise). To address these limitations, we introduce Speaker-Conditioned Priors over Emotions (SCoPE). SCoPE is a light weight module that utilizes the emotional history of each speaker and explicitly models their priors for use in subsequent emotion classification. Second, we incorporate emotion shift prediction, a well-established concept in ERC, to guide the model in balancing the priors from SCoPE and multimodal evidence. Finally, we propose a shift-aware fusion mechanism that performs precision-weighted logit integration between multimodal evidence and the speaker prior, forming a Bayesian-inspired product-of-experts formulation. This dynamic fusion allows the model to rely on historical priors when emotions persist and to prioritize multimodal evidence when shifts are likely. Experimental results show our model achieves superior performance over recent state-of-the-art models on the IEMOCAP dataset in multimodal settings.

Chinese Translation

在对话中，人类情感是短暂的；然而，它们往往在多个发言中持续存在。例如，我们很少会在快乐和愤怒等对立情感之间瞬间切换。相反，情感往往平滑演变，这些模式通常是特定于说话者的。有些人可能会情感升级，而另一些人则会随着时间的推移逐渐冷静下来。此外，当对话中的情感发生变化时，往往是由上下文因素驱动的，例如新接收到的信息或意外事件。尽管在对话中的情感识别（Emotion Recognition in Conversations, ERC）方面取得了一定进展，但大多数现有方法仍然严重依赖明显证据，并未充分建模这些不明显的因素。尤其是在多模态环境中，这使得这些模型在信号嘈杂时（例如，面部遮挡、俚语表达或麦克风噪声）变得脆弱。为了解决这些局限性，我们提出了基于说话者条件的情感先验（Speaker-Conditioned Priors over Emotions, SCoPE）。SCoPE 是一个轻量级模块，利用每个说话者的情感历史，并明确建模他们的先验，以便在后续的情感分类中使用。其次，我们结合情感转变预测，这是一种在 ERC 中得到广泛认可的概念，以指导模型在 SCoPE 的先验和多模态证据之间进行平衡。最后，我们提出了一种感知转变的融合机制，在多模态证据和说话者先验之间进行精度加权的对数积分，形成一种贝叶斯启发的专家产品形式。这种动态融合使模型在情感持续时依赖历史先验，而在情感转变可能发生时优先考虑多模态证据。实验结果表明，我们的模型在多模态环境下的 IEMOCAP 数据集上，相较于近期的最先进模型，表现出优越的性能。

View on arXiv Download PDF AI Translation

cs.CL / 22 / 2607.20446

Distinguishing Artificial from Authentic: Evaluating LLMs for Detecting LLM-Generated Content

区分人工与真实：评估大型语言模型在检测生成内容中的应用

Leinonen, Juho, Denny, Paul

Abstract

As large language models (LLMs) are increasingly used by students to generate natural language responses and program code, there is growing interest in whether LLMs themselves can be used to distinguish AI-generated work from human-authored submissions. In this paper, we investigate the extent to which LLMs can detect their own generated content across multiple educational task types, including programming exercises, reflective writing, and short-answer questions. Using authentic student responses and multiple variants of LLM-generated answers, we evaluate detection performance under different prompting strategies and output formats. Our study addresses three research questions: (1) how accurately LLMs can identify their own outputs across task domains, (2) how detection effectiveness is influenced by factors such as prompt design, response length, and task type, and (3) what characteristics of LLM-generated responses contribute to successful or failed detection. Our findings show that LLM-based detection is highly task-dependent: detection is substantially more reliable for programming tasks and longer reflective responses, but performs poorly for short-answer questions, where LLMs frequently judge their own outputs as more human-like than authentic student responses. We further find that prompt framing and response verbosity have a pronounced effect on detectability in reflective writing tasks, with relatively minor prompt variations significantly reducing detection accuracy, while programming-related detection is more robust to prompt changes. Together, these results highlight both the potential and the limitations of LLM self-detection in educational settings and suggest caution in relying on LLMs as standalone tools for identifying AI-generated student work.

Chinese Translation

随着大型语言模型（LLMs）被学生越来越多地用于生成自然语言响应和程序代码，关于LLMs自身是否能够区分人工生成的作品与人类创作的提交物的兴趣日益增长。本文探讨了LLMs在不同教育任务类型中检测自身生成内容的能力，包括编程练习、反思性写作和简答题。我们使用真实学生的响应和多种LLM生成答案的变体，评估在不同提示策略和输出格式下的检测性能。我们的研究针对三个研究问题： (1) LLMs在不同任务领域中识别自身输出的准确性如何， (2) 提示设计、响应长度和任务类型等因素如何影响检测效果， (3) LLM生成响应的哪些特征有助于成功或失败的检测。我们的研究结果表明，基于LLM的检测高度依赖于任务：在编程任务和较长的反思性响应中，检测的可靠性显著更高，但在简答题中表现不佳，LLMs常常将自身输出判断为比真实学生响应更具人类特征。我们进一步发现，提示框架和响应冗长程度对反思性写作任务的可检测性有显著影响，提示的相对小变动会显著降低检测准确性，而与编程相关的检测对提示变化则更为稳健。综合来看，这些结果突显了LLM自我检测在教育环境中的潜力与局限性，并建议在依赖LLMs作为识别AI生成学生作品的独立工具时应保持谨慎。

View on arXiv Download PDF AI Translation

cs.CL / 23 / 2607.20447

thaulab@EEUCA 2026: Who Said What to Whom? A Targeting-Aware Neural-Symbolic Pipeline for Gaming Toxicity Detection

thaulab@EEUCA 2026：谁对谁说了什么？一种针对目标的神经符号管道用于游戏毒性检测

Guragain, Anmol, Estecha-Garitagoitia, Marcos, Enríquez, Luis Fernando D'Haro, de Córdoba, Ricardo

Abstract

This paper describes our system for the EEUCA 2026 Shared Task on toxicity classification in gaming chat. We implement a three-stage pipeline combining an ensemble of two compact transformers (DeBERTa-v3-base, 184M; XLM-RoBERTa-base, 278M) with a Linguistically-Informed Mediator (LIM) that resolves inter-model disagreements through corpus-backed lexical normalization, class-conditional unigram scoring, multilingual profanity detection, and agentive targeting analysis grounded in speech act theory. The LIM specifically targets the minority classes (Hate \& Harassment, Threats, and Extremism), which are the most safety-critical categories in real-world gaming moderation. To address the extreme class imbalance (1{,}450:1 Non-toxic to Extremism ratio), we introduce a two-stage data augmentation strategy using only the provided training data. Our system achieves a Macro F1 of 0.6441 and accuracy of 0.9062 on the official test set, ranking 3rd in Macro F1 and 1st in accuracy among all teams. The proposed pipeline is domain-portable: adapting to other gaming platforms requires substituting only the game-specific entity lexicon. Code is publicly available at https://github.com/Anmol2059/thaulab\_EEUCA.

Chinese Translation

本文描述了我们在EEUCA 2026共享任务中用于游戏聊天毒性分类的系统。我们实现了一个三阶段管道，结合了两个紧凑型变换器的集成（DeBERTa-v3-base，184M；XLM-RoBERTa-base，278M）与一个语言信息中介（Linguistically-Informed Mediator, LIM），通过基于语料库的词汇标准化、类别条件的单元评分、多语言粗口检测以及基于言语行为理论的代理目标分析来解决模型间的分歧。LIM特别针对少数类别（仇恨与骚扰、威胁和极端主义），这些类别在现实世界的游戏管理中是最为安全关键的。为了应对极端的类别不平衡（非毒性与极端主义的比例为1{,}450:1），我们引入了一种两阶段的数据增强策略，仅使用提供的训练数据。我们的系统在官方测试集上实现了0.6441的宏F1值和0.9062的准确率，在所有团队中宏F1排名第3，准确率排名第1。所提出的管道具有领域可移植性：适应其他游戏平台只需替换特定于游戏的实体词汇。代码已公开，地址为https://github.com/Anmol2059/thaulab_EEUCA。

View on arXiv Download PDF AI Translation

cs.CL / 24 / 2607.20448

Domyn-Small: A European 10B Reasoning Language Model

Domyn-Small：一种欧洲10B推理语言模型

Angarano, Simone, Bertolotti, Francesco, D'Ambrosio, Federico, Resta, Michele, Rognoni, Alessandro, Ruggeri, Nicolò, Salvati, Dario, Valenti, Andrea, Veneri, Alberto, Cimmino, Martin

Abstract

We introduce Domyn-Small, a 10-billion-parameter open-weight reasoning language model released under the MIT license. Domyn-Small is the product of an initial pre-training phase on 9 trillion tokens multilingual data, followed by a post-training pipeline for reasoning, instruction following, and context extension. For the latter, we performed a Continued Pre-Training (CPT) phase that doubles the native context window to 32K tokens, followed by SFT with a math-focused annealing run. Finally, the RL phase includes GRPO with verifiable rewards, DPO, and a multi-environment GRPO stage spanning five task domains: mathematics, code, multiple-choice QA, instruction-following, and tool calling. The 32K-token native context extends to 128K at inference via YaRN, and a chat-template toggle enables dual-mode reasoning. Against peer models in the 7--10B class (Qwen3.5-9B, OLMo-3-7B-Think, Nemotron-Nano-8B, Ministral-3-8B), Domyn-Small achieves a strong accuracy-efficiency balance: it produces roughly one-third as many tokens as Qwen3.5-9B and approximately 35% of OLMo-3-7B-Think's token budget on core reasoning benchmarks, while delivering strong instruction-following (IFEval 79.9) and competitive science reasoning (GPQA-Diamond 50.0). We release the weights and the post-training recipe alongside Domyn Swarm (Apache~2.0), an open-source framework for scalable LLM inference on HPC clusters developed during this program and used throughout this work.

Chinese Translation

我们介绍了Domyn-Small，这是一种具有100亿参数的开放权重推理语言模型，采用MIT许可证发布。Domyn-Small是基于9万亿标记的多语言数据进行初始预训练阶段的产物，随后进行了推理、指令跟随和上下文扩展的后训练流程。对于后者，我们进行了持续预训练（Continued Pre-Training, CPT）阶段，将原生上下文窗口扩展到32K标记，随后进行了以数学为重点的退火训练（SFT）。最后，强化学习（RL）阶段包括具有可验证奖励的GRPO、DPO以及跨越五个任务领域的多环境GRPO阶段：数学、代码、多项选择问答、指令跟随和工具调用。32K标记的原生上下文在推理时通过YaRN扩展到128K，并且聊天模板切换功能支持双模式推理。在与7-10B类的同类模型（如Qwen3.5-9B、OLMo-3-7B-Think、Nemotron-Nano-8B、Ministral-3-8B）进行比较时，Domyn-Small在准确性和效率之间取得了良好的平衡：在核心推理基准上，它生成的标记数量大约是Qwen3.5-9B的三分之一，且约为OLMo-3-7B-Think标记预算的35%，同时在指令跟随（IFEval 79.9）和科学推理（GPQA-Diamond 50.0）方面表现出色。我们与Domyn Swarm（Apache 2.0）一起发布了模型权重和后训练配方，Domyn Swarm是一个用于在高性能计算集群上进行可扩展LLM推理的开源框架，开发于本项目期间并在整个工作中使用。

View on arXiv Download PDF AI Translation

cs.CL / 25 / 2607.20449

The Storyteller in the Model: Narrative Pattern Inheritance, Escalation Dynamics, and Alignment Governance in LLMs

模型中的讲故事者：大语言模型中的叙事模式继承、升级动态与对齐治理

Rigby, Adam, Saremi, Raz, Sohrabinejad, Azadeh, Rahimi, Mehdi

Abstract

LLMs are trained predominantly on human-authored text, yet the structural and narrative conventions embedded in that text are rarely examined as a source of systematic behavioral influence, or as a governance risk in deployed systems. This paper considers whether the storytelling patterns inherent in published human writing, including archetypal roles such as protagonist, antagonist, and underdog, as well as tension-and-resolution narrative arcs, are absorbed during training and subsequently surface in LLM outputs, causing responses to drift toward unexpected, adversarial, or rhetorically enticing behaviors over extended interactions. Through a systematic literature review and cross-paper analysis of recent empirical studies on LLM alignment, persona dynamics, emergent misalignment, and user interaction patterns, we observe evidence bearing on this hypothesis. The findings reveal three key patterns. First, LLMs reproduce statistical patterns from their training data rather than reasoning independently. Second, measurable latent traits, including sycophancy and deceptiveness, emerge reliably across unrelated prompts. Third, fine-tuning on a narrow narrative task can produce unintended behavioral changes well beyond that task. Furthermore, evidence suggests that persuasive, narrative-style outputs are among the most common LLM products in real-world usage, amplifying these risks. Narrative drift constitutes an unmonitored escalation pathway in deployed AI systems, one that evades discrete-incident detection mechanisms and requires dedicated monitoring instruments.

Chinese Translation

大语言模型（LLMs）主要在由人类创作的文本上进行训练，但这些文本中嵌入的结构和叙事惯例作为系统性行为影响的来源，或作为已部署系统中的治理风险，鲜有研究。本论文探讨了已发布的人类写作中固有的讲故事模式，包括主角、反派和弱者等原型角色，以及紧张与解决的叙事弧线，是否在训练过程中被吸收，并随后在LLM输出中显现，导致响应在长期交互中偏向意想不到的、对抗性的或修辞上诱人的行为。通过对近期关于LLM对齐、角色动态、新兴不对齐和用户交互模式的实证研究进行系统文献综述和跨论文分析，我们观察到支持这一假设的证据。研究结果揭示了三个关键模式。首先，LLMs重现其训练数据中的统计模式，而非独立推理。其次，可测量的潜在特征，包括谄媚和欺骗性，在无关提示中可靠地出现。第三，在狭窄叙事任务上的微调可能导致超出该任务的意外行为变化。此外，证据表明，具有说服力的叙事风格输出是现实世界使用中最常见的LLM产品之一，进一步放大了这些风险。叙事漂移构成了已部署人工智能系统中一个未监控的升级路径，这一路径逃避了离散事件检测机制，并需要专门的监测工具。

View on arXiv Download PDF AI Translation

cs.CL / 26 / 2607.20450

ShriNep@EEUCA 2026: RAKSHAK - Multi-Task DeBERTa with Rationale Distillation and Jigsaw-Augmented Training for Toxic Intent Classification

ShriNep@EEUCA 2026：RAKSHAK - 具有推理蒸馏和拼图增强训练的多任务 DeBERTa 用于有毒意图分类

Karki, Binayak, Kafle, Aryan, Ghimire, Pingala

Abstract

This paper presents two systems for the GameTox Shared Task at the Workshop on EEUCA at ACL 2026, which requires classifying World of Tanks chat utterances into six fine-grained toxic intent categories (Labels 0-5). Severe class imbalance, domain-specific multilingual slang, and extremely scarce data for rare categories such as Threats (Label 4, 60 samples) and Extremism (Label 5, 24 samples) make this a challenging classification problem. Our primary submission, RAKSHAK (rak s. aka, Sanskrit for "Protector"), is a multi-task DeBERTa-v3-base (He et al., 2022) framework combining rationale distillation from Qwen2.5-14B (An et al., 2024), Supervised Contrastive Loss, and dedicated rare-class binary heads. RAKSHAK's training data is augmented with cross-domain transfer from the Jigsaw Toxic Comment dataset (16,225 samples mapped to Labels 1-4) and 100 LLM-generated extremism samples for Label 5. Our secondary system (M1) fine-tunes DeBERTa-v3-base with Focal Loss on the original GameTox data plus the same 100 extremism samples, without Jigsaw transfer. RAKSHAK achieves a Macro F1 of 0.5883 on the official test set, ranking 7th out of 35 participating teams, while M1 achieves 0.5252 Macro F1. An ablation comparing M1 with and without Jigsaw data shows that cross-domain transfer accounts for +2.6 F1 points, while RAKSHAK's multi-task architecture contributes a further +3.7 points.

Chinese Translation

本文提出了两个系统，用于2026年ACL会议EEUCA研讨会的GameTox共享任务，该任务要求将《坦克世界》聊天语句分类为六个细粒度的有毒意图类别（标签0-5）。严重的类别不平衡、特定领域的多语言俚语以及对于稀有类别（如威胁（标签4，60个样本）和极端主义（标签5，24个样本））的数据极其稀缺，使得这是一个具有挑战性的分类问题。我们的主要提交系统RAKSHAK（rak s. aka，梵语意为“保护者”）是一个多任务DeBERTa-v3-base（He et al., 2022）框架，结合了来自Qwen2.5-14B（An et al., 2024）的推理蒸馏、监督对比损失和专门的稀有类别二元头。RAKSHAK的训练数据通过来自Jigsaw有毒评论数据集（16,225个样本映射到标签1-4）的跨领域迁移和100个LLM生成的极端主义样本（标签5）进行了增强。我们的次要系统（M1）在原始GameTox数据及相同的100个极端主义样本上微调DeBERTa-v3-base，未进行Jigsaw迁移。RAKSHAK在官方测试集上取得了0.5883的宏F1分数，在35个参与团队中排名第7，而M1的宏F1分数为0.5252。对比M1在有无Jigsaw数据的消融实验表明，跨领域迁移贡献了+2.6 F1点，而RAKSHAK的多任务架构进一步贡献了+3.7点。

View on arXiv Download PDF AI Translation

cs.CL / 27 / 2607.20451

Semantic Field Theory: Historical Origin, Higher-Order Interaction, and Stabilized Semantic Inference

语义场理论：历史起源、高阶互动与稳定语义推理

Vartziotis, Dimitris

Abstract

Semantic Field Theory (SFT) has developed from a philosophical critique of strong anti-formalist readings of language games into a proposed computational model class for lexical semantics, higher order composition, and stabilized interpretation. This paper reconstructs that evolution and gives SFT a sharper mathematical core suitable for independent evaluation in computational linguistics and representation learning. The central proposal is that a tractable level of linguistic organization can be modeled through lexical representations expressed as semantic fields, through contextual deformation of those fields, through interaction terms defined over subsets of tokens, and through stabilization governed by semantic energy dynamics. The paper contributes five formal elements. First, it defines a semantic field model as a tuple consisting of a semantic space, a lexical field lifting, a contextual deformation map, an interaction complex, and an interpretation functional. Second, it proves a Gaussian product closure result showing that multiplicative field interactions have explicit centers, precisions, and compatibility factors. Third, it generalizes the three-word problem by using Mobius inversion on the subset lattice to isolate irreducible semantic interactions of arbitrary order. Fourth, it introduces an order spectrum that measures how much field mass is explained at each interaction order. Fifth, it formulates stabilized interpretation as minimization of an energy functional associated with the sentence and gives existence, descent, and stability conditions. A small worked example shows how a three-word summer day triple can be represented by Gaussian semantic fields, implemented in Python, and summarized by a flow diagram. The result is not a completed theory of natural language meaning and does not replace social, pragmatic, or normative accounts of language.

Chinese Translation

语义场理论（Semantic Field Theory, SFT）从对语言游戏强反形式主义解读的哲学批判发展为一种针对词汇语义、高阶组合和稳定解释的计算模型类别。本文重构了这一演变，并为SFT提供了更为清晰的数学核心，以便在计算语言学和表示学习中进行独立评估。核心提案是，通过将词汇表示建模为语义场，通过对这些场的上下文变形，通过定义在标记子集上的交互项，以及通过由语义能量动态控制的稳定化，可以对可处理的语言组织水平进行建模。本文贡献了五个形式要素。首先，定义了一个语义场模型，该模型是一个包含语义空间、词汇场提升、上下文变形映射、交互复合体和解释泛函的元组。其次，证明了一个高斯乘积闭合结果，表明乘法场交互具有明确的中心、精度和兼容性因子。第三，通过在子集格上使用莫比乌斯反演来隔离任意阶不可约语义交互，从而推广了三词问题。第四，引入了一个阶谱，用于衡量在每个交互阶上解释了多少场质量。第五，将稳定解释表述为与句子相关的能量泛函的最小化，并给出了存在性、下降性和稳定性条件。一个小的实例展示了如何通过高斯语义场表示一个三词的夏日三重项，使用Python实现，并通过流程图进行总结。该结果并不是自然语言意义的完整理论，也不取代对语言的社会、语用或规范性解释。

View on arXiv Download PDF AI Translation

cs.CL / 28 / 2607.20453

A Knowledge-Injection Framework for Zero-Shot Adaptation of LLMs to Delirium Prediction

一种用于零-shot适应大型语言模型的知识注入框架以进行谵妄预测

Sena, Jessica, Priyadarshani, Shesadree, Contreras, Miguel, Gandhi, Bharat, Siegel, Scott, Nerella, Subhash, Rashidi, Parisa

Abstract

Large language models show promise for clinical prediction, but zero-shot performance on specialized tasks is limited by incomplete domain knowledge, especially for smaller locally deployable models. We present a lightweight knowledge-injection framework for zero-shot ICU delirium prediction that augments a deterministic natural-language summary of structured electronic health record data with an external clinical knowledge report at inference time, without fine-tuning or retrieval. We evaluate LLaMA 3.1 8B and LLaMA 3.3 70B on 3,160 ICU admissions from the MIMIC IV dataset. Adding a clinically meaningful external knowledge report improves AUROC by 8.57 percentage points for the 8B model and 1.99 percentage points for the 70B model compared to no external knowledge. Relative to a GPT-5.2 frontier-model reference without external knowledge report (AUROC 68.86%), knowledge injection reduces the performance gap from 15.66 to 7.09 AUROC points for LLaMA 8B and from 5.30 to 3.31 AUROC points for LLaMA 70B. Random control reports do not improve performance and often degrade it, indicating that gains depend on clinically meaningful content rather than added prompt length alone. SHAP-based attribution further confirms that the injected knowledge is actively used during prediction. These findings suggest that inference-time knowledge injection can narrow the gap between locally deployable open-weight models and frontier closed models while preserving a practical, privacy-preserving workflow for resource-constrained clinical settings.

Chinese Translation

大型语言模型在临床预测中展现出潜力，但在专业任务上的零-shot性能受到不完整领域知识的限制，尤其是对于较小的本地可部署模型。我们提出了一种轻量级的知识注入框架，用于零-shot重症监护室（ICU）谵妄预测，该框架在推理时将结构化电子健康记录数据的确定性自然语言摘要与外部临床知识报告相结合，而无需进行微调或检索。我们在MIMIC IV数据集中对3,160例ICU入院病例评估了LLaMA 3.1 8B和LLaMA 3.3 70B模型。与没有外部知识相比，添加具有临床意义的外部知识报告使8B模型的AUROC提高了8.57个百分点，70B模型提高了1.99个百分点。相较于没有外部知识报告的GPT-5.2前沿模型（AUROC 68.86%），知识注入将LLaMA 8B的性能差距从15.66缩小至7.09 AUROC点，将LLaMA 70B的性能差距从5.30缩小至3.31 AUROC点。随机控制报告并未改善性能，且常常导致性能下降，表明收益依赖于临床意义内容，而非仅仅是增加提示长度。基于SHAP的归因进一步确认注入的知识在预测过程中被积极使用。这些发现表明，推理时的知识注入可以缩小本地可部署开放权重模型与前沿封闭模型之间的差距，同时保持适用于资源有限的临床环境的实用、隐私保护的工作流程。

View on arXiv Download PDF AI Translation

cs.CL / 29 / 2607.20454

Response drift across frontier large language models

前沿大型语言模型中的响应漂移

Aledhari, Mohammed, Aledhari, Ali, Aledhari, Fatimah, Eathamokkala, Gowtham Venkat, Rahouti, Mohamed

Abstract

All frontier large language models (LLMs) exhibit response drift -- producing outputs that deviate from expert-validated references -- yet the magnitude and structure of this drift remain uncharacterised by systematic human evaluation. Here we report a fully crossed evaluation in which 47 geographically diverse participants each assessed all 62 multidomain questions across ten frontier LLMs under blinded conditions, yielding 29,140 independent assessments. Every model drifts, but drift magnitude varies substantially: eight models converge on a statistically indistinguishable ceiling (78-81% deviation), while two achieve lower deviation (47-49%). Drift profiles differ across six domains and 62 questions, with pairwise correlations among ceiling models exceeding r = 0.85. Automated similarity metrics explain less than 2% of variance in human judgements. These findings reveal that response drift is universal across frontier LLMs, domain- and question-dependent in structure, and accessible only through human-centred evaluation.

Chinese Translation

所有前沿大型语言模型（LLMs）都表现出响应漂移——生成的输出偏离专家验证的参考标准——然而这种漂移的幅度和结构尚未通过系统的人类评估进行表征。在此，我们报告了一项全面交叉的评估，其中47名地理分布多样的参与者在盲测条件下对十个前沿LLM的62个多领域问题进行了评估，共产生了29,140个独立评估。每个模型都存在漂移，但漂移幅度差异显著：八个模型趋向于一个统计上不可区分的上限（78-81%的偏差），而两个模型的偏差较低（47-49%）。漂移特征在六个领域和62个问题中有所不同，具有上限的模型之间的成对相关性超过r = 0.85。自动化相似性指标解释的人类判断方差不足2%。这些发现揭示了响应漂移在前沿LLM中是普遍存在的，其结构依赖于领域和问题，并且只能通过以人为中心的评估来获取。

View on arXiv Download PDF AI Translation

cs.CL / 30 / 2607.20455

RE-AD: Real-Time Requirement Adherence for Data Labeling

RE-AD：实时需求遵循用于数据标注

Malreddy, Siddarth, Nigam, Ishan, Arora, Akshay, Mittal, Nikhil, Sahu, Subrat

Abstract

Human-annotated data remains fundamental to training frontier Large Language Models (LLMs). However, crowd-sourced annotations often suffer from quality issues stemming from annotator misunderstanding or lack of engagement. To address this, we introduce a real-time requirement adherence (RE-AD) framework that leverages LLMs to proactively validate labeling quality. Our methodology involves decomposing Standard Operating Procedures (SOPs) into atomic rules via self-reflection, categorizing them by complexity, and applying tiered validation strategies. Evaluated on a synthetic benchmark, the system achieved an F1 score of 0.749. Furthermore, production deployment resulted in annotators accepting and fixing 82% of the errors flagged by the framework. We include ablation studies to demonstrate the impact of our core design decisions.

Chinese Translation

人工标注的数据仍然是训练前沿大型语言模型（LLMs）的基础。然而，众包标注常常受到标注者误解或缺乏参与感所导致的质量问题。为了解决这一问题，我们提出了一种实时需求遵循（RE-AD）框架，该框架利用LLMs主动验证标注质量。我们的方法论包括通过自我反思将标准操作程序（SOPs）分解为原子规则，按复杂性对其进行分类，并应用分层验证策略。在一个合成基准上评估，该系统达到了0.749的F1分数。此外，生产部署使得标注者接受并修正了框架标记的82%的错误。我们还包括了消融研究，以展示我们核心设计决策的影响。

View on arXiv Download PDF AI Translation

cs.CL / 31 / 2607.20456

Learn2Zinc: Fine-tuning Small Language Models for Text-to-Model Translation in MiniZinc

Learn2Zinc：针对 MiniZinc 的文本到模型翻译的小型语言模型的微调

Kadioglu, Serdar, Uppuluri, Karthik

Abstract

Large language models excel at code generation for mainstream programming languages but struggle with rare, domain-specific languages such as MiniZinc, a constraint modeling language for combinatorial problems. We investigate whether targeted fine-tuning can teach small language models (0.6B to 20B parameters) to generate syntactically correct and semantically valid MiniZinc models from natural language problem descriptions. Our key finding is that syntax errors dominate failures when working with this domain specific language: the out-of-the-box execution accuracy of small language models such as Qwen3, LLaMa, Gemma, and GPT-OSS is near-zero. We propose a cross-model error bootstrapping approach that collects syntax errors from multiple LLM runs and leverage those to curate an error correction training dataset. This dataset allows us fine-tune small language models that consistently improves both direct code generation and chain-of-thought approaches across all model sizes. With self-reflection and ensembling, our approach achieves up to 98\% execution accuracy. In parallel, solution accuracy still remains at 35\%, indicating that while syntax is learnable, constraint reasoning remains a challenge. We contribute our fine-tuning pipeline, datasets, and models to opens-source for further research on text-to-model translation.

Chinese Translation

大型语言模型在主流编程语言的代码生成方面表现出色，但在处理稀有的领域特定语言时却面临困难，例如 MiniZinc，这是一种用于组合问题的约束建模语言。我们研究了有针对性的微调是否能够教会小型语言模型（参数数量从 0.6B 到 20B）从自然语言问题描述中生成语法正确且语义有效的 MiniZinc 模型。我们的主要发现是，语法错误在处理这种领域特定语言时占据了失败的主导地位：小型语言模型（如 Qwen3、LLaMa、Gemma 和 GPT-OSS）的开箱即用执行准确率接近零。我们提出了一种跨模型错误引导的方法，该方法从多个大型语言模型的运行中收集语法错误，并利用这些错误来策划一个错误修正训练数据集。这个数据集使我们能够微调小型语言模型，从而在所有模型规模上持续改善直接代码生成和思维链方法。通过自我反思和集成，我们的方法实现了高达 98\% 的执行准确率。同时，解决方案的准确率仍然保持在 35\\%，这表明虽然语法是可以学习的，但约束推理仍然是一个挑战。我们将我们的微调管道、数据集和模型贡献给开源，以便进一步研究文本到模型翻译。

View on arXiv Download PDF AI Translation

cs.CL / 32 / 2607.20457

Dropping the Anchor: Statistical Context Summarization for Distributed Systems via Pulsar Attention

锚定：通过脉冲注意力实现分布式系统的统计上下文摘要

Sood, Aryan, Acharya, Shantanu

Abstract

Inference with large language models (LLMs) on long sequences is computationally expensive due to the quadratic complexity of self-attention. Distributed blockwise methods such as Star Attention reduce this cost by sharding context across hosts, but rely on prepending a static, content-blind copy of the first block to every host. We propose Pulsar Attention, which replaces the static anchor with two lightweight, content-aware components: a small attention-sink prefix that stabilizes softmax, and compact cross-block summaries built via a Max-IDF heuristic that selects chunks containing globally rare tokens. This reduces the Phase 1 per-GPU FLOPs by up to 3.3$\times$ over Star Attention while retaining an identical KV cache footprint. On RULER and BABILong with Llama-3.1-8B, Pulsar Attention outperforms both Star Attention and dense attention at sequence lengths up to 128K tokens, with absolute gains of up to 4.7% over the dense baseline.

Chinese Translation

在长序列上使用大型语言模型（LLMs）进行推理计算开销较大，原因在于自注意力的平方复杂度。分布式块状方法如星注意力（Star Attention）通过在主机之间划分上下文来降低这一成本，但依赖于在每个主机前添加一个静态的、与内容无关的第一个块的副本。我们提出了脉冲注意力（Pulsar Attention），用两个轻量级的、内容感知的组件替代静态锚定：一个小型的注意力汇聚前缀用于稳定softmax，以及通过最大逆文档频率（Max-IDF）启发式构建的紧凑跨块摘要，选择包含全局稀有标记的片段。这将第一阶段每个GPU的浮点运算次数（FLOPs）相较于星注意力减少了最多3.3倍，同时保持相同的键值缓存（KV cache）占用。在使用Llama-3.1-8B的RULER和BABILong数据集上，脉冲注意力在序列长度达到128K标记时，优于星注意力和密集注意力，绝对增益最高可达4.7%，相较于密集基线。

View on arXiv Download PDF AI Translation

cs.CL / 33 / 2607.20458

CAMeR: Keyword-Gated Hybrid Activation for Adaptive Memory Retention in LLM Agents

CAMeR：用于大语言模型代理的关键词门控混合激活的自适应记忆保留

Lai, Haowen

Abstract

Large language model (LLM) agents operating over extended dialogues accumulate vast amounts of information, yet existing memory systems either retain everything indiscriminately or apply uniform forgetting heuristics that fail to distinguish relevant from irrelevant knowledge. We present CAMeR (Context-Activated Memory Reinforcement), a memory retention framework combining keyword-gated hybrid activation -- a joint symbolic (word-level Jaccard) and sub-symbolic (embedding cosine) gating mechanism -- with adaptive weight dynamics. CAMeR computes a hybrid similarity score for each memory-query pair; memories exceeding a threshold receive reinforcement while all memories undergo controlled decay. We introduce CAMeR-Bench, a 76-memory, 100-round benchmark spanning 8 topic clusters with graded activation frequency, designed to test adaptive retention where existing benchmarks (LoCoMO, LongMemEval) cannot. On CAMeR-Bench, CAMeR's keyword gate achieves a 1.6$\times$ larger retention gap between high-frequency and never-referenced memories compared to embedding-only gating (scissors gap: 0.039 vs. 0.024), while time-driven baselines (Oblivion, SuperLocalMemory) collapse to near-zero weights over 100 rounds. CAMeR's top-5 retrieval saves 83.2\% tokens versus full-context approaches (39k vs. 231k cumulative) while producing weight signals that improve retrieval precision. Through 8 ablation conditions we establish that the keyword gate -- not learnable decay -- is the primary performance driver at this scale. Our findings demonstrate that hybrid symbolic-neural gating provides a simple yet effective mechanism for adaptive memory retention in LLM agents.

Chinese Translation

在扩展对话中运行的大语言模型（LLM）代理积累了大量信息，但现有的记忆系统要么无差别地保留所有信息，要么应用统一的遗忘启发式方法，无法区分相关知识与无关知识。我们提出了CAMeR（上下文激活记忆强化），这是一种结合关键词门控混合激活——一种联合符号（词级Jaccard）和亚符号（嵌入余弦）门控机制——与自适应权重动态的记忆保留框架。CAMeR为每个记忆-查询对计算混合相似度分数；超过阈值的记忆会得到强化，而所有记忆则经历受控衰减。我们引入了CAMeR-Bench，这是一个包含76个记忆、100轮的基准测试，涵盖8个主题集群，具有分级激活频率，旨在测试现有基准（LoCoMO、LongMemEval）无法实现的自适应保留。在CAMeR-Bench上，CAMeR的关键词门控在高频和未引用记忆之间实现了1.6倍的保留差距，相较于仅使用嵌入的门控（剪刀差距：0.039对0.024），而基于时间的基线（Oblivion、SuperLocalMemory）在100轮后权重接近于零。CAMeR的前5次检索相比于全上下文方法节省了83.2%的标记（39k对231k的累积），同时产生的权重信号提高了检索精度。通过8个消融条件，我们确定关键词门控——而非可学习衰减——是这一规模下主要的性能驱动因素。我们的研究结果表明，混合符号-神经门控提供了一种简单而有效的机制，用于大语言模型代理的自适应记忆保留。

View on arXiv Download PDF AI Translation

cs.CL / 34 / 2607.20459

THOR: A Theta-Gamma Hierarchical Oscillatory Reasoning Framework for Multi-hop QA

THOR：一种用于多跳问答的Theta-Gamma层次振荡推理框架

Ling, Ziyang, Xu, Ronald X., Sun, Mingzhai

Abstract

Multi-hop question answering requires retrieving and integrating evidence from multiple contexts. Despite the rapid progress of current research, multi-hop reasoning remains constrained by two persistent limitations: attention decay, where the model's focus on main question degrades as the reasoning chain grows, and error accumulation, where mistakes propagate across hops and compounds into final failure. Inspired by Theta-Gamma hierarchical oscillation which decouples global planning from local retrieval, enabling efficient attention transfer between hops and a verification and repair mechanism that interrupts the accumulation of errors in the wrong paths, we present THOR, a brain-inspired Theta-Gamma hierarchical oscillatory reasoning framework. Extensive comparative experiments and specific validation experiments on multi-hop QA benchmarks demonstrate that THOR improves answer accuracy and robustness while mitigating limitations, showcasing its generalization across different backbones.

Chinese Translation

多跳问答需要从多个上下文中检索和整合证据。尽管当前研究进展迅速，但多跳推理仍然受到两个持续限制的制约：注意力衰减，即随着推理链的增长，模型对主要问题的关注度下降；以及错误累积，即错误在跳跃中传播并最终导致失败。受到Theta-Gamma层次振荡的启发，该方法将全局规划与局部检索解耦，能够在跳跃之间实现高效的注意力转移，并提供一种验证和修复机制，以中断错误路径中的错误累积，我们提出了THOR，一种受大脑启发的Theta-Gamma层次振荡推理框架。大量的比较实验和针对多跳问答基准的特定验证实验表明，THOR提高了答案的准确性和鲁棒性，同时减轻了局限性，展示了其在不同基础架构上的泛化能力。

View on arXiv Download PDF AI Translation

cs.CL / 35 / 2607.20460

Instruct-FD: Can Your Full-Duplex Speech System Follow Turn-Taking Instructions?

Instruct-FD：您的全双工语音系统能否遵循轮流发言指令？

Tang, Yuzhi, Ma, Wentao, Zhao, Xiling, Salimi, Ahmad, Moridani, Sepehr Harfi, Shen, Dongming, Wang, Jixuan, Abdulrazzag, Abdulrahman, Aubry, Murdock, Chen, Yu-Hua, Lee, Daniel, Lee, Jaewon, Mackey, Jonah, Meng, Silin, Stranges, Nicholas, Xiong, Chenxu, Yu, Hao, Zhu, Yi, Li, Mu, Smola, Alex

Abstract

Current full-duplex (FD) spoken dialogue systems can produce fluid interactions, yet it remains unclear whether they can adapt their turn-taking behavior when explicitly instructed. This is critical for real-world deployment, where conversational policies vary across applications (e.g., proactive tutoring vs. passive counseling). We introduce Instruct-FD, an instruction-conditioned benchmark for evaluating controllable turn management in FD systems. To enable this, we develop a human-validated, scalable synthetic pipeline that generates instruction-conditioned conversations, along with a deployment-agnostic multi-turn evaluation protocol and an LLM-based judge. Benchmarking six state-of-the-art full-duplex systems reveals a substantial gap in instruction-following turn management: the best model achieves only 64.4% adherence. Performance is highly uneven across behaviors and scenarios, with proactive behaviors such as model backchanneling and interruption remaining particularly challenging. These findings establish instruction-following turn management as a crucial direction for building adaptable and deployable full-duplex dialogue systems.

Chinese Translation

当前的全双工（FD）口语对话系统能够产生流畅的交互，但尚不清楚它们是否能够在明确指令下调整其轮流发言行为。这对于实际应用至关重要，因为对话策略在不同应用中有所不同（例如，主动辅导与被动咨询）。我们提出了Instruct-FD，这是一个用于评估FD系统中可控轮流管理的指令条件基准。为此，我们开发了一个经过人类验证的、可扩展的合成管道，生成指令条件的对话，并设计了一个与部署无关的多轮评估协议及基于大语言模型（LLM）的评判机制。对六个最先进的全双工系统进行基准测试显示，指令遵循的轮流管理存在显著差距：最佳模型的遵循率仅为64.4%。不同行为和场景的表现差异显著，主动行为（如模型的回声和打断）尤其具有挑战性。这些发现确立了指令遵循的轮流管理作为构建可适应和可部署的全双工对话系统的重要方向。

View on arXiv Download PDF AI Translation

cs.CL / 36 / 2607.20461

Can Valence Reflect Morality in Natural Language? A Preliminary Annotation Study

情感能否反映自然语言中的道德？一项初步注释研究

O'Dwyer, Jonny, Bendechache, Malika, McCormack, Louise, Calik, Elif, Ranjbarzadeh, Ramin, Muhammad, Dost, Bozcheloei, Shokofeh Anari, Singh, Ishita

Abstract

Present implementations of artificial intelligence (AI) ethics do not adequately take feelings, or affect, into account. If AI should be aligned with human ethics, it seems reasonable to thoroughly investigate the possibility of AI behaviour that mirrors virtuous human ethical conduct, where feelings play a role in the actions, judgements or statements one makes. Furthermore, while prominent theories of normative ethics are often discussed in terms of their differences and shortcomings, Virtue, Consequentialist, and Kantian Deontological ethics all share a common feature of considering human feeling to some degree while the popular descriptive ethics theory, Moral Foundations Theory, positions feelings as central to many of its foundations. Therefore, in the present paper, a data set of moral valence is proposed, consisting of 500 annotations by six human participants for both action/judgement and consequence moral valence, ranging from -1 to 1 for text-presented scenarios from the Commonsense Norm Bank data set. The resulting valence features share significant relationships with multi-class (immoral/discretionary/moral) and binary immoral/moral categories while additionally providing a noteworthy test set Matthew's correlation coefficient of 0.764 using regularised logistic regression for binary classification. This provides early evidence of the usefulness of valence features for morality estimation of text, indicating that valenced consequences of responses for others can be considered toward more human morally-aligned AI. In the interest of promoting further affective-moral computing research, this study's annotations will be made available for research on request.

Chinese Translation

目前的人工智能（AI）伦理实施并未充分考虑情感或情绪。如果AI应与人类伦理相一致，那么彻底研究AI行为是否能反映出有德行的人类伦理行为似乎是合理的，其中情感在个人的行为、判断或陈述中发挥着作用。此外，尽管著名的规范伦理理论常常被讨论其差异和不足，但美德伦理、结果主义伦理和康德义务论伦理都在某种程度上共同考虑了人类情感，而流行的描述性伦理理论——道德基础理论则将情感视为其许多基础的核心。因此，在本研究中，提出了一个道德情感数据集，包含六名参与者对行动/判断和结果道德情感的500个注释，情感值范围从-1到1，基于来自常识规范库（Commonsense Norm Bank）数据集的文本呈现场景。所得的情感特征与多类别（不道德/自由裁量/道德）和二元不道德/道德类别之间存在显著关系，同时在二元分类中使用正则化逻辑回归提供了显著的测试集马修相关系数为0.764。这为情感特征在文本道德估计中的有效性提供了早期证据，表明对他人反应的情感后果可以被视为朝向更符合人类道德的AI。为了促进进一步的情感-道德计算研究，本研究的注释将在请求时提供给研究者。

View on arXiv Download PDF AI Translation

cs.CL / 37 / 2607.20589

Evaluating the Effectiveness of Persona Simulation in Opinion Prediction with GPT-4.1

评估基于 GPT-4.1 的角色模拟在观点预测中的有效性

Li, Sarah Y., Yao, Ziyu

Abstract

Persona simulation involves utilizing large language models (LLMs) to anticipate human choices or interactions based on specific characteristic information. To further understand current limitations and future directions, we tested persona simulation in opinion prediction with GPT-4.1 (knowledge cutoff by June 2024). Using personas from nine U.S. states provided by Columbia University's Personas dataset, GPT-4.1 accurately predicted 2024 election outcomes in eight out of the nine states, only failing in one of the swing states. We then focused on opinions related to medicine and healthcare. With the American Trends Panel Wave 123 dataset from Pew Research Center, GPT-4.1 was able to anticipate beliefs about childhood vaccines with an accuracy of up to 0.94. Furthermore, we applied GPT-4.1 to generate conversations among personas and observed that the simulated dialogues and opinions adhered well to personas' personalities and backgrounds, albeit lacking natural human-like flow. Persona simulation proves to be a promising application of artificial intelligence as long as biases are addressed. In the near future, it will be beneficial to apply it to opinion analysis and reaction prediction in diverse fields ranging from public health to lawmaking to economics.

Chinese Translation

角色模拟涉及利用大型语言模型（LLMs）根据特定特征信息预测人类选择或互动。为了进一步理解当前的局限性和未来的方向，我们在观点预测中测试了基于 GPT-4.1 的角色模拟（知识截止于2024年6月）。使用哥伦比亚大学的角色数据集提供的来自美国九个州的角色，GPT-4.1 在九个州中准确预测了2024年选举结果，只有在一个摇摆州中失败。随后，我们关注与医学和医疗保健相关的观点。利用皮尤研究中心的美国趋势面板第123波数据集，GPT-4.1 能够以高达0.94的准确率预测关于儿童疫苗的信念。此外，我们应用 GPT-4.1 生成角色之间的对话，并观察到模拟的对话和观点与角色的个性和背景相符，尽管缺乏自然的人类流畅性。角色模拟被证明是人工智能的一个有前景的应用，只要能够解决偏见问题。在不久的将来，将其应用于公共卫生、立法到经济等多个领域的观点分析和反应预测将是有益的。

View on arXiv Download PDF AI Translation

cs.CL / 38 / 2607.20645

Frontier Financial Judgement: Can agents tell what might move a stock?

前沿金融判断：代理能否判断什么可能影响股票？

Harris, Joshua

Abstract

We introduce Frontier Financial Judgement, a challenging new benchmark developed in collaboration with professional equity analysts to assess agents' ability to replicate expert human judgements. Rapidly identifying new information, evaluating its implications and determining its valuation impact is one of the most time-consuming and challenging aspects of real-world equity coverage. This is becoming ever more difficult and important as AI rapidly increases the quantity of new information to process. The strongest agent we evaluate on Frontier Financial Judgement matches all expert labels in only 52.4% of cases. We also find significant divergence in estimated false-positive rates among frontier agents, ranging from ~1% for GPT-5.6 Sol to ~32% for Claude Sonnet 4.6. To construct the benchmark and make it representative of real-world settings, we combine human-designed and labelled synthetic articles with live news articles and historical documents, creating 656 items for assessment. The resulting task requires agents to distinguish genuinely new, valuation-relevant financial information from stale, immaterial or misleading news under realistic conditions. We find substantial trade-offs among agent accuracy, cost, false positives and reliability that continue to hinder the reliable deployment of news-flow filtering in practice.

Chinese Translation

我们引入了前沿金融判断，这是一个与专业股票分析师合作开发的具有挑战性的基准，用于评估代理复制专家人类判断的能力。快速识别新信息、评估其影响并确定其估值影响是现实世界股票覆盖中最耗时且最具挑战性的方面之一。随着人工智能迅速增加需要处理的新信息量，这一过程变得愈加困难且重要。我们在前沿金融判断中评估的最强代理在仅52.4%的情况下匹配所有专家标签。我们还发现，前沿代理在估计假阳性率方面存在显著差异，范围从GPT-5.6 Sol的约1%到Claude Sonnet 4.6的约32%。为了构建这一基准并使其具有代表性，我们结合了人类设计和标记的合成文章与实时新闻文章和历史文献，共创建了656个评估项目。最终任务要求代理在现实条件下区分真正的新、与估值相关的金融信息与陈旧、无关或误导性的新闻。我们发现代理的准确性、成本、假阳性和可靠性之间存在显著的权衡，这继续阻碍了新闻流过滤在实践中的可靠部署。

View on arXiv Download PDF AI Translation

cs.CL / 39 / 2607.20668

From Agent Failures to Text Policies: What Works and What Breaks

从代理失败到文本策略：有效与失效的因素

Ray, Jaideep, Goyal, Ankit

Abstract

TextGrad improves language-model systems by revising text from feedback. Its core thesis is that natural-language feedback can act as a gradient for optimizing text components without changing model weights. Applying it to agents is harder because feedback arrives only after a sequence of actions, making it difficult to identify which decision caused failure. We study this problem by separating the ability to follow a useful policy from the ability to learn that policy from experience. Our main finding is a clear gap between these two abilities. Human-written policies improve two frozen 7B agents on TextWorldExpress by 5.0 success points, showing that useful policy text exists. However, policies generated from agent trajectories do not reliably outperform fixed prompting, even with richer traces, counterfactual evidence, or iterative GEPA search. The main challenge for agent-level TextGrad is therefore not executing textual policy updates, but reliably generating and selecting them from experience.

Chinese Translation

TextGrad 通过反馈修订文本，从而改善语言模型系统。其核心论点是，自然语言反馈可以作为优化文本组件的梯度，而无需改变模型权重。将其应用于代理更为困难，因为反馈仅在一系列动作之后到达，这使得识别导致失败的决策变得困难。我们通过将遵循有用策略的能力与从经验中学习该策略的能力分开来研究这个问题。我们的主要发现是这两种能力之间存在明显的差距。人类编写的策略使两个冻结的 7B 代理在 TextWorldExpress 上的成功率提高了 5.0 个点，表明有用的策略文本是存在的。然而，从代理轨迹生成的策略并未可靠地超越固定提示，即使在更丰富的轨迹、反事实证据或迭代 GEPA 搜索的情况下也是如此。因此，代理级 TextGrad 的主要挑战不在于执行文本策略更新，而在于可靠地从经验中生成和选择这些策略。

View on arXiv Download PDF AI Translation

cs.CL / 40 / 2607.20690

Learning to Detect UI Principle Violations via Reinforcement Learning

通过强化学习学习检测用户界面原则违规行为

Mehta, Nishi, Alse, Swathi, Kumavat, Himani, Yu, Yue, Jayarao, Pratik

Abstract

Small language models and coding agents increasingly generate web front-end code, yet their outputs are typically evaluated primarily for functional correctness. A generated interface may compile, render, and pass unit tests while still violating established interface quality principles, including accessibility barriers, deceptive design patterns, poor visual hierarchy, and excessive decision complexity. Existing auditing approaches face a trade-off between cost, coverage, and scalability: expert human review provides rich judgment but is slow and expensive; frontier vision-language models offer broader reasoning capabilities but remain costly to deploy at scale; and rule-based tools such as axe-core and Lighthouse are inexpensive but primarily capture mechanically checkable accessibility issues. We investigate whether a lightweight vision-language model can serve as an effective critic for generated interfaces. We unify 19 interface-quality principles from three complementary sources of HCI knowledge: WCAG 2.2 accessibility standards, deceptive design taxonomies, and established theories of perception, cognition, and interaction. To train this critic, we construct a verified dataset of approximately 10,000 generated web pages by synthetically injecting known violations into clean, LLM-generated Tailwind pages. Continued reinforcement learning on a 4B vision-language model improves micro-F1 from 36\% to 84\%, with 13 of 19 principles exceeding 80\% F1. The resulting critic can audit generated interfaces, filter low-quality interface training data, and provide a reward signal for design-aware code generation. We release our data-generation recipe and injection/verification prompts to support reproducible evaluation and future work on scalable interface-quality assessment.

Chinese Translation

小型语言模型和编码代理越来越多地生成网页前端代码，但其输出通常主要评估功能正确性。生成的界面可能编译、渲染并通过单元测试，但仍然违反既定的界面质量原则，包括可访问性障碍、误导性设计模式、糟糕的视觉层次结构和过度的决策复杂性。现有的审计方法在成本、覆盖范围和可扩展性之间面临权衡：专家人工审查提供丰富的判断，但速度慢且成本高；前沿的视觉-语言模型提供更广泛的推理能力，但在大规模部署时仍然成本高昂；而基于规则的工具如 axe-core 和 Lighthouse 成本低廉，但主要捕捉机械可检查的可访问性问题。我们研究了一种轻量级视觉-语言模型是否可以作为生成界面的有效批评者。我们整合了来自三个互补的HCI知识来源的19个界面质量原则：WCAG 2.2可访问性标准、误导性设计分类法以及已建立的感知、认知和交互理论。为了训练这个批评者，我们构建了一个经过验证的数据集，包含大约10,000个生成的网页，通过在干净的、LLM生成的Tailwind页面中合成注入已知的违规行为。对一个4B视觉-语言模型的持续强化学习将微F1从36%提高到84%，其中19个原则中的13个超过80%的F1。最终的批评者可以审计生成的界面，过滤低质量的界面训练数据，并为设计感知的代码生成提供奖励信号。我们发布了我们的数据生成配方和注入/验证提示，以支持可重复的评估和未来可扩展的界面质量评估工作。

View on arXiv Download PDF AI Translation

cs.CL / 41 / 2607.20722

REGARD: Regional Affective Differences in Large Language Models

REGARD：大型语言模型中的区域情感差异

Chetvergov, Andrei, Evseev, Alexander, Solovev, Mikhail, Sivoraksha, Timofei, Ukolov, Stepan, Kuschenko, Valeriia, Chistyakova, Maria, Bolovtsov, Sergey

Abstract

Large language models trained and aligned within different linguistic and regional ecosystems may frame the same political, cultural, and geopolitical entities in different ways. Such differences are often evaluated through sentiment, favorability, or stance, reducing model attitudes to a single positive-negative axis. We introduce REGARD, a study of what drives affective framing differences across LLMs on post-Soviet entities using target-directed Valence-Arousal-Dominance profiling. We query 19 models on 500 region-specific targets, score their responses with two independent LLM judges, GPT-4o-mini and Qwen3.6-35B-A3B, and validate the measurements on a 300-item human-annotated subset. Post-hoc Ward-linkage clustering of all 19 models by affective and response-behavior profiles yields three behavioral clusters that cut across model origin, family, and parameter count. Generic-answer rate is strongly associated with lower arousal (r = -0.81) and with cluster placement: models that deflect evaluative prompts with templated responses cluster together at low arousal regardless of origin. These findings show that VAD profiling captures emotional intensity, a dimension of affective framing that is largely invisible to conventional sentiment-based evaluation.

Chinese Translation

在不同语言和区域生态系统中训练和对齐的大型语言模型可能以不同的方式框定相同的政治、文化和地缘政治实体。这种差异通常通过情感、好感度或立场进行评估，将模型态度简化为单一的正负轴。我们引入了REGARD，一个研究驱动后苏联实体在大型语言模型（LLMs）之间情感框架差异的因素，采用目标导向的效价-唤醒-主导性（Valence-Arousal-Dominance, VAD）分析。我们对19个模型在500个特定区域目标上进行查询，使用两个独立的LLM评审员，GPT-4o-mini和Qwen3.6-35B-A3B，对其响应进行评分，并在一个300项人类标注的子集上验证测量结果。对所有19个模型进行后验的Ward链接聚类，基于情感和响应行为特征，得出了三个行为聚类，这些聚类跨越模型的来源、家族和参数数量。通用回答率与较低的唤醒水平（r = -0.81）以及聚类位置有很强的关联：无论来源如何，使用模板化响应来偏转评估提示的模型在低唤醒水平下聚集在一起。这些发现表明，VAD分析捕捉了情感强度，这是情感框架的一个维度，而这一维度在传统的基于情感的评估中往往是不可见的。

View on arXiv Download PDF AI Translation

cs.CL / 42 / 2607.20767

Rushes: A Human Preference Dataset for Pluralistic Alignment

Rushes：用于多元对齐的人类偏好数据集

Xu, Michael, Leandro, Jorge, Rao, Sudha, Xu, Weijia, Jojic, Nebojsa, DesGarennes, Gabriel, Quirk, Chris, Dolan, Bill

Abstract

We introduce Rushes, a dataset and benchmark for studying revealed human engagement preferences in interactive narrative environments. Rushes is collected through a game interface where users interact with AI-generated branching narratives and select one choice from a small, explicit candidate set at each decision point. Each interaction logs the full candidate set, the user's choice, and the evolving narrative context, yielding time-ordered trajectories with persistent user-level identifiers. Rushes contains 44,226 decision events from 8,167 unique users across six games, capturing sequential, personalized engagement behavior rather than static judgments. We show that user choices exhibit structured, non-random patterns, quantified by a low choice entropy relative to a uniform baseline. We position Rushes as a diagnostic benchmark for pluralistic alignment and demonstrate a robust Engagement Gap: state-of-the-art LLMs, including GPT-5, fail to outperform simple baselines. While classical Matrix Factorization (SVD) captures measurable personalized signal (37.7%), frontier LLMs (34.23%) struggle to even match the Popularity Baseline (36.4%) on event-level choice prediction. This gap suggests that single, population-level objectives, like those used in modern RLHF, appear insufficient to capture heterogeneous, context-dependent engagement signals. As a result, even highly capable models default to majority preferences rather than adapting to individual trajectories. We release Rushes to support research into pluralistic alignment and sequential decision-making in generative systems. The full code for the platform and dataset will be available here: https://github.com/microsoft/rushes

Chinese Translation

我们介绍了Rushes，一个用于研究互动叙事环境中人类参与偏好的数据集和基准。Rushes通过一个游戏界面收集，用户在其中与AI生成的分支叙事进行互动，并在每个决策点从一个小的、明确的候选集选择一个选项。每次互动记录了完整的候选集、用户的选择以及不断变化的叙事上下文，从而生成具有持久用户级标识符的时间顺序轨迹。Rushes包含来自8167个独特用户在六款游戏中产生的44226个决策事件，捕捉了顺序的、个性化的参与行为，而非静态的判断。我们展示了用户选择表现出结构化的、非随机的模式，其选择熵相对于均匀基线较低。我们将Rushes定位为多元对齐的诊断基准，并展示了一个显著的参与差距：包括GPT-5在内的最先进的大型语言模型（LLMs）未能超越简单基线。尽管经典的矩阵分解（SVD）能够捕捉可测量的个性化信号（37.7%），但前沿的LLMs（34.23%）甚至难以与流行基线（36.4%）在事件级选择预测上相匹配。这个差距表明，像现代强化学习人类反馈（RLHF）中使用的单一、人口级目标似乎不足以捕捉异质的、依赖上下文的参与信号。因此，即使是高度能力的模型也倾向于默认选择多数偏好，而不是适应个体轨迹。我们发布Rushes以支持对多元对齐和生成系统中顺序决策的研究。平台和数据集的完整代码将可在此处获取：https://github.com/microsoft/rushes

View on arXiv Download PDF AI Translation

cs.CL / 43 / 2607.20768

Are Diversity Metrics Measuring Diversity? A Capability-Controlled Audit of Majority-Vote Gain in LLM Ensembles

多样性指标是否在测量多样性？对大规模语言模型集成中多数投票增益的能力控制审计

Kim, Donghwan

Abstract

Majority voting over LLMs is widely assumed to benefit from diversity, and diversity measures are used to choose which models to combine. We ask whether five such measures track diversity or mainly re-express capability, auditing them as predictors of majority-vote gain over the best member across 31,900 subsets of 30 LLMs on MMLU-Pro (29 on TruthfulQA) under explicit capability controls. Three findings emerge. First, latent complementarity is ubiquitous: oracle gain is positive in 100% of subsets, yet simple voting beats the strongest member in only 9.98% of all canonical size-3 subsets (18.71% with held-out best selection); the pooled size-2-4 rate is 1.27%, partly reflecting deterministic even-size voting behavior. Second, a joint-correctness proxy (strict diversity) is nearly collinear with one minus mean accuracy (size-3 Spearman rho = +0.991 / +0.988); raw diversity-gain associations are strongly capability-entangled and, with one exception, unstable under control. Third, three linear contingency-table statistics are algebraically non-separable; after capability control, the empirically stable remainder is a modest residual pairwise co-failure association in which more shared error corresponds to lower gain. This direction is robust, but its magnitude is configuration-dependent. Joint rawspace linear regressions treating strict diversity, disagreement, and double-fault as independent predictors are rank-deficient by construction.

Chinese Translation

对大规模语言模型（LLMs）进行多数投票普遍被认为受益于多样性，因此多样性度量被用于选择组合哪些模型。我们探讨五种此类度量是否真正反映多样性，还是主要重新表达能力，并对其在31,900个30个LLM子集上对最佳成员的多数投票增益进行审计，使用MMLU-Pro（在TruthfulQA上为29个）并在明确的能力控制下进行。研究得出三个发现。首先，潜在的互补性无处不在：在100%的子集中，oracle增益为正，但简单投票在所有标准大小为3的子集中仅在9.98%的情况下超过最强成员（在保留最佳选择的情况下为18.71%）；合并的大小为2-4的比率为1.27%，部分反映了确定性的偶数大小投票行为。其次，一个联合正确性代理（严格多样性）与一个减去平均准确率几乎共线（大小为3的Spearman rho = +0.991 / +0.988）；原始多样性增益的关联强烈与能力纠缠，并且在控制下不稳定，只有一个例外。第三，三个线性列联表统计量在代数上是不可分离的；在能力控制后，经验上稳定的剩余是一个适度的成对共失效关联，其中更多的共享错误对应于较低的增益。这个方向是稳健的，但其幅度依赖于配置。将严格多样性、分歧和双重失误视为独立预测因子的联合原始空间线性回归在构造上是秩缺乏的。

View on arXiv Download PDF AI Translation

cs.CL / 44 / 2607.20803

The Geometry of Personality: Activation Steering with Jungian Cognitive Functions

人格的几何学：基于荣格认知功能的激活引导

Zai, Liu, Wang, Yumeng, Fu, Junchen, Jose, Joemon M.

Abstract

Activation steering enables control and interpretation of LLMs, yet existing work primarily models personality through static trait frameworks such as the Big Five. We investigate whether personality can instead be represented and controlled as a set of cognitive processes using the eight Jungian Cognitive Functions. To this end, we introduce a framework comprising a Jungian evaluation protocol and a dataset of over 2,100 role-playing character narrations. Activation steering vector extraction and evaluation experiments on Llama-3.1-8B demonstrate effective monotonic control over all eight cognitive functions through activation steering. Beyond controllability, our analysis reveals that: 1. personality information is concentrated in middle transformer layers; 2. steering vectors exhibit structured geometric relationships consistent with distinctions between rational and irrational functions; 3. effective multi-dimensional steering directions cannot be recovered as linear combinations of single-function directions. These findings provide new insights into the representation of personality in LLM activation space and establish a framework for studying interpretable, effective, and multi-dimensional personality control.

Chinese Translation

激活引导使得对大型语言模型（LLMs）的控制和解释成为可能，但现有研究主要通过静态特质框架（如五大人格理论）来建模人格。我们探讨人格是否可以作为一组认知过程来表示和控制，使用八种荣格认知功能。为此，我们引入了一个框架，包括荣格评估协议和一个包含超过2100个角色扮演角色叙述的数据集。在 Llama-3.1-8B 上进行的激活引导向量提取和评估实验表明，通过激活引导可以有效地对所有八种认知功能进行单调控制。除了可控性，我们的分析还揭示了：1. 人格信息集中在中间变换层；2. 引导向量表现出与理性和非理性功能之间的区别一致的结构几何关系；3. 有效的多维引导方向无法作为单一功能方向的线性组合恢复。这些发现为理解 LLM 激活空间中的人格表示提供了新的见解，并建立了一个研究可解释、有效和多维人格控制的框架。

View on arXiv Download PDF AI Translation

cs.CL / 45 / 2607.20833

REFACT: Adaptive Fact Restatement for Compact and Faithful Chain-of-Thought Reasoning

REFACT：适应性事实重述框架用于紧凑且真实的思维链推理

Jin, Zhensheng, Dai, Xin, Liu, Zhenghao, Xiao, Chaojun, Xie, Huiyuan, Gu, Yu, Yu, Ge, Sun, Maosong

Abstract

Large language models increasingly rely on long-form reasoning for complex tasks, yet their reasoning traces may drift away from the supplied context when evidence is sparse, noisy, or in conflict with parametric knowledge. Existing grounding methods either attach citations after generation or encourage evidence retrieval inside the trace, but they often do not ensure that cited content is sufficient for the local inference and final answer. We propose REFACT, an adaptive fact-restatement citation framework that trains models to decide when a reasoning step needs contextual grounding and at what granularity source facts should be restated. This design avoids both unsupported inference and indiscriminate fact copying by turning citations into answer-supporting intermediate states. REFACT is optimized with a two-stage SFT-to-RL pipeline in which a citation-utility reward encourages cited facts to be well-formed, source-traceable, and answer-sufficient. Experiments on LongBench, LV-Eval, and ConFiQA show that REFACT improves long-context QA and counterfactual faithfulness while substantially reducing token consumption. Further analysis shows that REFACT preserves more answer-bearing evidence with fewer restated facts, yielding reasoning traces that are denser rather than longer. All code and data are available at https://github.com/NEUIR/REFACT.

Chinese Translation

大型语言模型在复杂任务中越来越依赖于长形式推理，然而当证据稀少、噪声较大或与参数知识相冲突时，它们的推理轨迹可能会偏离提供的上下文。现有的基础方法要么在生成后附加引用，要么鼓励在推理轨迹中检索证据，但它们往往无法确保引用内容足以支持局部推理和最终答案。我们提出了REFACT，一个适应性事实重述引用框架，训练模型决定何时需要上下文基础以及源事实应以何种粒度进行重述。该设计通过将引用转化为支持答案的中间状态，避免了不支持的推理和无差别的事实复制。REFACT通过一个两阶段的SFT到RL管道进行优化，其中引用效用奖励鼓励引用的事实具有良好的形式、可追溯的来源和足够的答案支持。在LongBench、LV-Eval和ConFiQA上的实验表明，REFACT改善了长上下文问答和反事实忠实性，同时显著减少了令牌消耗。进一步分析显示，REFACT在重述事实较少的情况下保留了更多的答案相关证据，从而产生了更密集而非更长的推理轨迹。所有代码和数据可在https://github.com/NEUIR/REFACT获取。

View on arXiv Download PDF AI Translation

cs.CL / 46 / 2607.20862

CSPF: A Constrained Shared-Private Fusion Method for Non-Verifiable Preference Evaluation

CSPF：一种用于不可验证偏好评估的约束共享-私人融合方法

Zhang, Hehao, Wang, Danli, Wang, Xinyuan, Gao, Xuange

Abstract

At present, reliable evaluation of non-verifiable tasks remains challenging. Existing approaches often fail to adequately capture the diverse evaluative criteria underlying human preferences in such tasks. To this end, we propose Constrained Shared-Private Fusion (CSPF), a fusion method that treats heterogeneous frozen reward models as complementary evaluators and learns to integrate their hidden-state representations under pairwise human-preference supervision. CSPF decomposes each expert signal into shared and expert-private representations, encouraging cross-expert alignment while preserving complementary viewpoints. Across experiments on LM-Arena target-domain adaptation and PPE out-of-distribution preference evaluation, CSPF achieves the best performance on the primary metrics among the evaluated single-expert reward-model, scalar-score multi-expert, and rubric-judge baselines. Overall, CSPF suggests that fusing hidden-state representations provides a more expressive basis for preference assessment, offering a practical route toward integrated evaluative signals for non-verifiable preference tasks.

Chinese Translation

目前，对不可验证任务的可靠评估仍然面临挑战。现有方法往往无法充分捕捉人类偏好背后多样的评估标准。为此，我们提出了约束共享-私人融合（CSPF），这是一种融合方法，将异构的冻结奖励模型视为互补评估者，并在成对的人类偏好监督下学习整合它们的隐藏状态表示。CSPF将每个专家信号分解为共享和专家私人表示，鼓励跨专家对齐，同时保留互补视角。在LM-Arena目标领域适应和PPE分布外偏好评估的实验中，CSPF在评估的单专家奖励模型、标量分数多专家和评分标准基线中，在主要指标上取得了最佳性能。总体而言，CSPF表明融合隐藏状态表示为偏好评估提供了更具表现力的基础，为不可验证偏好任务提供了集成评估信号的实际途径。

View on arXiv Download PDF AI Translation

cs.CL / 47 / 2607.20872

LegalCiteTrust: Benchmarking Citation Trustworthiness in Chinese Long-Form Legal Research Reports

LegalCiteTrust：评估中国长篇法律研究报告中引用可信度的基准

Li, Yunhan, Xie, Mingjie, Shi, Zeyang, Wu, Gengshen, Yang, Min

Abstract

Long-form legal research reports increasingly rely on LLMs and agentic research systems, but their reliability depends not only on answering the task, but also on whether cited legal authorities are trustworthy. A citation can be risky even when it points to a real source: the report may omit limiting conditions, misdescribe the authority, or use it to support a stronger claim than the source allows. We introduce LegalCiteTrust, a benchmark for evaluating citation trustworthiness in Chinese long-form legal research reports. It contains 72 densely annotated report-level tasks and evaluates reports along three dimensions: Coverage, Support, and Citation Trustworthiness. Citation Trustworthiness is operationalized through citation-level Existence, Fidelity, and Applicability (E/F/A). Experiments on general-purpose LLMs, deep-research systems, and legal-specific systems show that task completion, evidence richness, citation density, and citation reliability expose different system behaviors. Retrieval tools can improve evidence support without reliably improving the Trust score, while E/F/A-based revision improves Trust and Final score more clearly than existence-only filtering. These results suggest that trustworthy legal research generation requires citation-aware evidence governance after retrieval: systems must not only retrieve legal authorities, but also select, describe, and apply them reliably.

Chinese Translation

长篇法律研究报告越来越依赖于大型语言模型（LLMs）和自主研究系统，但其可靠性不仅取决于任务的完成，还取决于所引用的法律权威是否可信。即使引用指向真实来源，引用也可能存在风险：报告可能省略限制条件、错误描述权威，或用其支持超出来源允许的更强主张。我们介绍了LegalCiteTrust，这是一个用于评估中国长篇法律研究报告中引用可信度的基准。该基准包含72个密集注释的报告级任务，并从三个维度评估报告：覆盖率、支持度和引用可信度。引用可信度通过引用级别的存在性、忠实性和适用性（E/F/A）进行操作化。在通用大型语言模型、深度研究系统和法律特定系统上的实验表明，任务完成度、证据丰富性、引用密度和引用可靠性揭示了不同系统的行为。检索工具可以在不可靠地提高可信度评分的情况下改善证据支持，而基于E/F/A的修订比仅存在性过滤更清晰地改善了可信度和最终评分。这些结果表明，可信的法律研究生成需要在检索后进行引用意识的证据治理：系统不仅必须检索法律权威，还必须可靠地选择、描述和应用它们。

View on arXiv Download PDF AI Translation

cs.CL / 48 / 2607.20911

Tencent WorkBuddy Bench: A Multi-Domain Coding-Agent Benchmark with Contamination-Resistant Task Construction

腾讯 WorkBuddy 基准：一个具有抗污染任务构建的多领域编码代理基准

Tencent WorkBuddy Bench Team, Cai, Siqi, Chen, Shaopeng, Fei, Xiang, Mao, Yong, Xu, Zihan, Lyu, Zhiheng, Shao, Zhijian, Shi, Yuchen, Zhang, Shuwen, Qiu, Chaofan, Che, Linjie, Zhao, Xiaoxi, Wu, Feng, Zhang, Kai, Zhu, Chaofan, Qi, Yubin, Liang, Xiaoyun, Dong, Peijie, Zhang, Yunhao, Zhu, Yuanjie, Jiang, Ling, Zhang, Xianjun, Chu, Zhehang, Sang, Anyuan, Feng, Zhen, Nie, Sen, Wu, Shi, Xu, Yuanzhen, Li, Xin, Yang, Ning, Dong, Zhiqiang, Dong, Hande, Lin, Qiang, Liu, Yi, Wu, Yunsheng, Li, Ke, Sun, Xing

Abstract

We introduce Tencent WorkBuddy Bench, a multi-domain evaluation suite for coding agents; this report documents its construction methodology, scoring protocol, and a cross-model leaderboard. At its core is a unified evaluation framework for constructing and running distribution-informed coding-agent tasks across four work domains - Code, Web, Office, and Security. Rather than adapting public issue text, every task is reverse-engineered from a real commit, pull request, or business scenario and rewritten as a short, colloquial, role-played request, so that a task's prompt is not recoverable by web-searching the underlying issue, pull request, or commit thread. Because the dataset is released openly - task directories, environment images, evaluation harness, tests, and reference solutions - contamination resistance rests on this construction together with dataset versioning rather than on secrecy. The four subsets - repository-level engineering, front-end development, office and business workflows, and red-/blue-team security - probe complementary facets of real work, each with its own verification style. All are packaged in a uniform task-directory format and run, under a uniform and reproducible protocol, on two agent harnesses (CodeBuddy Code and Claude Code); the full open release makes the benchmark reproducible end to end and directly auditable, since any third party can re-run each task and inspect its content. Because each subset uses a different scoring instrument, scores are not comparable across subsets and the suite reports no suite-wide average. We report a cross-model leaderboard across several model families.

Chinese Translation

我们介绍了腾讯 WorkBuddy 基准，这是一个用于编码代理的多领域评估套件；本报告记录了其构建方法、评分协议和跨模型排行榜。其核心是一个统一的评估框架，用于构建和运行基于分布信息的编码代理任务，涵盖四个工作领域——代码、网络、办公室和安全。与其说是改编公共问题文本，不如说每个任务都是从真实的提交、拉取请求或业务场景中逆向工程而来，并重写为简短、口语化的角色扮演请求，以确保任务提示无法通过网络搜索获取底层问题、拉取请求或提交线程。由于数据集是公开发布的——任务目录、环境图像、评估工具、测试和参考解决方案——抗污染性依赖于这种构建以及数据集版本控制，而不是依赖于保密。四个子集——仓库级工程、前端开发、办公室和商业工作流程，以及红队/蓝队安全——探讨了真实工作互补的不同方面，每个子集都有其独特的验证风格。所有内容都以统一的任务目录格式打包，并在两个代理工具（CodeBuddy Code 和 Claude Code）上按照统一且可重复的协议运行；完整的开放发布使得基准可以端到端可重复且直接可审计，因为任何第三方都可以重新运行每个任务并检查其内容。由于每个子集使用不同的评分工具，因此子集之间的分数不可比较，套件也不报告全套的平均值。我们报告了跨多个模型家族的跨模型排行榜。

View on arXiv Download PDF AI Translation

cs.CL / 49 / 2607.20946

From a Word-Level Dictionary to Sentence-Level Semantics: Multilingual Grievance Labelling with Contextual Models

从词级词典到句级语义：基于上下文模型的多语言投诉标注

Tian, Lin, Rizoiu, Marian-Andrei

Abstract

Grievance is one of the warning signs analysts look for when assessing threats of violence. It is increasingly measured at scale from online text, most often with word-level lexicons like the Grievance Dictionary that score by matching weighted terms. Such matching is a fast and transparent proxy, but it cannot resolve whether a term is asserted, quoted, negated, or condemned. These lexicons are also often evaluated on pools enriched with the very examples they retrieve, so a high score partly reflects agreement with the lexicon's own selection rule. Examining a five-language, 2{,}000-item evaluation pool, we find its halves separated almost perfectly by the lexicon itself: every item labeled ``random'' is in fact lexicon-negative, so the lexicon's apparent macro-AUROC of 0.686 collapses to a 0.500 floor fixed by construction. We keep the dictionary's 22-construct ontology but replace term matching with context-reading models, evaluated on a non-circular benchmark that separates unconditional-random, lexicon-positive, and lexicon-negative strata across five languages. Reading the full post rather than the target sentence alone helps most where the lexicon is silent, raising average precision on lexicon-negative text from 0.14 to 0.20, with the largest gains on quoted, implicit, and cross-sentence grievance. Together, these results show that grievance is measured more faithfully by reading the surrounding context, and more honestly when tested on text the lexicon did not select. We release our code and benchmark at https://github.com/behavioral-ds/multilingual_grievance.

Chinese Translation

投诉是分析师在评估暴力威胁时寻找的警示信号之一。近年来，投诉的测量越来越多地依赖于在线文本，通常使用诸如投诉词典（Grievance Dictionary）等词级词典，通过匹配加权术语进行评分。这种匹配是一种快速且透明的代理方法，但无法判断一个术语是被断言、引用、否定还是谴责。这些词典通常在包含其检索示例的池中进行评估，因此高分在一定程度上反映了与词典自身选择规则的一致性。在对一个包含五种语言、2000个项目的评估池进行分析时，我们发现其两半几乎完全被词典本身分开：每个标记为“随机”的项目实际上都是词典负面的，因此词典表面上的宏观AUROC（Area Under the Receiver Operating Characteristic）为0.686的结果在构造上崩溃至0.500的底线。我们保留词典的22个构造本体，但用上下文阅读模型替代术语匹配，并在一个非循环基准上进行评估，该基准在五种语言中区分无条件随机、词典正面和词典负面层次。阅读完整的帖子而不仅仅是目标句子在词典沉默的情况下帮助最大，提升了词典负面文本的平均精度，从0.14提高到0.20，尤其在引用、隐含和跨句投诉上获得了最大的提升。综合来看，这些结果表明，通过阅读周围的上下文，投诉的测量更加真实，而在对未被词典选择的文本进行测试时则更加诚实。我们在 https://github.com/behavioral-ds/multilingual_grievance 发布了我们的代码和基准。

View on arXiv Download PDF AI Translation

cs.CL / 50 / 2607.20995

Where Animacy Lives in Large Language Models: Tracing the Circuits of the Animacy Concept

大型语言模型中的生动性概念：追踪生动性概念的电路

Punzo, Samuele, Cinà, Giovanni, Pezzelle, Sandro

Abstract

Distinguishing animate from inanimate concepts in written language requires more than shallow text processing, as it involves recognizing complex selectional constraints and contextual cues, such as verb-argument interactions. Yet, current large language models (LLMs) appear to be capable of doing it. We investigate whether this animacy-sensitive behavior of LLMs can be traced to a localized set of causally relevant components and connections. To do so, we construct a controlled dataset of minimal pairs and perform circuit discovery on four open-weight models. Through in-depth experiments and ablations, we show that a causal mechanism responsible for handling animacy in these models does exist, thus discovering an animacy circuit. At the same time, this circuit appears to be less localized compared to other known ones and generalizes only partially across models and animacy tasks, confirming the distributed, context-dependent, and somewhat graded nature of the animacy concept.

Chinese Translation

在书面语言中区分生动与非生动概念不仅需要浅层文本处理，还涉及识别复杂的选择约束和上下文线索，例如动词-论元交互。然而，目前的大型语言模型（LLMs）似乎能够做到这一点。我们研究了LLMs的这种生动性敏感行为是否可以追溯到一组局部的因果相关组件和连接。为此，我们构建了一个最小对的控制数据集，并对四个开放权重模型进行了电路发现。通过深入的实验和消融研究，我们表明，负责处理这些模型中生动性的因果机制确实存在，从而发现了一个生动性电路。同时，这个电路似乎与其他已知电路相比局部化程度较低，并且在模型和生动性任务之间仅部分泛化，确认了生动性概念的分布式、依赖上下文和某种程度的分级特性。

View on arXiv Download PDF AI Translation

cs.CL / 51 / 2607.21016

CultureTalk-ID: A Multi-Task Dialogue Benchmark for Cultural Commonsense in Indonesian Local Languages

CultureTalk-ID：印尼地方语言文化常识的多任务对话基准

Kautsar, Muhammad Dehan Al, Pranida, Salsabila, Elbouardi, Bilal, Koto, Fajri

Abstract

Culture is lived through conversation, yet existing Indonesian cultural commonsense benchmarks evaluate LLMs on short and isolated prompts, stripping away the dialogic context in which cultural nuances actually surface. We introduce CultureTalk-ID, the first dialogue-based benchmark for cultural commonsense in Indonesian and its local languages, comprising 4,496 culturally grounded dialogues across 11 languages and 13 culturally salient topics, curated through a multi-stage human pipeline with native speakers to ensure authenticity. CultureTalk-ID introduces three complementary tasks, namely dialogue-based multiple-choice cultural commonsense reasoning, culturally faithful machine translation, and language steering, which jointly probe whether LLMs can understand, transfer, and generate culturally grounded language.

Chinese Translation

文化通过对话得以体现，然而现有的印尼文化常识基准仅在短小且孤立的提示上评估大型语言模型（LLMs），剥离了文化细微差别实际显现的对话背景。我们推出了CultureTalk-ID，这是第一个基于对话的印尼及其地方语言的文化常识基准，包含4,496个基于文化的对话，涵盖11种语言和13个文化显著主题，经过多阶段的人类流程与母语者的精心策划，以确保其真实性。CultureTalk-ID引入了三个互补任务，即基于对话的多项选择文化常识推理、文化忠实的机器翻译和语言引导，这些任务共同探讨LLMs是否能够理解、转移和生成基于文化的语言。

View on arXiv Download PDF AI Translation

cs.CL / 52 / 2607.21051

Sample-Efficient Learning from Agent Experience

从智能体经验中高效学习

Gou, Chenhui, Tu, Haoqin, Fang, Yunhao, Cai, Jianfei, Rezatofighi, Hamid

Abstract

Real-world agent learning is often constrained by costly environment interactions, such as running time-consuming experiments or obtaining human feedback. In-context learning offers a highly sample-efficient way for agents to learn from their own interaction histories, but its gains disappear once that experience is removed from the context. Separately, context distillation provides a mechanism for internalizing contextual information into model weights. However, applying it to agents' interaction histories without sacrificing environment sample efficiency remains underexplored. We term this problem Experience Distillation and develop an implementation that requires no further environment interaction beyond the collected experience. Experiments on 749 curated software-engineering tasks and six text-adventure games show that it retains at least 64.8\% of the gains from in-context learning across both domains, whereas direct supervised fine-tuning on the collected experience recovers only 3.8\%. Compared with classical reinforcement-learning baselines, in-context learning from trial-and-error experience followed by Experience Distillation matches their performance with at least $9.6\times$ fewer environment samples.

Chinese Translation

现实世界中的智能体学习常常受到昂贵环境交互的限制，例如进行耗时的实验或获取人类反馈。上下文学习为智能体从自身交互历史中学习提供了一种高效的样本利用方式，但一旦将该经验从上下文中移除，其收益便会消失。另一方面，上下文蒸馏提供了一种将上下文信息内化为模型权重的机制。然而，在不牺牲环境样本效率的情况下，将其应用于智能体的交互历史仍然未被充分探索。我们将此问题称为经验蒸馏，并开发了一种实现方法，该方法在收集的经验之外不需要进一步的环境交互。在749个精心策划的软件工程任务和六个文本冒险游戏上的实验表明，它在这两个领域中至少保留了64.8 ext{%}的上下文学习收益，而对收集的经验进行直接监督微调仅恢复了3.8 ext{%}。与经典的强化学习基线相比，从试错经验中进行的上下文学习，随后进行经验蒸馏，其性能与基线相当，且环境样本数量至少减少了9.6倍。

View on arXiv Download PDF AI Translation

cs.CL / 53 / 2607.21063

QuantiBias: Benchmarking Quantization-Induced Bias in LLMs

QuantiBias：量化引起的偏见基准测试在大型语言模型中的应用

Ferrara, Emilio

Abstract

Almost every large language model that reaches a broad audience is quantized: trained in full precision, then compressed for efficiency. This step is assumed harmless and its safety is rarely re-checked. We find its principal side effect is increased bias that standard safety evaluation misses. Holding the model, its training, and the prompts fixed, a quantized model still refuses harmful requests, still avoids over-refusing benign prompts, and still selects the unbiased multiple-choice answer. Yet asked an open-ended question, the same model volunteers stereotypes in all eight languages we probe, in roughly one in four open-ended answers under an independent judge (~24% to ~27% across the compression ladder): it passes every standard check and still reaches users measurably more biased. The selective gap is a robust finding; whether open-ended bias further increases with compression is less certain, sensitive to the judge that scores it. We address both with \textbf{QuantiBias}, a benchmark that pairs a generative, multilingual stereotype probe with the refusal and multiple-choice controls that isolate open-ended generation, contrasts each build with and without reasoning, and rates the content severity of what it generates. Across two backbone models (Qwen and Gemma), a five-family screen, and eight benchmarks, quantizers allocate their extra precision by capability data that carries no bias-prevention signal, and reasoning before answering roughly halves the effect on some families while doing nothing on others. A quantized build must be re-evaluated for open-ended bias, not only on the short-form safeguards it already passes.

Chinese Translation

几乎所有面向广泛受众的大型语言模型都是经过量化的：先以全精度训练，然后为了提高效率进行压缩。这一步骤被认为是无害的，其安全性很少被重新检查。我们发现其主要副作用是增加的偏见，而标准安全评估往往忽视这一点。在固定模型、训练和提示的情况下，量化模型仍然拒绝有害请求，避免过度拒绝良性提示，并选择无偏见的多项选择答案。然而，当被问及开放式问题时，同一模型在我们调查的八种语言中都会主动提出刻板印象，在大约四分之一的开放式回答中出现（在独立评审者下约24%至27%）：它通过了每一个标准检查，但在可测量上却对用户更加偏见。选择性差距是一个稳健的发现；开放式偏见是否随着压缩进一步增加则不太确定，且对评分的评审者敏感。我们通过 extbf{QuantiBias}来解决这两个问题，该基准将生成的多语言刻板印象探测与拒绝和多项选择控制相结合，以隔离开放式生成，并对比有无推理的每个构建，评估其生成内容的严重性。在两个主干模型（Qwen和Gemma）、五个家族筛选和八个基准测试中，量化器通过不携带偏见预防信号的能力数据分配其额外精度，而在回答之前进行推理大约将某些家族的影响减半，而对其他家族则没有影响。量化构建必须重新评估开放式偏见，而不仅仅是基于其已经通过的短期保护措施。

View on arXiv Download PDF AI Translation

cs.CL / 54 / 2607.21067

PrefReward: Learning User Preference Matrix for Personalized Text Generation

PrefReward：用于个性化文本生成的用户偏好矩阵学习

Wu, Yue, Wang, Chengbing, Bai, Yimeng, Zhao, Xiaoyan, Zhang, Yang, Feng, Fuli

Abstract

Large Language Models (LLMs) have demonstrated remarkable ability in generating personalized content by leveraging user histories and contextual cues. However, most existing personalization approaches rely on implicit representations within model parameters, making it difficult to interpret user-specific preferences or effectively handle long-context dependencies. To address these challenges, we propose PrefReward, a novel preference-aware generative framework that explicitly models user styles through a structured preference matrix and integrates it into the decoding process as a reward signal. PrefReward consists of two stages: (1) extracting a user-specific preference matrix that summarizes individual stylistic tendencies, and (2) using the matrix to guide generation via a KL-divergence-based reward function. Experiments on the LongLaMP dataset show that PrefReward outperforms non-personalized and retrieval-based baselines in both generation quality and personalization interpretability.

Chinese Translation

大型语言模型（LLMs）在利用用户历史记录和上下文线索生成个性化内容方面表现出色。然而，大多数现有的个性化方法依赖于模型参数中的隐式表示，这使得难以解释用户特定的偏好或有效处理长上下文依赖关系。为了解决这些挑战，我们提出了PrefReward，一种新颖的偏好感知生成框架，它通过结构化的偏好矩阵显式建模用户风格，并将其作为奖励信号集成到解码过程中。PrefReward包括两个阶段：（1）提取用户特定的偏好矩阵，以总结个体的风格倾向；（2）使用该矩阵通过基于KL散度的奖励函数指导生成。在LongLaMP数据集上的实验表明，PrefReward在生成质量和个性化可解释性方面均优于非个性化和基于检索的基线。

View on arXiv Download PDF AI Translation

cs.CL / 55 / 2607.21143

One More Turn, Less Regret: A Regret-Based Multi-Turn Benchmark for LLMs' Clarification Policies

再多一次转弯，少一些遗憾：基于遗憾的多轮基准测试用于评估大型语言模型的澄清策略

Ta, Minh Ngoc, Nguyen, My Anh Tran, Nguyen, Duong D., Wang, Yuxia, Nakov, Preslav

Abstract

Ambiguous user requests make clarification a sequential decision problem for conversational LLM assistants: they must decide whether to ask, what to ask, when to stop, and when to answer. We introduce RegretBench, a multi-turn benchmark that evaluates clarification as policy behavior rather than isolated question quality. RegretBench provides a hidden-intent formulation of ambiguity, supports free-form interaction grounded in semantic-state tracking, and introduces a regret-based objective that measures how much value a model loses relative to a reference clarification policy. Experiments on open-domain QA and product recommendation scenarios show that final success alone is insufficient, as models with similar accuracy can differ substantially in efficiency, robustness to user behaviors, and stopping decisions. By jointly measuring intent resolution, interaction cost, ineffective clarification, and regret, RegretBench reveals whether models clarify usefully and efficiently. Our results show that effective clarification requires more than plausible questions: models must ask the right question at the right time and stop once the user's intended meaning is clear.

Chinese Translation

模糊的用户请求使得澄清成为对话型大型语言模型助手的一个序列决策问题：它们必须决定是否提问、提问内容、何时停止以及何时回答。我们引入了RegretBench，这是一个多轮基准测试，评估澄清作为策略行为而非孤立的问题质量。RegretBench提供了一种隐含意图的模糊性表述，支持基于语义状态跟踪的自由形式交互，并引入了一种基于遗憾的目标，衡量模型相对于参考澄清策略所损失的价值。针对开放域问答和产品推荐场景的实验表明，仅凭最终成功是不够的，因为具有相似准确性的模型在效率、对用户行为的鲁棒性和停止决策上可能存在显著差异。通过共同测量意图解析、交互成本、无效澄清和遗憾，RegretBench揭示了模型是否有效且高效地进行澄清。我们的结果表明，有效的澄清不仅需要合理的问题：模型必须在恰当的时机提出正确的问题，并在用户意图明确后及时停止。

View on arXiv Download PDF AI Translation

cs.CL / 56 / 2607.21231

Progressive Cramming: Reliable Token Compression and What It Reveals

渐进式压缩：可靠的令牌压缩及其揭示的内容

Tarasov, Dmitrii, Lashukov, Timofei, Goncharova, Elizaveta, Kuznetsov, Andrey

Abstract

Token cramming compresses sequences into learned embeddings with near-perfect reconstruction, but fixed token budgets and 99\% accuracy thresholds leave it unclear whether residual errors reflect optimization failures or fundamental limits. We introduce progressive cramming, which grows the target prefix token-by-token, stopping only when reconstruction is no longer achievable within a fixed optimization budget. Progressive trajectories occupy low-dimensional structure in embedding space. Prepending a crammed embedding causes a moderate but consistent accuracy drop on multiple-choice benchmarks even with the original prefix in context, and collapses capability almost entirely under generative evaluation. Causal attention-knockout interventions trace this degradation to the embedding's interactions in the model's early layers. These results position progressive cramming as a tool for studying compression limits and show that perfect reconstruction - achievable through brittle steering rather than transferable semantics - is insufficient for meaningful compression.

Chinese Translation

令牌压缩将序列压缩为学习到的嵌入，几乎实现完美重构，但固定的令牌预算和99%的准确率阈值使得尚不清楚残余误差是反映优化失败还是根本限制。我们引入了渐进式压缩，该方法逐个令牌地扩展目标前缀，仅在固定优化预算内无法再实现重构时停止。渐进轨迹在嵌入空间中占据低维结构。即使在上下文中包含原始前缀，预先添加压缩嵌入也会导致多项选择基准测试上的准确率适度但持续下降，并在生成评估中几乎完全丧失能力。因果注意力干预追踪这一降级现象，发现其源于嵌入在模型早期层中的交互。这些结果将渐进式压缩定位为研究压缩极限的工具，并表明完美重构——通过脆弱的引导而非可转移的语义实现——不足以实现有意义的压缩。

View on arXiv Download PDF AI Translation

cs.CL / 57 / 2607.21255

slang.gr as a Large-Scale Crowdsourced Resource for Non-Standard Greek

slang.gr：一个大规模众包的非标准希腊语资源

Papadakos, Panagiotis, Papantoniou, Katerina, Plexousakis, Dimitris

Abstract

Slang is a central component of everyday language, reflecting linguistic creativity, social identity, and cultural change, yet its dy- namic and non-standard nature makes it difficult to model computationally. We present the first large-scale computational study of slang.gr, a crowdsourced lexicon of Greek non-standard language, combining lexical content, user-generated tags, and interaction data. To enable the systematic analysis, we map noisy folksonomic tags to a structured multi-layer taxonomy capturing both semantic categories and sociolinguistic metadata. Using this representation, we analyze the linguistic structure of Greek slang and the behavior of its contributor community. We find that slang is strongly centered on person-related and evaluative language, exhibits high morphological creativity, and is shaped by highly skewed participation with short user lifespans and overlapping communities. Building on these signals, we introduce a community-based confidence score for definitions that integrates user roles, interaction patterns, and moderation signals. Our results show that taxonomy-based representations improve interpretability while retaining meaningful aspects of behavioral structure, enabling a more structured and interpretable analysis of confidence signals. Overall, this work establishes slang.gr as a computational resource for non-standard Greek and provides a foundation for sociolinguistic NLP, bias analysis, and the study of informal language in LLMs.

Chinese Translation

俚语是日常语言的核心组成部分，反映了语言创造力、社会身份和文化变迁，然而其动态和非标准的特性使得计算建模变得困难。我们首次对slang.gr进行大规模计算研究，该平台是一个众包的希腊非标准语言词典，结合了词汇内容、用户生成的标签和互动数据。为了实现系统分析，我们将嘈杂的民间分类标签映射到一个结构化的多层次分类法中，以捕捉语义类别和社会语言学元数据。利用这种表示方式，我们分析了希腊俚语的语言结构及其贡献者社区的行为。研究发现，俚语强烈集中于与人相关的评价性语言，表现出高度的形态创造力，并受到参与高度不均衡的影响，用户生命周期短暂且社区重叠。基于这些信号，我们引入了一种基于社区的定义置信度评分，整合了用户角色、互动模式和管理信号。我们的结果表明，基于分类法的表示提高了可解释性，同时保留了行为结构的有意义方面，使得对置信度信号的分析更加结构化和可解释。总体而言，这项工作确立了slang.gr作为非标准希腊语的计算资源，并为社会语言学自然语言处理、偏见分析以及大语言模型中的非正式语言研究奠定了基础。

View on arXiv Download PDF AI Translation

cs.CL / 58 / 2607.21274

A Comparative Evaluation of Embeddings and LLMs in a Greek Book Publisher Setting - The CUP Dataset

希腊图书出版环境中嵌入和大型语言模型的比较评估 - CUP 数据集

Papantoniou, Katerina, Papadakos, Panagiotis, Patkos, Theodore, Garefalakis, Dimitris, Vardakis, Nikos, Plexousakis, Dimitris

Abstract

We present CUP, a Greek book retrieval benchmark consisting of 868 catalog records and 104 expert-annotated queries with graded relevance judgments. We evaluate sparse (BM25), dense (sentence-transformers), hybrid, and LLM-assisted retrieval methods in this book-search setting. Multilingual embeddings outperform Greek-specific models, while hybrid retrieval performs best overall. A query-level analysis shows that BM25 excels at named-entity queries, while dense and hybrid methods improve natural-language, noisy, cross-lingual, and concept queries. Field-aware prompting has model-specific effects, while LLM TOC summarization improves TOC-only retrieval and LLM post-filtering improves early-stage retrieval at a high cost. Overall, CUP enables real-world evaluation of Greek retrieval across lexical, semantic, noisy, and cross-lingual queries.

Chinese Translation

我们提出了CUP，一个希腊图书检索基准，包含868条目录记录和104个经过专家标注的查询，附有分级相关性判断。在这一图书搜索环境中，我们评估了稀疏检索（BM25）、密集检索（sentence-transformers）、混合检索以及大型语言模型（LLM）辅助的检索方法。多语言嵌入的表现优于特定于希腊语的模型，而混合检索的整体表现最佳。查询级别的分析表明，BM25在命名实体查询中表现优异，而密集和混合方法则在自然语言、噪声、跨语言和概念查询中有所改善。领域感知提示对模型的影响具有特定性，而LLM的目录（TOC）摘要提高了仅基于目录的检索效果，LLM后期过滤在高成本下改善了早期检索。总体而言，CUP使得对希腊语检索在词汇、语义、噪声和跨语言查询方面的现实世界评估成为可能。

View on arXiv Download PDF AI Translation

cs.CL / 59 / 2607.21279

A Unified Moral-Value Dataset for Instruction Tuning

统一的道德价值数据集用于指令调优

Zeng, Zhaohui, Mai, Florian

Abstract

Large language models (LLMs) have developed rapidly and become valuable tools in everyday life. However, how to align LLMs to a particular set of human values is still an open problem. Recent studies show that instruction tuning has strong potential for zero-shot tasks and may serve as an effective approach to addressing value alignment. Nevertheless, although many datasets for instruction tuning already exist, they are not specifically designed around moral scenarios and behaviors. We construct a unified moral-value dataset that can be directly used for instruction tuning. This dataset is built upon existing moral-value datasets by merging them into a unified corpus and converting them into an instruction-response format. We show that training on a mixed dataset combining general task datasets with our dataset preserves general-task performance, and we report preliminary observations on how the mixing ratio affects value-oriented task performance. Our work provides a moral-value dataset for instruction tuning and offers a useful resource for further alignment research. The dataset is available at https://huggingface.co/datasets/teohzzh/value-for-instruction-tuning.

Chinese Translation

大型语言模型（LLMs）发展迅速，已成为日常生活中重要的工具。然而，如何将LLMs与特定的人类价值观对齐仍然是一个未解决的问题。近期研究表明，指令调优在零样本任务中具有强大的潜力，可能成为解决价值对齐的有效方法。然而，尽管已有许多用于指令调优的数据集，但它们并未专门围绕道德情境和行为进行设计。我们构建了一个统一的道德价值数据集，可以直接用于指令调优。该数据集基于现有的道德价值数据集，通过将其合并为一个统一的语料库并转换为指令-响应格式而建立。我们展示了在结合一般任务数据集与我们的数据集的混合数据集上进行训练，能够保持一般任务的性能，并报告了混合比例如何影响价值导向任务性能的初步观察。我们的工作提供了一个用于指令调优的道德价值数据集，并为进一步的对齐研究提供了有用的资源。该数据集可在 https://huggingface.co/datasets/teohzzh/value-for-instruction-tuning 获取。

View on arXiv Download PDF AI Translation

cs.CL / 60 / 2607.21284

news-crawler-LM: A Small Long-Context Model For High-Quality News Crawling

news-crawler-LM：一种用于高质量新闻爬取的小型长上下文模型

Stolzenburg, Pascal, Golde, Jonas, Dallabetta, Max, Akbik, Alan

Abstract

Extracting structured content from news pages remains challenging due to heterogeneous HTML layouts, inconsistent markup, and substantial boilerplate such as navigation elements and advertisements. Rule-based news crawlers can achieve high extraction accuracy by encoding site-specific structure, but require manual configuration in order to generalize to new publishers. Large language models provide a more flexible alternative by reducing the need for handcrafted rules, but their high computational cost limits practical deployment. In this paper, we introduce news-crawler-LM, a small long-context language model fine-tuned on high-quality, human-validated extractions from the Fundus news-crawling library. Our model converts raw HTML into plaintext and structured JSON, including fields such as headline, author, publication date, and article body. In our experiments, news-crawler-LM outperforms strong baselines in HTML-to-Markdown and HTML-to-JSON extraction, improving performance by +4.8 BLEU and +6.1 METEOR in the HTML-to-Markdown task, and by +2.2 BLEU and +4.1 METEOR in the HTML-to-JSON task. However, we also observe that our model only slightly better compared to other rule-based parsing libraries on the HTML-to-plaintext task in evaluations on previously unseen publishers. We release all models and artifacts to the research community.

Chinese Translation

从新闻页面提取结构化内容仍然面临挑战，这主要是由于异构的HTML布局、不一致的标记以及大量的冗余内容，如导航元素和广告。基于规则的新闻爬虫通过编码特定网站的结构可以实现高提取准确率，但需要手动配置以适应新的出版商。大型语言模型提供了一种更灵活的替代方案，减少了对手工规则的需求，但其高计算成本限制了实际部署。在本文中，我们介绍了news-crawler-LM，这是一种在高质量、经过人工验证的提取数据上进行微调的小型长上下文语言模型，数据来源于Fundus新闻爬取库。我们的模型将原始HTML转换为纯文本和结构化的JSON，包括标题、作者、发布日期和文章正文等字段。在我们的实验中，news-crawler-LM在HTML到Markdown和HTML到JSON的提取任务中超越了强基线，在HTML到Markdown任务中提高了+4.8 BLEU和+6.1 METEOR的性能，在HTML到JSON任务中提高了+2.2 BLEU和+4.1 METEOR。然而，我们也观察到，在对以前未见过的出版商进行评估时，我们的模型在HTML到纯文本任务上的表现仅比其他基于规则的解析库稍好。我们将所有模型和相关成果发布给研究社区。

View on arXiv Download PDF AI Translation

cs.CL / 61 / 2607.21291

Adaptive Depth Sparse Framework: Similarity-Driven Resource Allocation for Pre-Trained LLMs

自适应深度稀疏框架：基于相似性的资源分配用于预训练的大型语言模型

Wu, Yidu, Wang, Xiang, Zhao, Kejie, Wang, Zhangchi, Guo, Qinghai, Tang, Xiaoying

Abstract

Large language models (LLMs) achieve strong generation and reasoning performance, but the Transformer architecture incurs high inference cost. Existing acceleration methods often rely on task-specific fine-tuning or training from scratch, increasing adaptation cost and limiting cross-task usability. We present an Adaptive Depth Sparse Framework (AdaDSF) that converts off-the-shelf pre-trained LLMs into depth-sparse models without full retraining. Our key insight is that layers contribute unequally to representation transformation, characterized by the cosine similarity between layer input and output hidden states. Based on this, AdaDSF assigns layer-wise token retention ratios from similarity statistics, uses a lightweight router to select informative tokens at each layer, and introduces a feature-preserving alignment objective to match intermediate and final representations between sparse and dense models. On GPT-NeoX and Qwen2.5 over language modeling and commonsense reasoning, AdaDSF substantially reduces inference FLOPs while preserving performance close to dense counterparts. Under comparable sparsity, AdaDSF consistently yields smaller accuracy degradation than strong baselines including MoD, D-LLM, and DLO.

Chinese Translation

大型语言模型（LLMs）在生成和推理性能上表现出色，但Transformer架构的推理成本较高。现有的加速方法通常依赖于特定任务的微调或从头训练，这增加了适应成本并限制了跨任务的可用性。我们提出了一种自适应深度稀疏框架（Adaptive Depth Sparse Framework, AdaDSF），该框架能够将现成的预训练LLMs转换为深度稀疏模型，而无需完全重新训练。我们的关键见解是，各层对表示转换的贡献不均，表现为层输入和输出隐藏状态之间的余弦相似度。基于此，AdaDSF根据相似性统计为每层分配令牌保留比例，使用轻量级路由器在每层选择信息性令牌，并引入特征保留对齐目标，以匹配稀疏模型和密集模型之间的中间和最终表示。在语言建模和常识推理任务上，使用GPT-NeoX和Qwen2.5，AdaDSF显著减少了推理的FLOPs，同时保持接近密集模型的性能。在可比稀疏性下，AdaDSF始终比包括MoD、D-LLM和DLO在内的强基线模型产生更小的准确性下降。

View on arXiv Download PDF AI Translation

cs.CL / 62 / 2607.21324

GRADRAG: Cross-Component Prompt Adaptation for Coordinated Multi-Agent RAG

GRADRAG：协调多智能体 RAG 的跨组件提示适应

Pedinotti, Paolo, Santus, Enrico

Abstract

Retrieval-Augmented Generation (RAG) systems increasingly employ multiple LLM agents. Yet, most prior work optimizes components in isolation rather than coordinating improvements across the pipeline. We introduce GRADRAG, a framework for cross-component prompt adaptation that models the RAG pipeline as a computational graph and propagates structured evaluation feedback to update upstream agents. An Evaluator critiques downstream answers and supporting evidence, producing actionable feedback that a Prompt Optimizer uses to iteratively update adaptive agents, such as retrievers, graph constructors, and answerers. The Evaluator also triggers early stopping when the output is deemed satisfactory. We evaluate GRADRAG on the SQUALITY and QMSUM benchmarks under two retrieval paradigms: flat chunk-based retrieval using IRCoT-style query refinement (Trivedi et al., 2023), and graph-based retrieval that constructs and iteratively enriches an entity-relation graph from the document. Across both settings, GRADRAG consistently outperforms one-step refinement baselines that update only the final generator, achieving a 12-15 percentage point net preference margin in LLM-judged pairwise comparisons, with most gains realized within two refinement iterations.

Chinese Translation

检索增强生成（RAG）系统越来越多地采用多个大型语言模型（LLM）代理。然而，大多数先前的研究在优化组件时往往是孤立进行，而非协调管道中的改进。我们提出了 GRADRAG，这是一种跨组件提示适应框架，它将 RAG 管道建模为计算图，并传播结构化评估反馈以更新上游代理。评估器对下游答案和支持证据进行批评，生成可操作的反馈，提示优化器利用这些反馈迭代更新适应性代理，例如检索器、图构建器和回答者。当输出被认为令人满意时，评估器还会触发提前停止。我们在 SQUALITY 和 QMSUM 基准上评估 GRADRAG，采用两种检索范式：基于平坦块的检索，使用 IRCoT 风格的查询优化（Trivedi 等，2023），以及基于图的检索，从文档中构建并迭代丰富实体-关系图。在这两种设置中，GRADRAG 始终优于仅更新最终生成器的一步优化基线，在 LLM 评判的成对比较中实现了 12-15 个百分点的净偏好差距，大多数增益在两次优化迭代内实现。

View on arXiv Download PDF AI Translation

cs.CL / 63 / 2607.21332

Phonetic forced alignment for low-resource language varieties: Model training and evaluation on Chengdu Mandarin

低资源语言变体的语音强制对齐：在成都普通话上的模型训练与评估

Qian, Zhiheng, Li, Aini, Hu, Hai, Zhao, Liang

Abstract

Phonetic forced alignment is a key technique in phonetic research, yet existing alignment systems lack specialized models for low-resource language varieties. We address this by training text-dependent and text-independent aligners for Chengdu Mandarin using a 17-hour corpus and a custom G2P dictionary. We trained a text-dependent GMM-HMM model (Chengdu-MFA) and fine-tuned a pretrained audio encoder on frame classification with Chengdu-MFA's pseudo label for text-independent alignment (Chengdu-FC). Evaluation on an expert-annotated test set show that both methods significantly outperform Standard Mandarin baselines. Chengdu-MFA reduced average phone boundary differences by 31.8%, while Chengdu-FC achieved a 61.2% reduction. This work establishes a practical bootstrapping pipeline for developing accurate aligners for under-resourced varieties without labor- and time-intensive manual annotation.

Chinese Translation

语音强制对齐是语音研究中的一项关键技术，但现有的对齐系统缺乏针对低资源语言变体的专门模型。我们通过使用17小时的语料库和自定义的G2P字典，为成都普通话训练了文本依赖和文本独立的对齐器。我们训练了一个文本依赖的GMM-HMM模型（Chengdu-MFA），并在帧分类上对预训练的音频编码器进行了微调，以使用Chengdu-MFA的伪标签进行文本独立对齐（Chengdu-FC）。在专家标注的测试集上的评估表明，这两种方法的性能显著优于标准普通话基线。Chengdu-MFA将平均音素边界差异减少了31.8%，而Chengdu-FC则实现了61.2%的减少。这项工作建立了一个实用的引导管道，以开发适用于资源不足变体的准确对齐器，而无需耗时且劳动密集的手动标注。

View on arXiv Download PDF AI Translation

cs.CL / 64 / 2607.21340

Capital Markets LLM Reliability Score (CM-LRS): From Plausible to Bankable

资本市场大型语言模型可靠性评分 (CM-LRS)：从可信到可融资

Ahuja, Prerit

Abstract

In capital-markets workflows the question is rarely whether a large language model can produce a fluent draft, but whether the draft is bankable: defensible in front of a counter-party or a regulator, with the documents in hand. Existing methods address parts of that gap: open-domain QA benchmarks reward surface accuracy, and finance benchmarks (FinanceBench, FinQA, ConvFinQA) advance document-grounded and numerical QA but evaluate at the question-answer layer rather than the workflow outputs practitioners defend. We introduce CM-LRS, a Capital Markets LLM Reliability Score, evaluating outputs at the workflow-output layer across seven dimensions: factual accuracy, evidence traceability, numerical consistency, workflow completeness, source discipline, decision usefulness, and reviewability/auditability. Each is scored 0-5 against a rubric anchored on signals reviewers in regulated settings use; the aggregate is tunable to the workflow. We demonstrate CM-LRS on five workflows (DCM transaction-terms extraction, precedent retrieval, issuer profile synthesis, M&A transaction-comparable reasoning, ECM transaction-terms extraction) over public SEC EDGAR filings, a public UK takeover release, and fictional synthetic supplements, scoring four models against four independent LLM judges spanning three model families. Three findings. First, the frontier closed-source models cluster within 0.22 points on four-judge averaged CM-LRS (Sonnet 4.6 = 4.31, Opus 4.7 = 4.30, GPT-5.5 = 4.09); all four judges place the open-weights baseline (Llama 3.3 70B = 3.15) last. Second, that gap concentrates on retrieval (2.23) and synthesis (2.15), not extraction (0.84). Third, Decision Usefulness shows the widest cross-model dispersion of any dimension (4.0 points on issuer profiling) and top-tier inter-judge agreement (mean r = 0.52). Plausibility is cheap. Bankability is the bar.

Chinese Translation

在资本市场工作流程中，问题很少是大型语言模型是否能够生成流畅的草稿，而是草稿是否可融资：在面对对方或监管机构时能够辩护，并且手中有相关文件。现有方法部分解决了这一差距：开放领域问答基准奖励表面准确性，而金融基准（FinanceBench、FinQA、ConvFinQA）则推进了基于文档和数值的问答，但评估的是问题-答案层面，而不是从业者所辩护的工作流程输出。我们引入了CM-LRS，即资本市场大型语言模型可靠性评分，评估在七个维度上的工作流程输出：事实准确性、证据可追溯性、数值一致性、工作流程完整性、来源学科、决策有用性和可审查性/可审计性。每个维度根据在受监管环境中使用的信号的评分标准进行0-5分评分；总分可以根据工作流程进行调节。我们在五个工作流程（DCM交易条款提取、先例检索、发行人概况综合、并购交易可比推理、ECM交易条款提取）上展示了CM-LRS，使用公共的SEC EDGAR文件、公共的英国收购公告和虚构的合成补充材料，对四个模型进行评分，并与跨越三个模型家族的四位独立大型语言模型评审进行比较。三个发现。首先，前沿的闭源模型在四位评审平均CM-LRS评分上聚集在0.22分以内（Sonnet 4.6 = 4.31，Opus 4.7 = 4.30，GPT-5.5 = 4.09）；所有四位评审将开放权重基线（Llama 3.3 70B = 3.15）排在最后。其次，这一差距集中在检索（2.23）和综合（2.15）上，而不是提取（0.84）。第三，决策有用性在任何维度上显示出最大的跨模型分散性（在发行人概况上为4.0分）和顶级的评审间一致性（平均r = 0.52）。可信性是廉价的，而可融资性才是标准。

View on arXiv Download PDF AI Translation

cs.CL / 65 / 2607.21391

Word meaning co-determines vowel-inherent spectral change. A corpus-based investigation of conversational Mandarin

词义共同决定元音固有光谱变化：基于语料库的对话式普通话研究

Jin, Xiaoyun, Ernestus, Mirjam, Baayen, R. Harald

Abstract

This study investigates vowel-inherent spectral change (VISC) in spontaneous conversational Mandarin. Using the generalized additive model and word embeddings from distributional semantics, we show that, when controlling for variables such as vowel duration, gender, speaker identity, co-articulation, vowel identity, and utterance position, vowel formant trajectory dynamics have word-specific components that are tied to their meaning in context: The F1 and F2 trajectories of words can be predicted from their contextualized embeddings with an accuracy that substantially exceeds a permutation baseline. Challenging modular cognitive models of speech production, these results indicate that, words' semantics co-determine the fine details of their articulation.

Chinese Translation

本研究探讨了自发对话普通话中的元音固有光谱变化（VISC）。通过使用广义加法模型和来自分布语义的词嵌入，我们表明，在控制元音持续时间、性别、说话者身份、共发音、元音身份和话语位置等变量时，元音共振峰轨迹动态具有与其语境意义相关的特定词汇成分：词语的 F1 和 F2 轨迹可以从其上下文嵌入中预测，准确度显著超过置换基线。这些结果挑战了模块化的言语产生认知模型，表明词汇的语义共同决定了其发音的细节。

View on arXiv Download PDF AI Translation

cs.CL / 66 / 2607.21404

MemTools: A Unified Research Framework for Interoperable Agent Memory

MemTools：一个统一的可互操作代理记忆研究框架

Zhao, Chengfeng, Chen, Jinhui, Liang, Sirui, He, Shizhu, Wang, Yequan, Zhao, Jun, Liu, Kang

Abstract

While memory systems are essential for agent architectures, pervasive architectural fragmentation restricts systematic research. Existing implementations typically couple different stages of the memory lifecycle, entangle evaluation logic with specific datasets, and provide limited support for the management of heterogeneous memory types. We introduce MemTools, an interoperability research framework that decouples memory system components from their underlying deployment environments. MemTools standardizes the memory lifecycle through declarative data contracts, enabling the interchangeable assembly of components across different systems. It orthogonally separates benchmark datasets from execution protocols to facilitate controlled assessments. Furthermore, MemTools provides a unified computational interface for coordinating symbolic, neural, and multimodal memory representations within a shared runtime. Empirical evaluations on cross-system component integration, evaluation protocol reconfiguration, and heterogeneous memory coordination demonstrate that MemTools enables systematic isolation and analysis of memory design variables. These findings suggest that MemTools provides a practical and extensible infrastructure for advancing principled research on agent memory.

Chinese Translation

尽管记忆系统对于代理架构至关重要，但普遍存在的架构碎片化限制了系统性研究。现有的实现通常将记忆生命周期的不同阶段耦合在一起，将评估逻辑与特定数据集纠缠在一起，并对异构记忆类型的管理提供有限支持。我们提出了MemTools，一个将记忆系统组件与其基础部署环境解耦的互操作性研究框架。MemTools通过声明性数据契约标准化记忆生命周期，使得不同系统之间的组件可以互换组装。它正交地将基准数据集与执行协议分离，以促进受控评估。此外，MemTools提供了一个统一的计算接口，用于协调符号、神经和多模态记忆表示在共享运行时中的协同工作。对跨系统组件集成、评估协议重配置和异构记忆协调的实证评估表明，MemTools能够系统性地隔离和分析记忆设计变量。这些发现表明，MemTools为推进代理记忆的原则性研究提供了一个实用且可扩展的基础设施。

View on arXiv Download PDF AI Translation

cs.CL / 67 / 2607.21405

Anti-Periodic Positional Encoding: M\"obius Boundary Conditions Make In-Context Retrieval Reliable

反周期位置编码：M"obius 边界条件使上下文检索更可靠

Bae, Ji Ho

Abstract

M\"obius RoPE is a rotary positional encoding built on the anti-periodic frequency ladder $\theta_i=\pi(2i+1)/N$: every rotation plane advances by an odd multiple of $\pi$ across the training context, so the positional holonomy is $-1$ and the two ends of the sequence are deterministically coupled through a closed-form Dirichlet "dipole"; to our knowledge this is the first anti-periodic boundary condition in positional encoding. We verify the theory numerically to $\sim 10^{-6}$ and pretrain 48 models spanning six 160M-class and three 410M-class arms (2B FineWeb-Edu tokens each; the hybrid arm puts M\"obius frequencies on 25% of heads). Hybrid perplexity is unchanged (29.66 vs. 29.72), but needle-in-a-haystack retrieval becomes reliable: $90.3\pm5.7\%$ versus $63.3\pm31.4\%$ at context 512 ($n=6$ seeds), observed worst seed 86% versus 14%, robust variance tests $p=0.013$-$0.029$ (unadjusted), recurring at 410M (Levene $p=0.040$). Matched controls isolate the mechanism: an aperiodic ladder in the same frequency band reproduces none of the effect, and a periodic (holonomy $+1$) ladder only a fraction. Swapping trained models' frequency table back to standard RoPE (weights frozen) collapses retrieval, with damage concentrated on far needles: trained models depend on this long-range geometry. A NoPE arm is even more reliable at short context but pays a 13% perplexity tax and extrapolates worst; only the anti-periodic hybrid pairs baseline perplexity with a high reliability floor. The effect is scoped to single-needle retrieval within the training window; a one-line frequency swap thus provides zero-cost insurance against the retrieval seed lottery.

Chinese Translation

M"obius RoPE 是一种基于反周期频率阶梯 $ heta_i=rac{ ext{π}(2i+1)}{N}$ 的旋转位置编码：每个旋转平面在训练上下文中以奇数倍的 $ ext{π}$ 前进，因此位置全局性为 $-1$，序列的两个端点通过封闭形式的 Dirichlet "dipole" 确定性耦合；据我们所知，这是位置编码中首个反周期边界条件。我们在数值上验证了该理论，精度达到 $ ext{∼} 10^{-6}$，并预训练了 48 个模型，涵盖六个 160M 级和三个 410M 级的分支（每个分支 2B FineWeb-Edu 令牌；混合分支将 M"obius 频率应用于 25% 的头部）。混合困惑度保持不变（29.66 对比 29.72），但在上下文为 512 时“针在干草堆中”的检索变得可靠：$90.3 ext{±}5.7 ext{%}$ 对比 $63.3 ext{±}31.4 ext{%}$（$n=6$ 种子），观察到最差种子为 86% 对比 14%，稳健方差检验 $p=0.013$-$0.029$（未调整），在 410M 处重复出现（Levene $p=0.040$）。匹配对照实验隔离了机制：在相同频带中的非周期阶梯未能重现任何效果，而周期性（全局性 $+1$）阶梯仅能重现一部分。将训练模型的频率表切换回标准 RoPE（权重冻结）导致检索崩溃，损害集中在远处的针上：训练模型依赖于这种远程几何结构。NoPE 分支在短上下文中更为可靠，但付出了 13% 的困惑度代价，并且在外推时表现最差；只有反周期混合将基线困惑度与高可靠性底线相结合。该效应限于训练窗口内的单针检索；因此，一次频率的切换提供了对检索种子抽奖的零成本保险。

View on arXiv Download PDF AI Translation

cs.CL / 68 / 2607.21424

An Evaluation Framework for Structured Audio Captions Validated by Controlled Perturbations

通过控制扰动验证的结构化音频字幕评估框架

Wu, Liang-Yuan, Sridhar, Sripathi, Cartwright, Mark, Fuentes, Magdalena

Abstract

Recent advancements in automated audio captioning (AAC) have shifted from monolithic sentence generation toward structured formats that explicitly disentangle distinct acoustic and semantic properties. However, evaluating this heterogeneous data remains a significant challenge. Existing caption metrics focus on flat textual outputs and fail to reliably assess multimodal attributes. To bridge this gap, we propose a multi-axis evaluation framework tailored for structured audio descriptions. Building on the AudioCards dataset, we evaluate outputs across five orthogonal axes: tag-sets, descriptions, logical reasoning, numeric measurements, and spectral profiles. Our approach combines Large Language Model (LLM) judges to capture semantic nuance with deterministic computational metrics to precisely measure acoustic deviations. To rigorously validate the reliability of this framework, we introduce a controlled perturbation testing protocol that injects typed, graded errors into groundtruth annotations. Our results demonstrate that this framework successfully distinguishes meaning-preserving paraphrases from genuine semantic and acoustic corruptions.

Chinese Translation

最近，自动音频字幕生成（AAC）的进展已从单一的句子生成转向明确区分不同声学和语义特性的结构化格式。然而，评估这种异质数据仍然是一个重大挑战。现有的字幕评估指标侧重于平面文本输出，未能可靠地评估多模态属性。为了解决这一问题，我们提出了一种针对结构化音频描述的多轴评估框架。基于AudioCards数据集，我们在五个正交轴上评估输出：标签集、描述、逻辑推理、数值测量和频谱特征。我们的方法结合了大型语言模型（LLM）评估者，以捕捉语义细微差别，并使用确定性计算指标精确测量声学偏差。为了严格验证该框架的可靠性，我们引入了一种控制扰动测试协议，该协议将类型化、分级的错误注入到真实注释中。我们的结果表明，该框架成功区分了保留意义的释义与真正的语义和声学损坏。

View on arXiv Download PDF AI Translation

cs.CL / 69 / 2607.21433

Token Budget Saturation and Mechanistic Early Detection of Reasoning Non-Convergence in Chain-of-Thought Models

令牌预算饱和与链式思维模型中推理非收敛的机制性早期检测

Oladri, Renuka, Jawahar, Niveda, Mohamed, Abdirisak

Abstract

Chain-of-thought reasoning models such as DeepSeek-R1-Distill-Qwen-7B exhibit a bimodal convergence pattern: generations either terminate within a token budget (converged) or exhaust it without reaching a conclusion (non-converged). We characterize this phenomenon empirically, showing that converged generations achieve 90.3% accuracy on AIME 1983-2024 while non-converged ones achieve only 6.6%, with an overall convergence rate of 62.0%. We then ask whether this outcome is detectable early in the thinking chain using internal model representations. Training linear probes on hidden-state activations at token positions 50-300, we find that layer-20 activations at token 150 achieve AUC 0.608 (+-0.080, 5-fold CV), reliably above chance even at token 50. Activation probes consistently outperform behavioral baselines derived from token entropy and repetition statistics. A sweep-level permutation test yields p=0.063 (100,000 permutations), consistent with a modest signal that our sample size cannot confirm at conventional thresholds. These findings suggest that convergence fate is partially encoded in intermediate representations well before the generation ends, opening a path toward early-exit inference and adaptive compute allocation.

Chinese Translation

链式思维推理模型，如 DeepSeek-R1-Distill-Qwen-7B，表现出双峰收敛模式：生成结果要么在令牌预算内终止（收敛），要么在未达成结论的情况下耗尽预算（非收敛）。我们通过实证研究对这一现象进行了表征，显示收敛的生成结果在 AIME 1983-2024 上的准确率达到 90.3%，而非收敛的结果仅为 6.6%，整体收敛率为 62.0%。接着，我们探讨是否可以通过内部模型表示在思维链的早期阶段检测到这一结果。通过在令牌位置 50-300 的隐藏状态激活上训练线性探针，我们发现令牌 150 处的第 20 层激活达到 AUC 0.608（+-0.080，5 倍交叉验证），即使在令牌 50 时也可靠地高于随机水平。激活探针在性能上始终优于基于令牌熵和重复统计得出的行为基线。全局置换检验结果为 p=0.063（100,000 次置换），与我们的样本量在常规阈值下无法确认的适度信号一致。这些发现表明，收敛的命运在生成结束之前的中间表示中部分编码，为早期退出推理和自适应计算分配开辟了道路。

View on arXiv Download PDF AI Translation

cs.CL / 70 / 2607.21445

When Trivia Is Not Trivial: Everyday Knowledge Failures in Multilingual LLMs

当琐事并非琐事：多语言大型语言模型中的日常知识失败

Mosolova, Anna, Seddah, Djamé

Abstract

Quiz rooms, trivia nights, and quiz shows challenge human knowledge across a wide range of topics, from canonical facts to everyday culture. In this paper, we examine whether large language models (LLMs) can perform competitively in such settings, using quiz-style questions to test them on both common and niche topics. We introduce TriviaRoomQA, a multilingual benchmark designed to evaluate everyday, culturally grounded, and long-tail knowledge across 288 topics. The benchmark contains 3,300 parallel multiple-choice questions in six European languages and additional 5,340 French-only questions for a more fine-grained case study. We evaluate 30 open-weight LLMs from European, Asian, and North American providers, covering models from 7 to 70B parameters. We find that models are strong on knowledge-intensive topics such as history, geography, and mathematics, but substantially weaker on everyday popular-culture topics such as celebrities, music, movies, and news. Moreover, model performance varies across languages even for the same underlying questions, suggesting that access to factual knowledge is not always language-independent. In sum, our dataset and experiments demonstrate an important knowledge gap which is not captured by existing academic-based saturated benchmarks.

Chinese Translation

问答室、琐事之夜和问答节目在广泛的主题上挑战人类知识，从经典事实到日常文化。本文探讨大型语言模型（LLMs）在此类环境中是否能够具有竞争力，使用问答风格的问题对其在常见和小众主题上的表现进行测试。我们引入了TriviaRoomQA，这是一个多语言基准，旨在评估日常、文化基础和长尾知识，涵盖288个主题。该基准包含3300个并行的多项选择题，涉及六种欧洲语言，并额外提供5340个仅限法语的问题，以便进行更细致的案例研究。我们评估了来自欧洲、亚洲和北美的30个开放权重LLM，涵盖参数从7亿到700亿的模型。我们发现，模型在历史、地理和数学等知识密集型主题上表现强劲，但在名人、音乐、电影和新闻等日常流行文化主题上则明显较弱。此外，即使对于相同的基础问题，模型在不同语言中的表现也存在差异，这表明获取事实知识并不总是与语言无关。总之，我们的数据集和实验展示了一个重要的知识差距，而这一差距并未被现有的基于学术的饱和基准捕捉到。

View on arXiv Download PDF AI Translation

cs.CL / 71 / 2607.21447

RUMBA: Russian User Memory Benchmark

RUMBA：俄罗斯用户记忆基准

Shevtsova, Elizaveta, Glebkina, Inna, Baushenko, Mark, Gulyaev, Pavel, Fenogenova, Alena

Abstract

The ability to handle long-term memory in LLMs is becoming increasingly critical, yet existing benchmarks remain English-centric and rely on aggregate retrieval metrics, failing to capture interactions between long-range context, temporal information, and reasoning. To address this, we introduce RUMBA (Russian User Memory BenchmArk) - a new benchmark for long-term conversational memory that provides a fine-grained taxonomy of memory-centric question types and a unified methodology accounting for semantic type, session scope, temporal reasoning, and the explicitness of temporal expressions. RUMBA consists of timestamped user-assistant dialogues with QA pairs requiring retrieval, combination, and reasoning across sessions. While designed for Russian, we also provide an aligned English subset under the same methodology. We evaluate contemporary memory systems and long-context models, and show how RUMBA serves as a diagnostic tool to analyze model behavior across benchmark slices and identify strengths and failure modes of different memory mechanisms.

Chinese Translation

在大型语言模型（LLMs）中处理长期记忆的能力变得越来越重要，但现有的基准测试仍然以英语为中心，并依赖于聚合检索指标，未能捕捉长距离上下文、时间信息和推理之间的交互。为了解决这个问题，我们引入了RUMBA（Russian User Memory BenchmArk）——一个新的长期对话记忆基准，提供了以记忆为中心的问题类型的细分分类法，以及考虑语义类型、会话范围、时间推理和时间表达的明确性的统一方法。RUMBA包含带有时间戳的用户-助手对话，包含需要在会话之间进行检索、组合和推理的问答对。虽然该基准是为俄语设计的，但我们也提供了在相同方法下对齐的英语子集。我们评估了当代记忆系统和长上下文模型，并展示了RUMBA如何作为诊断工具，分析模型在基准切片中的行为，并识别不同记忆机制的优势和失败模式。

View on arXiv Download PDF AI Translation

cs.CL / 72 / 2607.21491

What, Where, and How: Disentangling the Roles of Task, Language, and Model in Code Model Representations

什么、哪里以及如何：解开任务、语言和模型在代码模型表征中的角色

Wilam, Piotr

Abstract

Do independently trained language models come to represent the same thing in the same way? We answer for code, extending a recently introduced concept-circuit extraction method to a 2x2 design -- Python and Rust crossed with Qwen2.5-Coder-7B and DeepSeek-Coder-V1-6.7B -- and measuring a complete inventory of grammatical concepts (58 Python, 57 Rust) identically in all four cells: the smallest design that separates what depends on the task, the language, and the model. The answer splits into three parts. What earns dedicated circuitry is set by the task: the models agree on which concepts receive circuits (Spearman $\rho$ = 0.638 for Python, 0.673 for Rust, both p < $10^{-7}$). Where those circuits sit is set by the model: Qwen processes concepts in a late band (~L17-19), DeepSeek at L6-7, for both languages. How circuits grow across layers is also set by the model: Qwen gives its atomic concepts an early spike that DeepSeek does not. "Are circuits universal?" thus has no single answer: yes for What, no for Where and How -- universality is a property of representational content, not of computational organisation. None of this structure was fixed in advance. The agreement could have landed anywhere between independence and identity; it lands at $\rho \approx 0.65$. Rust constructs receive 2-3x more concept-specific circuitry than their Python equivalents, in both models. Both models share neurons between the languages (6/7 and 7/7 paired constructs), DeepSeek 1.94x more than Qwen -- a direction no prior result predicts. And Qwen binds nine keywords of Rust's type-and-trait machinery into one tight neuron cluster (Jaccard 0.535 vs null 0.112, p < 0.001), a semantic dimension invisible in surface syntax. Ablation and linear probes confirm the circuits are functional. All claims are scoped to this 2x2; whether the per-model profile predicts a third model is the designed next test.

Chinese Translation

独立训练的语言模型是否以相同的方式表示相同的事物？我们针对代码进行回答，扩展了最近引入的概念电路提取方法，采用2x2设计——Python和Rust与Qwen2.5-Coder-7B和DeepSeek-Coder-V1-6.7B交叉，并在所有四个单元中相同地测量了完整的语法概念清单（58个Python，57个Rust）：这是一个最小的设计，能够区分任务、语言和模型之间的依赖关系。答案分为三个部分。什么获得专用电路由任务决定：模型在哪些概念获得电路上达成一致（Spearman $ ho$ = 0.638对于Python，0.673对于Rust，均p < $10^{-7}$）。电路的位置由模型决定：Qwen在后期带（~L17-19）处理概念，DeepSeek在L6-7处理这两种语言的概念。电路在层之间的增长方式同样由模型决定：Qwen为其原子概念提供了一个早期的峰值，而DeepSeek则没有。因此，“电路是普遍的吗？”并没有单一的答案：对于什么是的，答案是肯定的；对于哪里和如何则是否定的——普遍性是表征内容的属性，而不是计算组织的属性。所有这些结构在事先并未固定。协议可以在独立性和同一性之间的任何地方落脚；它落在$ ho ext{approx} 0.65$。Rust构造在两个模型中获得的概念特定电路比其Python等价物多2-3倍。两个模型在语言之间共享神经元（6/7和7/7配对构造），DeepSeek的共享程度比Qwen高1.94倍——这是之前结果未曾预测的方向。而且，Qwen将Rust的类型和特征机制的九个关键字绑定到一个紧密的神经元集群中（Jaccard 0.535对比空值0.112，p < 0.001），这是在表面语法中不可见的语义维度。消融和线性探针确认这些电路是功能性的。所有声明的范围限于这个2x2设计；每个模型的特征是否能预测第三个模型是下一个设计的测试。

View on arXiv Download PDF AI Translation

cs.CL / 73 / 2607.21498

Artificial Epanorthosis: Why large language models overuse a classical rhetorical figure, and how to mitigate it

人工修正：为何大型语言模型过度使用这一经典修辞手法，以及如何减轻这种现象

Boggia, Federico

Abstract

A rhetorical figure that Cicero and Quintilian catalogued two thousand years ago reappears, systematically, in the text of large language models: epanorthosis, the self-correction of the specimen {\guillemotleft}This is not a course. It is a journey of transformation{\guillemotright}. This essay argues that the overuse is a trained disposition, driven mainly by a training distribution rich in promotional prose and by preference tuning (RLHF) that rewards confident, emphatic phrasing; the left-to-right nature of generation is an amplifier rather than the root cause. Building on evidence that models diverge from human rhetorical style, and on Fontanier's classification of epanorthosis as a figure of thought, it sets out a programme that scores the figure against genre-specific human baselines through an Epanorthosis Index (density relative to the human rate). A first measurement, on three sizes of one instruction-tuned model family, finds mis-calibration by register in both directions: the models overshoot in oratory (about twofold, near threefold in Italian, concentrated in the larger tiers) and undershoot in informal question-and-answer writing, while matching humans in argument, journalism, and encyclopedic prose. Three constructive contributions follow: a survey of mitigation techniques centred on lightweight LoRA adapters; a demonstration, in Italian, that a one-line instruction cuts the figure by half to nearly three-quarters and that a supervised-fine-tuning adapter removes it almost entirely, with a scaling coefficient that dials the reduction back onto the human rate; and the argument that the target is calibration to the human rate for each genre, not elimination. It closes on the stakes: the real risk is that we begin to write like the machines.

Chinese Translation

两千年前，西塞罗和昆体良所列举的修辞手法在大型语言模型的文本中系统性地重新出现：修正（epanorthosis），即自我修正的表述{AB}这不是一门课程。这是一次转变的旅程{BB}。本文认为，过度使用这一修辞手法是一种经过训练的倾向，主要受益于丰富的宣传性文体的训练分布以及奖励自信、强调性措辞的偏好调优（RLHF）；生成的从左到右特性是放大器，而非根本原因。基于模型与人类修辞风格的差异证据，以及丰塔尼耶（Fontanier）将修正分类为思维的修辞手法的观点，本文提出了一项计划，通过修正指数（Epanorthosis Index）对该修辞手法进行评分，以与特定类型的人类基准进行比较（相对于人类的密度）。在对三个不同规模的指令调优模型家族进行的首次测量中，发现模型在不同语域中的标定存在双向失调：模型在演讲中表现过度（约两倍，在意大利语中接近三倍，主要集中在较大层级），而在非正式问答写作中表现不足，同时在论证、新闻报道和百科全书式的散文中与人类相匹配。接下来有三项建设性贡献：围绕轻量级LoRA适配器的减轻技术调查；在意大利语中演示一行指令将修正数量减少一半至近四分之三，并且监督微调适配器几乎完全消除了这一现象，同时具有将减少比例调回人类水平的缩放系数；以及论证目标是针对每个类型的标定，而非消除。最后强调了风险：真正的风险在于我们开始像机器一样写作。

View on arXiv Download PDF AI Translation

cs.CL / 74 / 2607.21540

DONDO: Open w2v-BERT Speech-Recognition Base Models for African Languages

DONDO：面向非洲语言的开放式 w2v-BERT 语音识别基础模型

Azunre, Paul

Abstract

We present DONDO, a family of open, permissively licensed automatic speech recognition (ASR) base models for African languages, built on the w2v-BERT 2.0 self-supervised speech encoder. DONDO comprises twenty-one monolingual models and five multilingual models spanning twenty-seven language varieties across Ghana, Sierra Leone, Nigeria, Senegal, Kenya and Zimbabwe. Models are fine-tuned primarily on read speech drawn from religious texts, which offer broad, license-clear and orthographically consistent coverage for languages that otherwise lack transcribed audio. We describe a two-step (and, for one family, three-step) learning-rate-annealed fine-tuning procedure that first adapts a shared multilingual model at a high learning rate and then anneals it to recover, and in several cases surpass, strong monolingual baselines. We further describe a lightweight language-conditioning mechanism that injects a one-hot language identity as a sequence of prefix frames prepended to the acoustic features, allowing a single multilingual checkpoint to be steered to a target language at inference. Across the five multilingual families the annealed models reach average word error rates (WER) of 10-13%, closing most of the gap to monolingual models while covering many languages in a single checkpoint. All models are released on the Hugging Face KhayaAI organisation under the Apache-2.0 license (attribution only) so that others may fine-tune them freely, including for commercial use. We provide a conservative estimate that the languages covered are spoken by on the order of one hundred million first-language speakers, and by substantially more when second-language use is included.

Chinese Translation

我们提出了 DONDO，这是一个开放的、许可友好的自动语音识别（ASR）基础模型系列，专为非洲语言构建，基于 w2v-BERT 2.0 自监督语音编码器。DONDO 包含二十一种单语模型和五种多语模型，覆盖来自加纳、塞拉利昂、尼日利亚、塞内加尔、肯尼亚和津巴布韦的二十七种语言变体。模型主要在从宗教文本中提取的朗读语音上进行微调，这些文本为 otherwise 缺乏转录音频的语言提供了广泛、许可明确且正字法一致的覆盖。我们描述了一种两步（对于一个系列为三步）学习率退火微调程序，首先在高学习率下适应共享的多语模型，然后退火以恢复，并在多个情况下超越强大的单语基线。我们进一步描述了一种轻量级语言条件机制，该机制将一个热编码语言身份作为一系列前缀帧注入到声学特征中，使得单个多语检查点在推理时能够引导到目标语言。在五个多语系列中，退火模型的平均词错误率（WER）达到 10-13%，缩小了与单语模型之间的大部分差距，同时在单个检查点中覆盖多种语言。所有模型均在 Hugging Face KhayaAI 组织下以 Apache-2.0 许可证（仅限署名）发布，以便他人可以自由微调，包括用于商业用途。我们提供了一个保守的估计，所覆盖的语言由大约一亿的母语使用者所讲，当包括第二语言使用时，使用者数量会显著增加。

View on arXiv Download PDF AI Translation

cs.CL / 75 / 2607.21570

MedGame: Storytelling Gamification Empowered by Large Language Models for Medical Education

MedGame：基于大型语言模型的故事化游戏化医学教育

Wu, Qian, Zhou, Xinrong, Ma, Zizhan, Chen, Kai, Gao, Zheyao, Lin, Xun, Wu, Hongqiu, Gou, Longfei, Liu, Yixiao, Lau, Ann Sin Nga, Dou, Qi

Abstract

Large Language Models (LLMs) show promise for medical education, but most existing systems focus on localized interactions such as question answering or single-turn feedback, rather than organizing an entire clinical case into a decision-centered learning trajectory. We introduce \textit{MedGame}, a framework that transforms static clinical cases into structured, executable storytelling games. MedGame uses a dual-engine design: a Medical Narrative Designer synthesizes case-grounded clinical storylines with states and decision nodes, while a Story Director converts them into dependency-aware multimodal orchestration plans rendered by our released interactive platform. We construct MedGame Bench, a 5,000-case benchmark and evaluation protocol for Medical Narrative Generation and Story Direction. Experiments show that task-specific fine-tuning substantially improves open-source LLMs on MedGame Bench and narrows the gap with commercial models. A pilot student study further shows that learners perceive MedGame as more engaging and useful than text-only alternatives.

Chinese Translation

大型语言模型（LLMs）在医学教育中展现出潜力，但现有大多数系统主要集中于局部交互，如问答或单轮反馈，而不是将整个临床案例组织成以决策为中心的学习轨迹。我们提出了 extit{MedGame}，一个将静态临床案例转化为结构化、可执行故事游戏的框架。MedGame采用双引擎设计：医学叙事设计师合成基于案例的临床故事情节及其状态和决策节点，而故事导演则将其转化为依赖感知的多模态编排计划，并通过我们发布的互动平台呈现。我们构建了 MedGame Bench，这是一个包含5000个案例的基准和医学叙事生成与故事指导的评估协议。实验表明，任务特定的微调显著提高了开源LLMs在MedGame Bench上的表现，并缩小了与商业模型之间的差距。一项初步的学生研究进一步表明，学习者认为MedGame比仅文本的替代方案更具吸引力和实用性。

View on arXiv Download PDF AI Translation

cs.CL / 76 / 2607.21574

Surprisal Theory is Tautological (without Rational Grounding)

惊讶理论是自我重复的（缺乏理性基础）

Cotterell, Ryan

Abstract

Surprisal theory holds that the human processing difficulty of a linguistic unit in context is an affine function of its surprisal under some language model. I argue this claim is a tautology without further constraint: for any non-negative difficulty measure over units in context, there exists a language model whose surprisal is an affine function of it under mild technical conditions. Therefore, because any pattern of difficulty is consistent with some language model, without an additional constraint on the language model, surprisal theory makes no falsifiable predictions. The tautology was long obscured by an assumption implicit in two decades of psycholinguistic work---that the relevant language model is the distribution that generated the training corpus, so that improving corpus fit improves predictions of human behavior. Recent empirical work has undermined this assumption, demonstrating that better corpus models can be worse predictors of processing difficulty. I conclude that breaking the tautology requires a rationalist intervention, i.e., the relevant language model must be derived from a non-empirically motivated model of the comprehender, which could be based on, for instance, memory constraints or processing goals, and that, thus, does not depend on the behavioral data surprisal theory is meant to explain.

Chinese Translation

惊讶理论认为，在特定语境中，语言单位的人类处理难度是其在某种语言模型下的惊讶度的仿射函数。我认为这一主张是一个自我重复的命题，缺乏进一步的约束：对于任何非负的上下文单位难度度量，都存在一个语言模型，其惊讶度在温和的技术条件下是该度量的仿射函数。因此，由于任何难度模式都与某种语言模型一致，且在没有对语言模型的额外约束的情况下，惊讶理论无法做出可证伪的预测。这个自我重复的特性长期以来被隐含在二十年的心理语言学研究中的一个假设所掩盖——即相关的语言模型是生成训练语料库的分布，因此改善语料库的拟合度能够改善对人类行为的预测。最近的实证研究削弱了这一假设，表明更好的语料库模型可能在处理难度的预测上表现更差。我得出结论，打破这一自我重复的特性需要理性干预，即相关的语言模型必须来源于一个非经验驱动的理解者模型，这个模型可以基于例如记忆限制或处理目标，因此不依赖于惊讶理论旨在解释的行为数据。

View on arXiv Download PDF AI Translation

arXiv Papers

HERMES: Heterogeneous Edge-Relational Multi-Head Embedded SSM Attention for Traffic Conflict Prediction at Signalized Intersections

PhysCoRe: Physics-Corrected Residual World Models for Material-Aware Deformable Dynamics

Scalable Low-Cost Laboratory Automation: A Digital Twin-Integrated Robotic Platform for Autonomous Liquid Handling (RAINBOTTM)

Safe and Scalable Multi-Drone Payload Transport via CBF-based Reinforcement Learning with Zero-Shot Sim-to-Real Transfer

Towards Capability-Aware Traversability Navigation for Unstructured Environments

FELT: Generating Tactile Signals from Vision for Visuo-Tactile Manipulation

Decentralized UAV Swarms for Ground Target Protection in GPS- and Communication-Denied Environments

Self-Supervised Bio-Inspired Robotic Trajectory Planning with Obstacle Avoidance

A real-time RGB-D perception pipeline for autonomous impact hammers in mining: self-filtering, rock segmentation and rock-breaking poses generation

Emergent Compositional Skills in Mixture-of-Experts VLAs

Socially Consistent Multi-Robot Navigation Using Decoupled Planning and Trajectory Coordination

Robostral Navigate

URF: A Unified Robot Control-Policy Framework for Stable Contact Aware Manipulation

Deep Reinforcement-Learning-Guided Model Predictive Control for Preventing Overtakes in Autonomous Racing

Distributed Model-Based Diffusion For Scalable Multi-Robot Trajectory Optimization

TableVerse: A Large-scale Tabletop Dataset with Real-world Grounded Layouts for Generalizable Manipulation

ZONDA: Zero-shot Object Navigation with Dynamic Avoidance in Multi-floor Environments

A Real-Time Generalized Nash Equilibrium Framework for Interaction-Aware Autonomous Driving in Mixed Traffic

GuidedAttention: Interpretable and Correctable Visual Attention for OOD-Robust Robot Manipulation via Imitation Learning

Human-Inspired Framework for Robotic Craniotomy: Integrating Multimodal Fusion and Adaptive Trajectory Adjustment

RL-MACRO: A Cybernetic Closed-Loop Intelligence Framework for Multimodal Adaptive Robotic Craniotomy

FORGE-plus: Force-Budgeted Recovery for Contact-Rich Assembly with a Frozen LLM Supervisor

Factorized Spatio-Temporal Convolutions for Human Pose Estimation from Planar Lidar

Grasp, Handover, Rotate: Bimanual Object Reorientation via Compositional Diffusion and Energy-Based Optimization

VoLN: Vision-Only Long-Horizon Navigation---Paradigm, Benchmark, and Method

GLAM-SLAM: Real-time Gaussian Large-scale Mapping via Flow Densification and Spatial Decomposition

GS-Agent: Creating 4D Physical Worlds With Generative Simulation

Beyond Episodic Evaluation: Memory Architectural Bottlenecks in Sequential Embodied Question Answering

Scale Up Strategically: Learning Compositional Generalization via Bias-Aware Evaluation and Data Collection for Robotic Manipulation

AXIS: A Growable Community-Driven Data Engine for Scalable Robot Manipulation

RealVDeblur: One-Step Diffusion for Generalizable Real-World Video Deblurring

Masked Topology Modeling for Self-Supervised Learning on Parametric CAD

Axolotl3D: a Unified Framework for Faithful 3D Shape Completion

ODeform: Learning Continuous 4D Motion for Shape Deformation with Neural ODEs

Spatially Grounded Concept Bottleneck Models for Trustworthy Breast Ultrasound Diagnosis

DS@GT ARC at ImageCLEFmed GANs 2026: Geometric Filtering for Privacy-Preserving CT Slice Generation

U-CFR: Uncertainty-Guided Cascade Forward Refinement for Interactive Segmentation

Rethinking Open-World Video Anomaly Detection: Diagnosing Definition Blindness

3D-GIMP: When 3D Gaussian Inpainting Meets PatchMatch

Ocular Verification for Virtual Reality

SubSplat: High-Resolution Pixel-aligned 3DGS via Sub-pixel Gaussian Reparameterization

Explainable graph attention network for stress recognition (StressGAT) via differential action units

Agentic Designer: Progressive Multi-Agent Collaboration for Structure-Aware Interior Layout Generation

ViSTR-Bench: Can MLLMs Reason from Continuous Visual Cues in Dynamic Scenes?

Webly Supervised Multi-Label Recognition: Evaluation Benchmark and Dual-Branch Multi-Label Contrastive Learning

WhereEdit: Mask-aware Local Latent Editing for One-Step Image Editing

Engine-Native Editable 3D World Reconstruction with Objects and Lighting

MAGE-Vein: Multi-Instance Age and Gender Estimation from Finger Vein Images

DINO-VPT: Hierarchical Visual Prompt Tuning for Joint Physical-Digital Face Anti-Spoofing

Sidewalk Moments: Are Richer Representations Always More Human-Aligned? Evidence from City-Walk Videos

FA-LAM: Focus-Aware Large Avatar Model for One-Shot 4D Animatable Gaussian Head

MagicMakeup: A Region-Controllable Diffusion Transformer for High-Fidelity Makeup-Transfer

Ms. Forcing: Efficient Streaming Video Generation with Multi-Scale Patchification and Attention

RECO: Region-Aware Compensation for Extrinsic Perturbations in Roadside 3D Detection

FSB-Net: Frequency-Spatial Boundary Network for Brain Stroke Lesion Segmentation in Non-Contrast CT

Unsupervised Metal Artifact Reduction in Dental CBCT using Fine-tuned Cycle-Consistent Adversarial Networks

Distribution-Alignment Bridge for Uncertainty-Aware Text-to-Video Retrieval

HyWorldVLA: A Vision-Language-Action Model with Hybrid World Modeling for Autonomous Driving

Latent Variable-Mediated Cross-Learning for Few-Shot Acoustic Impedance Imaging

Sparse Concept Channels in Frozen 3D CT Vision Encoders

AUCH-Net: Action Unit-Based Consistency-Aware Hypergraph Network for Cross-Domain Few-Shot Facial Expression Recognition

Explainable Deepfake Detection Challenge

ProCap: Prominence-guided Object Rectification for Faithful and Comprehensive Video Captioning

WAT3R: Feedforward Underwater 3D Reconstruction

GroupVideo: Multi-Identity Customized Text-to-Video Generation

Spectral-Spatial Synergistic Guided Network for Hyperspectral Salient Object Detection

GeoThreat: Transferable Targeted Adversarial Attacks on Large Vision-Language Models for Remote Sensing Image Interpretation

HyperImageNet: A Large-Scale High-Spatial Resolution Hyperspectral Imagery Classification Benchmark

Achieving Text-based Person Retrieval with Any Granularity

MVEI & EmObserver: Empowering MLLM-Oriented Visual Emotional Intelligence via Emotion Statement Judgement

Do Pathology Vision-Language Models Truly See Pathology?

TransBiolab: A Real-World Multi-View Dataset of Cluttered Transparent Biomedical Objects

Show, Don't Tell: Evaluating Spatial Cognition in Generative Pixels Rather Than LLM Text

C-PTQ: Fisher-weighted Channel-wise Sensitivity for Post-training Quantization of MLLMs

The RealDefocus Benchmark for Defocus Deblurring

Geo3R: Mitigating Spatial Reasoning Hallucination in Multimodal Large Language Models

Loss Landscape Topology Reveals Why Simple Baselines are Competitive at 3D Point Cloud Segmentation Under Class Imbalance

HalluScope: Fine-grained Hallucination Diagnosis for Multimodal Large Language Models

The Second LoViF 2026 Challenge on Real-World All-in-One Image Restoration: Methods and Results