Daily Papers · 2026-10-03
📄 论文日报 | 来源：Hugging Face Daily Papers
🗓 Episode：2026.10.02 | 流式视频主动记忆；音视频扩散奖励路由
1. OneStreamer：统一流式视频交互中的感知、记忆与主动响应
- 英文标题：OneStreamer: Unifying Perception, Memory, and Proactive Response in Streaming Video Interaction
- arXiv：2610.01762
💡 一句话解读：流式视频AI需要一边看一边记住可能有用的信息，并在合适时机主动回答，但实时感知和长期记忆很难兼顾。OneStreamer让模型在观看时自动生成带时间戳的画面描述和事件总结作为可复用记忆，同时用新训练方法减少无效等待，从而在多个视频理解测试中表现最佳。
流式视频大语言模型（Streaming video LLMs）必须在尚不知未来任务相关性时保留证据，并在证据充分时做出响应。难点在于形成可复用的事实记忆而不损害实时感知。我们提出OneStreamer，它通过共享的主动生成过程，联合学习与查询无关的证据记录和任务响应。其主动分层字幕记忆（Proactive Hierarchical Caption Memory, PHCM）生成基于时间的局部细节字幕和已完成事件的摘要。流式字幕目标在训练中监督对已观察视频前缀的解读。推理时，模型生成的记录补充近期视觉窗口，提供可复用的事实上下文而无需重新访问历史视觉特征。主动状态转移学习（Proactive State Transition Learning, PSTL）通过保留所有输出锚点的监督并选择代表性的状态变化和状态持续标记，减少了重复等待状态的主导地位。我们进一步开发了流式数据合成流程，使输出内容和时机与可用证据对齐。将生成的流式字幕和问答与清理后的开源数据结合，得到OneStreamer-1M，一个广泛覆盖的流式视频交互数据集，包含超过一百万条记录，涵盖多种任务。我们的4B模型在所有八个评估的流式视频理解基准上均取得了相比方法的最佳结果。消融实验表明，保留生成的字幕提高了历史问答性能而不降低实时感知。PSTL在仅监督27.5%的标注状态标记的情况下，也优于密集状态监督。这些结果共同支持主动生成作为连接流式视频交互中感知、记忆形成和及时响应的共享学习接口。
📄 English Abstract
Streaming video LLMs must retain evidence before its relevance to future tasks is known and respond when sufficient evidence becomes available. The challenge is to form reusable factual memory without compromising real-time perception. We introduce OneStreamer, which jointly learns query-independent evidence recording and task response through a shared proactive generation process. Its Proactive Hierarchical Caption Memory (PHCM) produces time-grounded local-detail captions and summaries of completed events. Streaming caption targets supervise the interpretation of observed video prefixes during training. At inference, model-generated records complement a recent visual window, providing reusable factual context without revisiting historical visual features. Proactive State Transition Learning (PSTL) reduces the dominance of repeated waiting states by preserving supervision at all output anchors and selecting representative state-change and state-persistence tokens. We further develop a streaming data synthesis pipeline that aligns output content and timing with available evidence. Combining the resulting streaming captions and QA with cleaned open-source data yields OneStreamer-1M, a broad-coverage streaming video interaction dataset with over one million records spanning diverse tasks. Our 4B model achieves the best results among the compared methods across all eight evaluated streaming video understanding benchmarks. Ablations show that retaining generated captions improves historical QA without degrading real-time perception. PSTL also outperforms dense state supervision while supervising only 27.5% of annotated state tokens. Together, these results support proactive generation as a shared learning interface connecting perception, memory formation, and timely response in streaming video interaction.
2. 自适应奖励路由：通过前向过程强化学习实现联合音视频扩散的动态多奖励优化
- 英文标题：Adaptive Reward Routing: Dynamic Multi-Reward Optimization for Joint Audio-Video Diffusion via Forward-Process RL
- arXiv：2609.37200
💡 一句话解读：这篇研究要解决的是：训练能同时生成音频和视频的AI模型时，多个目标（画质、音质、音画对齐等）互相打架，而现有方法用固定权重和固定位置来调整，跟不上模型变化。作者提出“自适应奖励路由”，一边动态决定该在哪些跨模态层更新，一边动态协调各奖励的权重，从而让训练更稳、效果更好。
多奖励引导的强化学习（RL）为提升联合音视频扩散模型在多个互补目标上的表现提供了有前景的途径，这些目标包括各模态自身的质量、跨模态语义对齐以及时间同步。然而，其效果取决于训练过程中不断变化的两个量：奖励驱动的更新应作用于何处，以及相互竞争的奖励应如何组合。现有方法往往依赖固定的路由和奖励权重，无法跟踪模型功能的演变。为解决这些局限，我们提出自适应奖励路由（Adaptive Reward Routing），在联合音视频扩散模型的前向过程强化学习（即 DiffusionNFT）中，联合自适应地调整更新位置与奖励协调。方法包含两个部分：(i) 跨模态影响引导路由（定位更新）：我们使用双向交叉注意力响应作为演化中的跨模态影响的高效代理，动态重新加权 token 感知损失，并在跨模态层间缩放梯度，而无需额外的模型干预。(ii) 偏好保持的模态感知重加权（协调奖励）：我们将预定义权重保留为偏好先验，并在预热后使用分支特定的奖励梯度交互作为残差修正。这能在不让主导奖励压制微弱但关键目标的情况下，解决演化中的冲突。大量实验表明，在模态质量、语义一致性和音视频同步方面，该方法相比强 RL 基线持续取得改进。消融实验与机制分析进一步验证了自适应更新路由与奖励协调的互补优势。
📄 English Abstract
Multi-reward guided reinforcement learning (i.e., RL) offers a promising way to improve joint audio-video diffusion models along several complementary objectives, including modality-specific quality, cross-modal semantic alignment, and temporal synchronization. Its effectiveness, however, depends on two quantities that change during training: where reward-driven updates should act, and how competing rewards should be combined. Existing methods tend to rely on fixed routing and reward weights, failing to track evolving model functions. To address these limitations, we propose Adaptive Reward Routing to jointly adapt update locations and reward coordination during forward-process RL (i.e., DiffusionNFT) of joint audio-video diffusion models. Our method consists of two components. (i) Cross-Modal Influence-Guided Routing (Localizing Updates): We use bidirectional cross-attention responses as an efficient proxy for evolving cross-modal influence, dynamically reweighting token-aware losses and scaling gradients across cross-modal layers without additional model interventions. (ii) Preference-Preserving Modality-Aware Reweighting (Coordinating Rewards): We preserve predefined weights as preference priors and use branch-specific reward-gradient interactions as residual corrections after warm-up. This resolves evolving conflicts without letting dominant rewards suppress weak but essential objectives. Extensive experiments demonstrate consistent improvements in modality quality, semantic consistency, and audio-video synchronization over strong RL baselines. Ablations and mechanism analyses further validate the complementary benefits of adaptive update routing and reward coordination.
3. 超越记忆：利用显式信念状态驾驭长时程智能体
- 英文标题：Beyond Memory: Harnessing Long-Horizon Agents with Explicit Belief States
- arXiv：2610.01415
💡 一句话解读：长时程智能体在复杂任务中容易迷失，因为仅靠记忆历史无法保证对当前世界的连贯理解。本文提出PoS框架，在推理时构建并持续维护显式的信念状态，将当前世界状态与未完成的任务需求结合起来，并检测和恢复“信念陷阱”，从而提升长时程任务表现。
大型语言模型（LLM）智能体如今能承担日益复杂的任务，但它们将交互历史组织为记忆的方式并不能确保对当前世界形成连贯的理解。我们提出了PoS，一个推理时框架，它构建并持续维护显式的信念状态作为智能体的决策上下文。每个信念将当前世界状态的估计与未解决的任务需求相结合，明确智能体仍需学习和完成的内容。为了保持该信念可靠且可操作，PoS验证其一致性并监控任务进展，以检测“信念陷阱”——即智能体继续行动却未朝着目标取得有意义进展的情况。随后，根据陷阱模式与未解决任务需求的类型定制恢复策略。在涵盖执行与诊断的四个基准测试上，实验表明PoS在所有三个LLM主干上均取得最高整体性能。消融实验证明了一致性验证与恢复的重要性，而上下文扩展实验则显示了对上下文增长的韧性。总之，这些结果支持将信念构建与持续维护作为长时程上下文管理的基础，超越单纯的历史保留与压缩。
📄 English Abstract
Large language model (LLM) agents can now undertake increasingly complex tasks, but the way they organize interaction history into memory does not ensure a coherent understanding of the current world. We introduce PoS, an inference-time framework that constructs and continually maintains explicit belief states as the agent's decision context. Each belief combines an estimate of the current world state with unresolved task requirements, making explicit what the agent still needs to learn and accomplish. To keep this belief reliable and actionable, PoS validates its consistency and monitors task progress to detect Belief Trapping, where the agent continues to act without making meaningful progress toward the goal. Recovery is then tailored to both the trapping pattern and the type of unresolved task requirement. Experiments on four benchmarks spanning execution and diagnosis show that PoS achieves the highest overall performance on every benchmark with all three LLM backbones. Ablations demonstrate the importance of consistency validation and recovery, while context-scaling experiments show resilience to context growth. Together, these results support belief construction and continual maintenance as a foundation for long-horizon context management beyond history retention and compression.
4. 智能体先验引导的策略学习
- 英文标题：Agent Priors-guided Policy Learning
- arXiv：2609.35690
💡 一句话解读：机器人看几个示范就要学会新任务，但技能组合和技能泛化互相拖累：组合时只看到技能的名字，不知道它什么时候管用。本文让每个技能在训练时带上“结构先验”（比如抓取只取决于夹爪相对物体的位姿），并把这个先验同时写进语言接口，让组合时能知道技能适用范围。方法叫 APPL，在多个任务上提升了泛化与组合能力。
从少量示范中学习的机器人通常需要两种泛化能力：组合泛化（compositional generalization）将技能重新组合以解决新任务，技能泛化（skill generalization）则让每个技能背后的策略在新情境中有效。二者相互依赖，但在组合与所调用技能之间会丢失信息：技能在何处有效由其策略训练时的结构决定，而组合只能通过名称、指令或符号算子等独立描述看到该技能，这些描述省略了结构。我们的核心思想是将每个策略的结构先验（structural prior）作为组合与技能之间接口的一部分。结构先验说明一个行为依赖于什么，例如抓取只取决于夹爪相对于物体的位姿；将其写入训练可塑造策略的泛化范围，用语言表述则能告诉组合该策略在何处适用。我们在“智能体先验引导的策略学习”（Agent Priors-guided Policy Learning, APPL）中实现了这一思想：一个构建智能体将完整示范分割为可复用技能，为每个技能提出若干结构先验，并为每个先验训练和验证一个策略；一个运行时智能体则在这些先验特定策略间选择，并利用其接口将它们组合以达成新任务目标。在 MetaWorld 和长时程 ManiSkill 任务上，APPL 提升了分布外技能泛化，并实现了此前未见的技能组合；消融接口信息会显著降低性能。这些结果支持将训练时的结构假设作为技能学习与技能组合之间的桥梁。
📄 English Abstract
Robots that learn from a few demonstrations often require two forms of generalization. Compositional generalization recombines skills to solve new tasks, and skill generalization lets the learned policy behind each skill work in new situations. The two depend on each other, yet information is lost between composition and the skills it calls. Where a skill works is determined by the structure its policy is trained with, while composition sees the skill only through a separate description, such as a name, an instruction, or a symbolic operator, that omits this structure. Our key idea is to use each policy's structural prior as part of the interface between composition and the skill. A structural prior states what a behavior depends on, for example that a grasp depends only on the gripper's pose relative to the object. Built into training, it shapes where the policy generalizes; stated in language, it tells composition where the policy applies. We instantiate this idea in Agent Priors-guided Policy Learning (APPL). A construction agent segments complete demonstrations into reusable skills, proposes several structural priors for each skill, and trains and verifies one policy per prior. A runtime agent then selects among these prior-specific policies and composes them toward new task goals using their interfaces. Across MetaWorld and long-horizon ManiSkill tasks, APPL improves out-of-distribution skill generalization and enables previously unseen skill compositions; ablating the interface information substantially reduces performance. These results support the use of training-time structural assumptions as a bridge between skill learning and skill composition.
5. 分层连续扩散语言模型
- 英文标题：Hierarchical Continuous Diffusion Language Models
- arXiv：2610.02193
💡 一句话解读：扩散语言模型并行生成文字时，各词独立抽取，丢掉了词与词之间的关联；连续扩散虽用共享状态缓解，但状态与最终词脱节。本文提出分层连续扩散语言模型，让连续隐状态成为唯一生成载体，每步从中读出词、再把词反馈回去指导下一步，从而在数独、倒计时和语言建模上超过同类模型。
离散扩散语言模型为需要双向推理和全局约束满足的任务提供了自回归生成之外的有力替代方案，但它们存在一个结构性瓶颈：并行解码时，每个词元都从其边缘分布独立采样，切断了同时解码的词元之间的统计依赖。连续扩散语言模型通过去噪一个共享的连续状态来避免这一问题，但其去噪器只能看到该状态，因此在最终解码之前，没有任何东西将该状态与合法的词元配置绑定起来。为解决这一问题，我们提出分层连续扩散语言模型（HC-DLM），它在单一且原理化的去噪过程中，将离散词元生成与连续隐轨迹耦合起来，其训练目标由词元似然的变分下界推导而来。与近期将连续上下文附加到自包含离散链上的方法不同，HC-DLM 让隐状态成为唯一持久的生成状态：每一步都从中读出词元，并将其反馈作为下一步隐状态更新的支架。在结构化推理（数独）、数学规划（倒计时）和语言建模（LM1B）任务上，在模型规模匹配的情况下，HC-DLM 在数独和倒计时的谜题准确率以及 LM1B 的生成困惑度方面均优于离散和连续扩散基线。项目页面：https://hc-dlm.github.io/。
📄 English Abstract
Discrete diffusion language models offer a compelling alternative to autoregressive generation for tasks demanding bidirectional reasoning and global constraint satisfaction. Yet they share a structural bottleneck: when decoding in parallel, each token is sampled independently from its marginal, severing the statistical dependencies among the tokens decoded together. Continuous diffusion language models avoid this by denoising a shared continuous state, but their denoiser sees only that state, so nothing ties it to a valid token configuration until it is finally decoded. To address this, we propose Hierarchical Continuous Diffusion Language Models (HC-DLM), which couple discrete token generation with a continuous latent trajectory in a single, principled denoising process, whose training objective is derived from a variational bound on the token likelihood. In contrast to recent methods that attach continuous context to a self-contained discrete chain, HC-DLM makes the latent the only persistent generative state: tokens are read out from it at every step and feed back as a scaffold for the next latent update. On structured reasoning (Sudoku), mathematical planning (Countdown) and language modeling (LM1B), HC-DLM improves over discrete and continuous diffusion baselines at matched model size, in puzzle accuracy on Sudoku and Countdown and in generative perplexity on LM1B. Project page: https://hc-dlm.github.io/.
6. World Observer：面向持久世界建模的联合行动者-观察者生成
- 英文标题：World Observer: Joint Actor-Observer Generation for Persistent World Modeling
- arXiv：2610.02162
💡 一句话解读：视频世界模型通常只从主角视角模拟环境，物体一旦离开画面就丢失了它的演化信息，等它再出现时状态往往对不上。这篇论文提出让“观察者”与“行动者”分开：除了主角视角，还生成一个或多个全景观察者持续盯着场景中的特定区域，使离开视野的物体仍在被观察、持续演化，回来时状态能对上。
世界模型如何持续观察行动者当前视野之外的区域？视频世界模型模拟环境随智能体行动而演变的过程，但始终以行动者为中心。一旦物体离开行动者视野，模型便失去其演化的直接证据，往往无法在物体重新进入视野时保持其状态与动态。为此，我们提出 World Observer，通过联合生成一个用于智能体中心视角的透视行动者，以及一个或多个观察选定世界区域的全景观察者，将观察与行动解耦。这使得离开行动者视野的物体能在观察者中持续进行视觉演化，从而在重新进入视野时反映其更新后的状态。我们通过从共享全景源进行变形（warping）来为行动者与观察者建立显式几何对应，并引入高分辨率透视参考的“观察者汇（Observer Sink）”，以在重新进入时恢复精细外观。由于观察者与行动者解耦，它们可以自由放置在场景各处、扩展到多个位置以获得更广覆盖，并可由控制信号驱动以引导视野外的演化。为评估视野外演化，我们进一步引入了世界空间度量指标以及一个涵盖真实与合成场景的基准。World Observer 在显著提升视野外动态的同时，在视觉保真度、相机控制和三维一致性方面仍具竞争力。
📄 English Abstract
How can a world model continuously observe regions beyond the actor's current view? Video world models simulate how an environment evolves from an agent's actions, yet remain actor-centric. Once an object leaves the actor's view, they lose direct evidence of its evolution, often failing to preserve its state and dynamics upon re-entry. To address this, we introduce World Observer, which decouples observing from acting by jointly generating a perspective actor for the agent-centric view with one or more panoramic observers that watch selected world regions. This allows objects that leave the actor's view to remain visually evolving in an observer, so their updated states are reflected when they re-enter. We ground the actor and observers by warping from a shared panoramic source for explicit geometric correspondence, and introduce an Observer Sink of high-resolution perspective references to restore fine appearance upon re-entry. Since the observers are decoupled from the actor, they can be placed freely across the scene, extended to multiple locations for broader coverage, and driven by control signals to steer out-of-view evolution. To evaluate out-of-view evolution, we further introduce world-space metrics and a benchmark spanning real and synthetic scenes. World Observer substantially improves out-of-view dynamics while remaining competitive in visual fidelity, camera control, and 3D adherence.
7. 后训练中的锐化税
- 英文标题：Sharpening Tax in Post-Training
- arXiv：2610.01509
💡 一句话解读：这篇论文研究大语言模型经过强化学习后训练后，是否只是“磨尖”了原有能力，导致解题多样性下降。作者发现，在智能体任务中，预训练模型配合简单推理工具，虽然单次准确率低，但多次尝试的覆盖率反而更高；后训练让任务要么总被解决要么总失败，牺牲了覆盖率。他们提出“锐化税”指标量化这种损失，并给出一种自适应采样方法，能在训练中减少该损失。
一个新兴假说认为，大语言模型（LLM）的强化学习（RL）后训练仅仅锐化了基座模型的现有行为，在提高单次准确率的同时牺牲了解空间覆盖率。尽管这种权衡已在数学和编程任务中被观察到，但它未必适用于智能体任务，因为多轮工具使用和交互可能需要后训练期间新获得的能力。我们的惊人发现是，预训练LLM配备轻量推理工具后，可以充当有能力的智能体。尽管准确率（pass@1）远低，但在足够的测试时预算下，它们在解覆盖率（pass@K）上往往超过其后训练对应模型。我们进一步分析其机制，表明后训练将任务推向两个极端：总是被解决或从未被解决，从而以解覆盖率为代价提高了采样效率和一致性。为了衡量这一代价，我们提出了锐化税（Sharpening Tax），一种量化后训练后测试时可扩展性损失的诊断指标。在来自四个家族的14对基座/后训练模型和三个智能体基准（共42个案例）中，该税在大多数设置中普遍存在，可以从少量rollout中估计，并与其他指标良好相关。最后，我们提出了后验温度组采样（PTGS），一种简单的即插即用贝叶斯采样器，根据估计的难度为每个提示调整采样温度。在两个智能体环境的RL训练中应用PTGS，其支付的税比固定温度基线更小，在重复采样下解决了更多任务，同时提高了单次准确率。
📄 English Abstract
An emerging hypothesis about reinforcement learning (RL) post-training of large language models (LLMs) is that it merely sharpens existing behaviors of a base model, improving single-shot accuracy at the cost of solution coverage. Although this trade-off has been observed in math and coding tasks, it need not extend to agentic tasks, where multi-turn tool use and interaction may require capabilities newly acquired during post-training. Our surprising finding is that pre-trained LLMs, equipped with a light inference harness, can serve as capable agents. Despite far lower accuracy (pass@1), they often surpass their post-trained counterparts in solution coverage (pass@K) given a sufficient test-time budget. We further analyze the underlying mechanism and show that post-training pushes tasks toward two extremes, always solved or never solved, and thereby improves sampling efficiency and consistency at the cost of solution coverage. To measure this cost, we propose Sharpening Tax, a diagnostic metric that quantifies the loss in test-time scalability after post-training. Across 14 base/post-trained model pairs from four families and three agentic benchmarks (42 cases in total), the tax is prevalent in most settings, can be estimated from a few rollouts, and correlates well with other metrics. Finally, we present posterior-tempered group sampling (PTGS), a simple plug-and-play Bayesian sampler that adapts the sampling temperature per prompt to its estimated difficulty. Applied during RL training in two agentic environments, PTGS pays a smaller tax than the fixed-temperature baseline, solving more tasks under repeated sampling while also improving single-shot accuracy.
8. ActiveSaddler：面向智能体执行框架优化的自动化课程学习
- 英文标题：ActiveSaddler: Automated Curriculum Learning for Agent Harness Optimization
- arXiv：2610.00906
💡 一句话解读：现有方法优化AI智能体的“执行框架”（提示、工具接口等）时，训练场景的顺序是固定不变的。但框架变了，最有用的训练场景也会变。本文提出ActiveSaddler，把训练课程也变成自动调整的：它把反复出现的失败归纳成可复用的“失败模式”，动态估计每种模式还能带来多少进步，在复习旧弱点和探索新场景之间自动平衡，让课程和框架一起进化。
自动化执行框架（harness）优化可以通过从执行反馈中迭代更新提示、工具接口和控制逻辑，显著提升大语言模型智能体（LLM agents）的表现。然而，现有方法主要优化执行框架如何更新，却在很大程度上固定了哪些训练场景产生驱动这些更新的反馈。随着执行框架的演化，对进一步优化最有用的场景可能发生变化，这表明训练课程本身也应与执行框架一同自适应。我们将这一被忽视的执行框架优化维度形式化为自动化课程学习问题，并提出ActiveSaddler。ActiveSaddler将演化中的课程建模为具有动态实例化优化目标的非平稳老虎机（non-stationary bandit），把反复出现的失败抽象为可复用的失败模式臂，估计进一步针对每种模式所能带来的潜在学习进展，并自适应地平衡重温已知弱点与探索未见场景以发现新弱点。优化结果持续更新已发现的失败模式集合及其优先级，使课程与执行框架协同演化。在GAIA2和Terminal-Bench 2.0上的实验表明，ActiveSaddler能持续发现更强的执行框架，相比使用优化前固定场景顺序的同一执行框架优化器，测试Pass@1分别提升了4.4和7.5个百分点。消融实验进一步表明，这些收益依赖于动态构建优化目标、估计其演化效用，以及平衡持续优化与新失败发现。总之，这些结果确立了自动化课程学习作为执行框架优化的一个新关键优化维度。
📄 English Abstract
Automated harness optimization can substantially improve LLM agents by iteratively updating their prompts, tool interfaces, and control logic from execution feedback. However, existing methods primarily optimize how the harness is updated while largely fixing which training scenarios generate the feedback that drives those updates. As the harness evolves, the scenarios most useful for further optimization can change, suggesting that the training curriculum itself should adapt alongside the harness. We formulate this missing dimension of harness optimization as an automated curriculum learning problem and introduce ActiveSaddler. ActiveSaddler models the evolving curriculum as a non-stationary bandit with dynamically instantiated optimization targets. It abstracts recurring failures into reusable failure-pattern arms, estimates the potential learning progress from further targeting each pattern, and adaptively balances revisiting known weaknesses with exploring unseen scenarios for new ones. Optimization outcomes continually update both the set of discovered failure patterns and their priorities, allowing the curriculum to co-evolve with the harness. Experiments on GAIA2 and Terminal-Bench 2.0 show that ActiveSaddler consistently discovers stronger harnesses, improving test Pass@1 by 4.4 and 7.5 percentage points over the same harness optimizer using a scenario order fixed before optimization, respectively. Ablations further show that these gains depend on dynamically constructing optimization targets, estimating their evolving utility, and balancing continued optimization with new failure discovery. Together, these results establish automated curriculum learning as a new crucial optimization dimension for harness optimization.
9. 可信AI审稿人缺失的一环：从修辞鲁棒性基准测试到SciCore评审
- 英文标题：A Missing Piece for Trustworthy AI Reviewers: From Benchmarking Rhetorical Robustness to SciCore Review
- arXiv：2609.39027
💡 一句话解读：AI审稿人可能因为论文措辞不同而给出不同评价，导致奖励修辞优化而非科学改进。作者提出“修辞鲁棒性”概念，并构建基准测试，发现现有审稿人存在虚假鲁棒性，于是提出SciCore双分支审稿模型，通过结合全文判断和结构化科学核心判断来降低对修辞的敏感度。
AI审稿人可能对以不同措辞报告相同科学内容的手稿给出不同判断，从而可能奖励修辞优化而非科学改进。我们将修辞鲁棒性（Rhetorical Robustness）定义为在内容保持不变的改写上保持稳定与在不同论文间保持区分度的联合要求。我们引入了RobustReview，一个包含1,260个手稿版本的受控全文基准，并评估了30种审稿人配置。该基准揭示了虚假鲁棒性（false robustness），即低改写敏感性与论文间评分崩溃同时出现，并表明人类对齐与修辞鲁棒性对审稿人的排名不同。此外，所评估的以内容为中心的提示协议并未在不同骨干模型上一致地提高鲁棒性。受这些发现启发，我们引入了SciCore，一种双分支审稿人，它将全文判断与基于提取的结构化科学核心的判断进行平均。这种设计结合了手稿级评估与旨在降低修辞敏感性的内容归一化视图。在我们主要的GPT-5.5比较中，SciCore在基准审稿人中实现了领先的联合稳定性-区分度特征，同时保持了有竞争力的人类对齐。这些结果将修辞鲁棒性确定为一个独特的评估目标，并展示了科学核心审稿在改进它方面的潜力。
📄 English Abstract
AI reviewers can assign different judgments to manuscripts that report the same science in different wording, potentially rewarding rhetorical optimization over scientific improvement. We formulate Rhetorical Robustness as the joint requirement of stability across content-preserving rewrites and discrimination across papers. We introduce RobustReview, a controlled full-manuscript benchmark with 1,260 manuscript versions, and evaluate 30 reviewer configurations. The benchmark reveals false robustness, where low rewrite sensitivity coincides with score collapse across papers, and shows that human alignment and rhetorical robustness rank reviewers differently. Moreover, the evaluated content-focused prompting protocol does not consistently improve robustness across backbones. Motivated by these findings, we introduce SciCore, a dual-branch reviewer that averages a full-manuscript judgment with a judgment based on an extracted, structured science core. This design combines manuscript-level assessment with a content-normalized view intended to reduce rhetorical sensitivity. In our primary GPT-5.5 comparison, SciCore achieves a leading joint stability-discrimination profile among the benchmarked reviewers while maintaining competitive human alignment. These results identify rhetorical robustness as a distinct evaluation target and demonstrate the potential of science-core review to improve it.
10. ROWBench：视频模型能否渲染程序所指定的内容？
- 英文标题：ROWBench: Do Video Models Render What the Program Specifies?
- arXiv：2610.02205
💡 一句话解读：现有视频生成模型（可编程世界模型）能否忠实呈现程序设定的世界事件，此前缺乏细粒度评测。作者构建了 PROWBench 基准，用 170 段程序化生成的场景和 600 段代理视频，记录实体状态与带时间戳的事件，再与生成视频对照，从实体控制、长时记忆、逻辑-渲染一致性和交互成功率等方面进行评测。
可编程世界模型将可执行动力学与视觉生成分离，为下一代游戏引擎提供了有前景的基础。然而，其视觉上对显式规则与交互的遵循程度尚未得到充分评估。现有基准测试评估视觉质量、可控性以及指令或物理遵循性，但很少检验对细粒度、程序指定的世界事件的保真度。我们提出 PROWBench，包含 170 个程序化构建的片段和 600 个代理视频，覆盖多样场景与交互。PROWBench 将实体状态和带时间戳的事件（包括相机视野之外的事件）记录为可重放的世界记录，并据此渲染同步视图与代理表示，从而使生成视频能够对照程序执行的可观测结果进行检查。一个可扩展框架可构建场景、控制行为，并能以不同表示（如粗糙 3D 和边界框）渲染每个相机视图。该基准涵盖第一人称和第三人称视角，并为部分片段提供同步多视图观测。基于这些记录，PROWBench 评估实体控制、长时程记忆，并通过两个基于视觉语言模型（VLM）的指标——逻辑-渲染对齐度（Logic-Render Alignment）和交互成功率（Interaction Success Rate）——评估对规定时间线的遵循以及对带时间戳的引擎记录事件的视觉实现。
📄 English Abstract
Programmable world models separate executable dynamics from visual generation, offering a promising foundation for next-generation game engines. However, their visual adherence to explicit rules and interactions remains insufficiently evaluated. Existing benchmarks assess visual quality, controllability, and instruction or physical adherence, but rarely test fidelity to fine-grained, program-specified world events. We introduce PROWBench, comprising 170 programmatically constructed episodes and 600 proxy videos covering diverse scenes and interactions. PROWBench logs entity states and timestamped events, including those outside the camera's field of view, as replayable world records, from which it renders synchronized views and proxy representations. This enables generated videos to be checked against the observable consequences of program execution. An extensible framework constructs scenes, controls behaviors, and can render each camera view in different representations, such as coarse 3D, and bounding boxes. The benchmark covers first- and third-person perspectives, with synchronized multi-view observations available for a subset of episodes. Grounded in these records, PROWBench evaluates entity control, long-horizon memory, and, with two VLM-based metrics, Logic-Render Alignment and Interaction Success Rate, adherence to the prescribed timeline and the visual realization of timestamped engine-recorded events.
11. AutoGUIWorld：将图像生成器用作 GUI 智能体的视觉世界模型
- 英文标题：AutoGUIWorld: Image Generators as Visual World Models for GUI Agent
- arXiv：2610.01215
💡 一句话解读：训练电脑操作智能体需要大量“操作-界面变化”的示例，但真实软件环境种类有限、装起来又麻烦。AutoGUIWorld 不运行真实软件，而是用图像生成器“脑补”界面变化、用规划器设计操作步骤，自动合成海量训练数据，让智能体在真实桌面任务上表现更好。
GUI 智能体需要高质量交互轨迹，以学习软件环境如何响应动作、维持状态并支持多步工作流。然而，可用轨迹的多样性受限于底层环境中可访问的应用、界面状态和工作流；扩大覆盖范围需要部署日益多样且复杂的软件，而专用应用还会带来额外的安装、配置和运行成本。我们提出 AutoGUIWorld，这是一个数据生成框架，它将图像生成器的视觉先验与规划器的任务知识相结合，在不部署或运行相应软件环境的情况下合成 GUI 交互轨迹。AutoGUIWorld 从操作系统上下文、视觉外观和界面状态的结构化规格中采样初始 GUI 场景，并基于这些场景生成任务；随后由规划器指定原子动作及其预期的视觉后果，而图像生成器则迭代编辑当前截图以生成后续观察。通过动作落地（action grounding）和转换级质量过滤，我们获得了覆盖 Ubuntu、Windows、macOS 和 Chrome 的 79,266 个带空间标注的步骤级训练样本。在 AutoGUIWorld 轨迹上微调 Qwen3.5-35B-A3B，使 OSWorld 上的平均任务得分从 33.0% 提升至 40.8%，ScienceBoard 上的任务成功率从 14.0% 提升至 32.2%。这些结果表明，生成的轨迹能够提升 GUI 智能体在真实桌面和科学任务上的表现。
📄 English Abstract
GUI agents require high-quality interaction trajectories to learn how software environments respond to actions, maintain state, and support multi-step workflows. However, the diversity of available trajectories is constrained by the applications, interface states, and workflows accessible in the underlying environments. Expanding this coverage requires deploying increasingly diverse and complex software, with specialized applications imposing additional installation, configuration, and runtime costs. We introduce AutoGUIWorld, a data generation framework that combines the visual priors of image generators with the task knowledge of a planner to synthesize GUI interaction trajectories without deploying or running the corresponding software environments. AutoGUIWorld samples initial GUI scenes from structured specifications of operating-system context, visual appearance, and interface state, and generates tasks conditioned on those scenes. A planner then specifies atomic actions and their intended visual consequences, while an image generator iteratively edits the current screenshot to produce subsequent observations. Action grounding and transition-level quality filtering yield 79,266 spatially annotated step-level training samples across Ubuntu, Windows, macOS, and Chrome. Fine-tuning Qwen3.5-35B-A3B on AutoGUIWorld trajectories improves the mean task score on OSWorld from 33.0% to 40.8% and the task success rate on ScienceBoard from 14.0% to 32.2%. These results show that generated trajectories improve GUI-agent performance on real desktop and scientific tasks.
12. 基于跨执行框架适配的检索增强技能优化
- 英文标题：Retrieval-Augmented Skill Optimization via Cross-Harness Adaptation
- arXiv：2609.38024
💡 一句话解读：现有方法优化AI智能体技能时，只靠反复试错，浪费了网上已有的海量技能。本文提出RASO，先检索相似技能来生成初始方案，再根据执行反馈不断检索优化，让智能体更快更好地完成任务。
智能体技能是一种可复用、可操作的自然语言产物，用于指导智能体在特定执行框架（harness）下有效完成任务。近期研究已探索智能体技能的优化，并积累了大量涵盖不同任务、领域和执行框架的公开技能。然而，尽管已有数百万公开共享的技能，现有技能优化方法在很大程度上忽视了这些积累的知识，而是仅依赖昂贵的智能体试运行（rollout）来为目标任务迭代优化技能。为解决这一问题，我们提出检索增强技能优化（Retrieval-Augmented Skill Optimization, RASO），该框架在整个技能优化过程中利用外部技能语料库作为先验知识。RASO从现有技能中检索相关知识，并通过跨执行框架适配（Cross-Harness Adaptation）将其适配到目标任务和执行框架，同时考虑领域和执行框架两方面的不匹配。RASO包含两个互补阶段：检索增强技能初始化（Retrieval-Augmented Skill Initialization, RASI）无需智能体试运行即可构建基于知识的初始技能；检索增强技能更新（Retrieval-Augmented Skill Update, RASU）则根据执行反馈引导检索外部知识，迭代优化技能。在四个智能体基准和两个模型上的大量实验表明，RASO始终优于未使用检索增强技能初始化与更新的基线方法。
📄 English Abstract
An agent skill is a reusable, actionable natural-language artifact that guides an agent to perform a task effectively under a given harness. Recent studies have explored the optimization of agent skills, contributing to a growing collection of publicly available skills spanning diverse tasks, domains, and harnesses. Despite millions of publicly shared skills, existing skill optimization methods largely overlook this accumulated knowledge, instead relying solely on expensive agent rollouts to iteratively refine skills for a target task. To address this, we propose \textbf{Retrieval-Augmented Skill Optimization (RASO)}, a framework that leverages an external skill corpus as prior knowledge throughout skill optimization. RASO retrieves relevant knowledge from existing skills and adapts it to the target task and harness via Cross-Harness Adaptation, accounting for mismatches in both domain and harness. RASO comprises two complementary stages: \textbf{Retrieval-Augmented Skill Initialization (RASI)} constructs a knowledge-grounded initial skill without requiring agent rollouts, while \textbf{Retrieval-Augmented Skill Update (RASU)} iteratively refines the skill by retrieving external knowledge guided by execution feedback. Across four agent benchmarks and two models, extensive experiments show that RASO consistently outperforms baselines without retrieval-augmented skill initialization and updating.
13. 让稀疏奖励算数：多奖励强化学习中的密度感知奖励聚合
- 英文标题：Make Sparse Rewards Count: Density-Aware Reward Aggregation for Multi-Reward RL
- arXiv：2610.00574
💡 一句话解读：多目标强化学习训练大模型时，不同目标的学习进度不均，稀疏出现的奖励信号容易被淹没。本文发现奖励的“优势能量”与它在批次中的活跃密度成正比，据此提出密度感知奖励聚合（DARA），给不常出现的奖励更大权重，从而更快学会各项要求。
多奖励强化学习用于训练大语言模型同时满足多种行为目标。GDPO 中使用的按奖励归一化能在同一批次内保留各奖励的相对信息，但不同目标的学习进度仍可能不均衡。我们通过“优势能量”（一个奖励在批次中平方优势之和）来研究这一现象。在理想化的 GDPO 归一化下，我们证明该能量与活跃组密度成正比，即该奖励在多少批次组中提供了非零相对优势。这揭示了批次层面残留的信号不平衡，并为校准各奖励贡献提供了依据。基于此关系，我们提出密度感知奖励聚合（DARA），推导出逆平方根密度校正，对不常活跃的奖励信号赋予更大权重。DARA 根据每个批次计算权重，在训练过程中自适应奖励活跃度的变化，且不改变底层策略优化目标。在工具调用和数学推理任务上的实验表明，DARA 比 GDPO 更快学到目标行为：在工具调用上最多减少 26% 的训练步数即达到高格式合规率，在数学推理上最多减少 65% 的步数即接近饱和的长度合规率，同时最终性能保持竞争力。代码见 https://github.com/zhaihaotian/DARA。
📄 English Abstract
Multi-reward reinforcement learning trains large language models to satisfy multiple behavioral objectives simultaneously. Reward-wise normalization, as used in GDPO, preserves reward-specific relative information within rollout groups, but different objectives can still exhibit uneven learning progress. We study this behavior through advantage energy, the sum of a reward's squared advantages over a batch. Under idealized GDPO normalization, we show that this energy is proportional to active-group density: the fraction of rollout groups in which the reward provides nonzero relative advantages. This reveals a residual batch-level signal imbalance and provides a basis for calibrating reward contributions. Based on this relation, we propose Density-Aware Reward Aggregation (DARA). We derive an inverse-square-root density correction that gives greater weight to signals from less frequently active rewards. DARA computes its weights from each rollout batch, adapting to changes in reward activity throughout training without modifying the underlying policy optimization objective. Experiments on tool calling and mathematical reasoning show that DARA learns the targeted behaviors faster than GDPO, reaching high format compliance in up to 26% fewer training steps on tool calling and near-saturated length compliance in up to 65% fewer steps on mathematical reasoning, while remaining competitive in final performance. Our code is available at https://github.com/zhaihaotian/DARA.
14. 去中心化 Master-Mind：多智能体路径规划中通过迭代意图去噪的联合动作精炼
- 英文标题：Decentralized Master-Mind: Joint Action Refinement through Iterative Intent Denoising in Multi-Agent Pathfinding
- arXiv：2609.32019
💡 一句话解读：多智能体在只能局部观察、需要通信的情况下找路，容易因各自独立选动作而拼出互相冲突的联合动作。DMM 把“一次性选动作”改成多轮通信中逐步修正动作意图（类似扩散模型去噪），让智能体在最终决定前先协调好，从而提升成功率并降低成本。
带通信的去中心化多智能体路径规划（MAPF）要求智能体在部分可观测条件下无碰撞地到达各自目标。基于专家数据训练的可学习策略是解决该问题的有效途径，但当同一情境下存在多个协调一致的联合动作时，从各智能体分布独立采样可能把局部有效的选择重新组合成不兼容的联合动作；即使各智能体动作分布学习正确，这种失败也可能源于最终采样机制。DMM（Decentralized Master-Mind）通过在多轮通信中用离散、迭代的动作意图精炼替代一次性动作采样来解决该问题，其灵感来自扩散模型中的去噪。智能体初始化随机动作意图，并通过局部通信进行精炼，在最终承诺前耦合各自的选择。DMM 先在专家 MAPF 解上通过模仿学习预训练，再用 MICPO（一种为多智能体、多轮动作精炼设计的无评论家组相对强化学习方法）进一步优化。DMM 通常比所评估的可学习基线取得更高的成功率和更低的解成本。在 1,600 个 MovingAI 任务上，经 MICPO 微调的 DMM 解决了 1,598 个，是所评估方法中覆盖率最高的，同时解成本接近最强基线。DMM 还可扩展到障碍物丰富环境中超过一百万个同时行动的智能体。这些结果表明，轮级意图精炼能在保持去中心化执行的同时改善联合动作协调。
📄 English Abstract
Decentralized multi-agent path finding (MAPF) with communication requires agents to reach individual goals without collisions under partial observability. Learnable policies trained on expert data provide an effective approach to this problem. However, when several coordinated joint actions are valid in the same context, independently sampling from per-agent distributions can recombine locally valid choices into incompatible joint actions. This failure can arise from the final sampling mechanism even when the per-agent action distributions are learned correctly. DMM (Decentralized Master-Mind) addresses this by replacing one-shot action sampling with discrete, iterative refinement of action intents across communication rounds, inspired by denoising in diffusion models. Agents initialize random action intents and refine them through local communication, coupling their choices before commitment. DMM is pretrained with imitation learning on expert MAPF solutions and further optimized with MICPO, a critic-free group-relative reinforcement-learning method designed for multi-agent, multi-round action refinement. DMM generally achieves higher success rates and lower solution costs than the evaluated learnable baselines. On 1,600 MovingAI tasks, DMM fine-tuned with MICPO solves 1,598, the highest coverage among the evaluated methods, while achieving solution costs close to those of the strongest baselines. DMM also scales to over one million simultaneously acting agents in obstacle-rich environments. These results show that round-level intent refinement can improve joint-action coordination while preserving decentralized execution.
15. E-MoE：面向非因子化扩散语言模型的增强混合专家
- 英文标题：E-MoE: Enhanced Mixture-of-Experts for Non-Factorized Diffusion Language Models
- arXiv：2609.37533
💡 一句话解读：掩码扩散模型逐位生成文本，各位置相互独立，导致步数少时质量差；已有方法用连续隐变量捕捉位置关联，却容易“隐变量被忽略”。本文提出 E-MoE，用专家混合骨干的专家路由决策作为离散共享隐变量，把反向过程变成多个独立分布的混合，在不增加激活参数的前提下提升少步生成质量。
掩码扩散模型（MDM）通过在每步去噪中逐步解掩若干词元来生成序列，但其反向过程通常按位置因子化，这在少步生成场景下限制了样本质量，而少步恰恰是扩散相对自回归解码速度优势最关键的区间。近期一类工作引入连续高斯隐变量并以变分自编码器训练，以捕捉位置间相关性，但这类方法易出现后验坍缩，即隐变量被悄然忽略。我们提出增强混合专家（E-MoE），将反向过程构建为对离散共享隐变量的因子化分布的混合，该隐变量由混合专家（MoE）骨干的专家路由决策给出，且相比因子化基线不增加激活参数量。在合成多模态基准、二值化 MNIST 和 LM1B 上，E-MoE 相比因子化基线改善了少步生成效果。
📄 English Abstract
Masked diffusion models (MDMs) generate sequences by progressively unmasking several tokens per denoising step, but their reverse process is typically factorized over positions, limiting sample quality in the few-step regime where diffusion's speed advantage over autoregressive decoding matters most. A recent line of work introduces a continuous Gaussian latent, trained as a variational autoencoder, to capture correlations across positions, but such approaches are prone to posterior collapse, where the latent is silently ignored. We propose Enhanced Mixture-of-Experts (E-MoE), which builds the reverse process as a mixture of factorized distributions over a discrete shared latent given by the expert-routing decisions of a Mixture-of-Experts (MoE) backbone, without increasing active parameters over the factorized baseline. Across synthetic multi-modal benchmarks, binarized MNIST, and LM1B, E-MoE improves few-step generation over factorized baselines.