今日速览
| 序号 | 标题 | 来源 | 日期 | 主题 | 推荐等级 |
|---|---|---|---|---|---|
| 1 | Dex-One2Many: Learning Dexterous Manipulation from a Single Human Demonstration | arXiv | 2026-10-08 | AI | 中 |
| 2 | Error-Propagation Modeling for Failure Attribution in LLM-Based Multi-Agent Systems | arXiv | 2026-10-08 | AI-Agent | 高 |
重点论文与技术动态
1. Dex-One2Many: Learning Dexterous Manipulation from a Single Human Demonstration
- 来源：arXiv
- 日期：2026-10-08
- 作者/机构：Jusuk Lee, Sungha Kim, Yeonsoo Park, Jonguk Cheon, Yoonkyo Jung, Yongjun You
- 主题标签：
AI,arXiv - 推荐等级：中
- 分类：cs.RO, cs.CV
一句话结论
提出Dex-One2Many框架，将单个人类视频抽象为序列场景图以引导强化学习，实现灵巧手操作的高效探索与强泛化。
核心内容
- 解决运动模仿泛化受限及纯强化学习在多阶段任务中高维探索困难的问题。
- 用序列场景图约束相对关系而非精确位姿，以采样多样化重置状态并提供密集奖励。
- 纯仿真训练可零样本迁移至真实多指手，5项任务中已知和未知场景分别超基线6.5%和71%。
方法与数据
- 场景图引导RL的real-to-sim-to-real框架；仅用单个人类视频，仿真器与硬件细节摘要未明确。
价值判断
- 值得关注：未知场景性能超基线71%，且实现零样本真机迁移。
- 可复用点：以场景图约束关系来初始化状态并提供密集奖励的短探索机制。
- 局限/待核查：场景图的提取方式摘要未明确。
摘要
While learning dexterous manipulation from a single human video offers a promising alternative to costly robot demonstrations, many recent methods predominantly imitate demonstrated motions. Such strict motion matching often limits generalization to initial object poses, goal poses, and grasps not shown in the video. Alternatively, discovering a policy via reinforcement learning (RL) allows for broad generalization, but without prior guidance, it struggles with high-dimensional exploration in complex, multi-stage tasks. To address these coupled generalization and exploration challenges, we present Dex-One2Many, a real-to-sim-to-real framework that learns a generalizable dexterous manipulation policy from a single human video. Our key insight is to abstract the video into sequential scene graphs that guide RL, enabling efficient exploration while preserving broad generalizability. The graphs serve as generative constraints for sampling diverse reset states and provide dense rewards for each stage. Because the graphs constrain relations rather than exact poses, these reset states cover object poses and grasps beyond the video, while initializing each stage from them with dense rewards keeps exploration short and guided. Trained entirely in simulation, Dex-One2Many transfers zero-shot to a real multi-fingered hand. Across five tool-use and manipulation tasks, Dex-One2Many exceeds baselines by 6.5% in seen configurations, while its robust generalization widens this gap to 71% in unseen scenarios.2. Error-Propagation Modeling for Failure Attribution in LLM-Based Multi-Agent Systems
- 来源：arXiv
- 日期：2026-10-08
- 作者/机构：Jiaqi Liao, Yuanzhao Zhai, Huanxi Liu, Xu Zhang, Zheming Zhuang, Dawei Feng
- 主题标签：
AI-Agent,arXiv - 推荐等级：高
- 分类：cs.AI
一句话结论
本文提出错误传播建模方法（EMFA），通过有效区分根本错误与下游症状，在LLM多智能体系统失败归因中实现了最先进的步骤级准确率。
核心内容
- 将归因目标定义为修正后能恢复失败执行的决定性“智能体-步骤”对。
- 解决现有方法忽略错误跨交互传播及持续死循环的问题，精准区分根因与症状。
- 在基准测试中取得SOTA步骤级归因准确率，并在智能体级别保持竞争力。
方法与数据
- 方法：构建失败轨迹的结构化表示以建模级联传播与持续交互循环，结合传播感知的候选筛选与反事实验证来定位错误。
- 数据：采用 Who&When 基准测试，在手工构建和算法生成两个子集上，步骤级准确率分别提升3.45和4.40个百分点。
价值判断
- 值得关注：显式建模多智能体交互中的错误级联传播与持续循环。
- 可复用点：“结构化轨迹+传播感知筛选+反事实验证”的归因流水线。
- 局限/待核查：反事实验证所需的额外计算开销与耗时，摘要未明确。