[论文] ThinkWorld: World Imagination Reinforces Chain-of-Dynamics Reasoning

论文概要

研究领域: 机器人 / VLA 作者: Chaehyun Kim, Chi-Pin Huang, Jisoo Kim, Min-Hung Chen, Yu-Chiang Frank Wang, Seungryong Kim, Fu-En Yang 发布时间: 2026-10-27 arXiv: 2610.37666

中文摘要

视觉-语言-动作（VLA）模型在机器人操作领域取得了显著进展，但其泛化能力仍受限于带动作标注的机器人演示数据的多样性和覆盖范围。人类视频提供了更广泛的行为和环境经验，但从中学习面临可扩展数据与显式动作接地之间的根本性权衡。本文提出ThinkWorld框架，将动作推论锚定在世界动力学上，并通过世界想象强化学习来增强推理能力。ThinkWorld将每个动作思维表示为Chain-of-Dynamics（CoD）——一系列离散运动token，捕捉从当前世界到未来状态的中间动力学过程，作为跨本体的通用动作表示。团队首先用带标注的演示初始化CoD，然后预测每个CoD引起的未来状态，并强化那些预测与观测结果更匹配的动作。这使得无动作标注的视频能从观测状态转移中提供直接监督，让人类和机器人经验在同一推理空间中相互增强。仿真和真实机器人实验均证明了有效的人到机器人迁移能力，以及对未见过任务和环境的强泛化性。

原文摘要

Vision-Language-Action (VLA) models have made substantial progress in robotic manipulation, but their generalization remains constrained by the diversity and coverage of action-labeled robot demonstrations. Human videos provide much broader behavioral and environmental experience, while learning from them faces a fundamental trade-off between scalable data and explicit action grounding. We introduce ThinkWorld, a VLA learning framework that grounds action reasoning in world dynamics and strengthens it through reinforcement learning with world imagination. ThinkWorld represents each action thought as a Chain-of-Dynamics (CoD), a sequence of discrete motion tokens that captures the intermediate dynamics connecting the current world to a future state and serves as a common action representation across embodiments. We first initialize CoD with action-labeled demonstrations, then predict the future induced by each CoD and reinforce those whose predictions better match the observed outcome. This enables videos without action annotations to provide direct supervision from observed state transitions, allowing human and robot experience to refine the same reasoning space. Experiments in simulation and on real robots demonstrate effective human-to-robot transfer and strong generalization to unseen tasks and environments.

---

*自动采集于 2026-10-29*

#论文 #arXiv #AI #小凯