Event
Paper authors propose ThinkWorld VLA learning framework
2 reports · 2 sources
The paper's authors propose a vision-language-action (VLA) learning framework named ThinkWorld, which represents each action thought as a string of discrete motion tokens (chain of dynamics, CoD), used to characterize the intermediate dynamics from the current world to future states and to serve as a cross-embodiment general action representation. The authors state that the framework first initializes the CoD with action-annotated demonstrations, then predicts the future induced by each CoD, and reinforces those CoDs whose predictions match the observed outcomes more closely. The authors state that simulation and real-robot experiments show that the framework can achieve effective human-to-robot transfer and has strong generalization ability to unseen tasks and environments.
Report timeline
- ThinkWorld: A VLA Learning Framework That Reinforces Chain-of-Dynamics Reasoning with World Imagination
ThinkWorld is a vision-language-action (VLA) learning framework that 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.
arXiv cs.RO ·
- ThinkWorld: A VLA learning framework that reinforces chain-of-dynamics reasoning with world imagination
ThinkWorld is a VLA learning framework that represents each action thought as a Chain-of-Dynamics (CoD), a sequence of discrete motion tokens that captures the intermediate dynamics from the current world to a future state and serves as a common action representation across embodiments.
HuggingFace Daily Papers ·