Jin Xu

A single visual interface, action views, bridges the gap between video generation models and diverse robotic embodiments, enabling unified action modeling and execution.

Trie-GRPO, a novel reinforcement learning algorithm based on action prefix trees, which enables step-level advantage estimation, is introduced, which resolves the credit assignment problem by isolating intermediate correct decisions from downstream errors, while effectively balancing exploration efficiency and depth compared to conventional search trees.

Achieving an 85.5% success rate in robotic manipulation while accelerating action generation by 2.7 times could redefine efficiency in VLA systems.

Multimodal models can now achieve state-of-the-art performance in real-world tasks like document understanding and audio-video comprehension with significantly reduced inference latency thanks to novel token-reduction techniques.