Yao Mu

Fine-grained metrics reveal that robots can recover from failures more effectively than previously thought, reshaping our understanding of their capabilities.

Current video generation models fail to maintain embodiment consistency and functional interaction in human-to-robot manipulation, revealing significant gaps in their transfer capabilities.

Integration of diverse robot policies can be streamlined from hours to minutes, revolutionizing how we deploy and evaluate robotic systems.

Current vision-language models struggle with process understanding in robotic manipulation, but targeted post-training can yield significant improvements.

Achieving 86.4% grasp stability and 83.3% real-world success, SynManDex bridges the gap between human dexterity and robotic manipulation.

AHA-WAM achieves a remarkable 92.80% success rate on RoboTwin while executing actions at 24.17 Hz, all without the need for prior robot-data training.

IDP achieves high-frequency robot control by enforcing action manifold constraints without the computational burden of iterative sampling.

Human-in-the-loop chunk-wise residual adaptation closes the reality gap for dexterous robot manipulation, boosting success rates by up to 43% compared to offline imitation learning.

Decoupling high-level VLM planning from low-level diffusion-based control lets robots reason like foundation models *and* execute precisely, outperforming end-to-end approaches in complex manipulation tasks.

Forget painstakingly creating 3D assets for robot training - ManiTwin automates the process, turning single images into simulation-ready objects at scale.