Yushen Zuo

UMI-Bridge, which uses UMI as an intermediate domain to align representations according to action equivalence rather than pixel similarity, supports action-anchored latent alignment for data-efficient robot learning and UMI-to-robot task transfer.

Action-conditioned objectives can significantly enhance the effectiveness of Euclidean-cost MPC by aligning latent representations with real task progress.

Users can alter video prompts on-the-fly, enabling a seamless interactive experience in long-form video generation.

Zero-robot post-training on high-fidelity UMI data achieves real-world performance that rivals traditional teleoperation methods, revolutionizing data efficiency in robotic manipulation.