Motion-Focused Latent Action Enables Cross-Embodiment VLA Training from Human EgoVideos figure
AlphaXiv 中文概览(可滚动查看)