Dex-One2Many Learns Dexterous Manipulation from Single Human Video
October 8, 2026
Dex-One2Many uses a real-to-sim-to-real framework to learn generalizable dexterous manipulation policies from a single human demonstration. By abstracting video into sequential scene graphs, the method overcomes the exploration challenges typical of reinforcement learning in high-dimensional tasks.
HOW THIS AFFECTS YOU
●
researcherYou can bridge the gap between imitation learning and reinforcement learning for complex robotic tasks.