Reproduce “EgoJEPA: Beyond Retargeting for Dexterous Manipulation from Egocentric Videos” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Meng Cao, Can Zhao, Jinjie Wei, Haonan Zhao, Yulong Zhang, Duo Han, Xiaofan Li, Ziwu Song and 7 more
EgoJEPA transfers human hand-motion knowledge to dexterous robots by learning from annotated hand movements and their effects on video, then adapting a shared action representation with robot-specific adapters. Across five real-world tasks and three embodiments, it averages 74% out-of-domain task success, compared with 65% for retargeting. This result is limited to the tested tasks and conditions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.