Reproduce “HuRo: Robotizing Human Videos for Scalable VLA Pretraining” 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
Jinho Jeong, Se June Joo, Jaehyun Kang, Dongyun Kim, Yena Kim, Hanjung Kim, Seon Joo Kim
HuRo turns human videos into robot-form episodes for pretraining vision-language-action policies, addressing the mismatch between human hands and robot inputs and actions. It retargets motion and replaces human arms with a robot. On four ALLEX tasks, completion rises from 51.5% to 80.3% overall and from 34.9% to 72.2% OOD after real-robot fine-tuning, but labels are estimated kinematics, not executable demonstrations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.