Reproduce “Open-HumanWAM: An Open Recipe for Scaling Human and Robot Experience for Embodied Foundation Models” 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
Juncheng Ma, Jiaxin Li, Jiankai Tu, Bo Liang, Guodong Lv, Yufan Deng, Junzhe Tang, Liangzhu Yang and 25 more
Open-HumanWAM studies whether curated first-person human videos can improve robot manipulation. It aligns hand motion and language across video sources into a shared action representation, then tests robot post-training and hybrid, goal-conditioned control. In simulation, scene-OOD success rises from 49.1% at 100 hours to 73.2% at 50,000 hours with IDM; this empirical result covers tested settings only.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.