Reproduce “$λ_0$: Scaling Open-World Egocentric Human Data for Humanoid Loco-Manipulation” 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

Lambda-humanoid Team, Chongyang Xu, Jin Chen, Jinhui Ye, Yilun Chen, Yiming Jiang, Zhao Wu, Hu Wei and 8 more

λ₀ trains humanoid policies using human activity data converted into robot-compatible whole-body supervision, followed by robot demonstrations. HUMANVERSE pairs egocentric video with tracked body motion and reconstructed hand motion, while alignment adjusts actions for the G1. Across four tasks, λ₀ succeeded in 25 of 40 trials and reached 80.4% weighted progress, but uses robot post-training and results cover the setup.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.