Reproduce “Tokens of Motion: Primitive Motion Learning for Human-Humanoid Skill Transfer” 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

Hanwen Ren, Meng Zheng, Wenzhe Cui, Benjamin Planche, Zhongpai Gao, Anwesa Choudhuri, Terrence Chen, Ziyan Wu

M-PACT represents human hand motion as reusable local primitives and adapts them to a robot by refitting an embodiment-specific decoder while keeping the representation fixed. On HumanoidEveryday, the refitted decoder reconstructs motion with median translation error of 3.442 cm and rotation error of 0.901°. These offline results do not establish task success; aperture error does not beat predicting no change.

Summary by alphaXiv.

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