Reproduce “BlinkMimic: Training a Humanoid Motion Tracker in Minutes on a Single GPU” 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
Fuyi Chen
BlinkMimic accelerates training a simulated Unitree G1 policy for all 40 LAFAN1 clips using a progress-triggered step-size schedule, deployment-condition failure probing to guide resets, and GPU-graph capture. On one RTX 4090, it reaches three consecutive clean checkpoints in 3.38 minutes on average, versus 7.4 minutes for InstantTrack v1.2. The evidence is limited to one robot and motion library in simulation, not real-robot performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.