Reproduce “PGX: Physically Grounded Execution for Long-Horizon 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
Kai Lu, Chenyang Ma, Ioannis Havoutis
PGX coordinates long-horizon humanoid loco-manipulation by using an external executor for task stages and spatial references while a vision-language-action policy handles manipulation. Its predicate-gated state machine and world-coordinate locking manage transitions and stance. In simulated coffee fetching, GR00T N1.6 success rose from 35% to 80% with fixed policy weights and instruction; this comparison does not isolate individual components.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.