Reproduce “ECHO: In-Context Behavior foundation model for Humanoid Whole-Body Control” 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
Xingyi Wang, Bin Li, Jiyuan Shi, Chenyun Zhang, Weiji Xie, Xiangyu Miao, Junsong Wu, Wei Song and 2 more
ECHO adapts humanoid motion tracking to changing physical conditions by using recent state-and-action history to infer context and correct a frozen policy, without test-time parameter updates. In HDR simulation on AMASS-2000-Random, ECHO-H raises success rate by 0.95 percentage points and lowers global MPJPE by 3.0% versus RMA-Teacher, but loses on root-relative MPJPE; the result is limited to this comparison.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.