Reproduce “EmbodiedRSI: Active Continual Robot Learning Through Hypothesis-Guided Co-Evolution” 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
Python Song, Zhixuan Liang, Kelsey Fu, Mengdi Wang, Junfeng Yang, Shilong Liu
EmbodiedRSI adapts a robot harness around a frozen vision-language-action model by maintaining competing code and skill hypotheses, selecting simulator experiments for expected information gain relative to cost, and retaining useful evidence. On RoboCasa365’s Composite-Unseen suite, it reports 71.3% success, versus 40.1% for Harness VLA. This supports harness-level adaptation on tested tasks, not model learning or universal gains.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.