Reproduce “RoboACE: Adversarial Co-Evolution for Robot Self-Improvement” 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
Delin Zhao, Zhengrong Yue, Junlin He, Yali Wang
RoboACE alternates revisions to a robot’s curriculum and execution harnesses around frozen foundation-model weights. Its curriculum harness uses capability evidence to select practice, while its execution harness tests reusable changes against fixed task snapshots. In simulation, the deployment scored 74.0% on LIBERO-PRO, versus 59.0% initially and 60.5% for direct π0.5 control; this development trajectory does not isolate curriculum’s effect.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.