Reproduce “Co-Evolving Robot Orchestrators and Policies through Deployment” 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
Xilun Zhang, Maggie Wang, Erik Bauer, Hong-Xing Yu, Huang Huang, Jiajun Wu, Marco Pavone
Robo-COP lets a robot improve both its vision-language-action policy and the language-model orchestrator that decides when to use it. It curates successful skill segments from mixed scripted and learned executions, selectively fine-tunes candidate policies, verifies them before adoption, and revises policy-dependent memory. Across ten simulated tasks, held-out success averaged 73.8%, versus 64.8% for a frozen policy; real-world evidence covers three tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.