Reproduce “RACaP: Agentic Reasoning, Acting, and Coding as Policies for Evolvable Robot Learning” 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
Zexi Li, Yehang Zhang, Haojian Huang, Bohan Zhou, Wenqian Li, Chenxu Wang, Yifan Chang, Yangkai Wei and 5 more
RACaP evolves reusable Policy APIs, runtime ReAct procedures, and experience memory before deployment, then uses a visual agent to select typed calls and adapt arguments, recovery, and stopping without generating code. In simulator benchmarks, frozen Phase 2 achieved 45.0% native success (81/180) on zero-shot LIBERO-PRO, versus 17.8% for RATS-base, but deployment on physical robots has not been demonstrated.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.