Reproduce “REA: A Routed Embodied Agent Runtime for Asynchronous Brain-Cerebellum 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
Pengfei Zhou, Yixing Ma, Zhengfeiyang Zhang, Haoyuan Chen, Zhenglin Wan, Hexin Wang, Wangbo Zhao, Yang You
REA coordinates robot policies asynchronously, using a router to select policies and allowing them to act without waiting for reasoning. On 3,000 episodes across 60 LIBERO-Pro tasks, REA-M achieved 95.9% success, versus 94.7% for Agents SDK plus GPT-6, with 25% fewer control steps per success. The comparison covers selected tasks and perturbations, with matched policy programs on 37 tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.