Reproduce “Reasoning on Call: Efficient Agentic Robot Control for Long-Horizon Robot Manipulation” 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
Zilang Cen, Chao Tang, Fangcheng Liu, Weishi Mi, Wei Hao, JiaYu Xu, Dehao Huang, Yehui Tang and 3 more
RoC coordinates long-horizon robot manipulation by placing an online monitor between a low-level policy and a reasoning agent. The monitor calls the agent when progress or a possible task transition needs review, allowing execution repair and plan revision. Across nine simulated tasks, RoC averages 99.6% success and uses 65.3% fewer tokens than GPT-as-Policy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.