Reproduce “Scaling Online Exploration for Robot Agents” 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
Runhan Huang, Xiangcheng Zhang, Haoyu Chen, Yitong Li, Yen-Jen Wang, Bohan Lyu, Haotian Ye, Chuye Hong and 12 more
RAX lets robots learn tasks by exploring actions and turning successful decisions into reusable programs. A frozen language model observes outcomes, revises code for perception, planning, and control, without updating its parameters or training a policy. The paper reports a 37.4-percentage-point average margin over the strongest prior learned model across five selected simulation comparisons, not all benchmarks or robot agents.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.