Reproduce “RoboRecover: Benchmarking Robot Policy Recovery under Execution Deviations” 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
Yang Li, Chen Zhao, Zhuoran Wang, Jiankang Wang, Chao Shao, Yihan Lin, Haitao Shen, Jing Zhang
RoboRecover evaluates whether robot policies can resume manipulation tasks after prior actions alter the scene. It reconstructs off-nominal intermediate states by replaying action prefixes and measures Overall Recovery Success Rate (RSR), averaging completion rates equally across scenarios. On LIBERO, UnifoLM leads initial-state success at 98.83%, while π0.5 leads recovery at 64.40%, versus 48.00% for UnifoLM; evidence is limited to simulated scenarios.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.