Reproduce “ASPIRE: Agentic /Skills Discovery for Robotics” 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
Runyu Lu, Yubo Wu, Ethan Kou, Haotian Lin, Letian Fu, Wenli Xiao, Ajay Mandlekar, Yinzhen Xu and 8 more
ASPIRE helps robot-programming agents improve by diagnosing failed programs and preserving verified repair guidance. It logs step-level evidence from perception, planning, and control, then repairs and reruns programs, with evolutionary search over alternatives. In simulation, average LIBERO-Pro success rises from 14% without the execution engine or search to 72% with both; zero-shot transfer reaches 30.5%, not reliable success on every task.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.