Reproduce “PyRUA-REAL: Robot Exploration and Adaptive Learning through RSI in the Real World” 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
Qiang Wang, Jianxin Bi, Shulong Jiang, Ruiyang Si, Wenbo Huang, Kaiqi Chen, Jiaxin Li, Liyun Li and 2 more
PyRUA-REAL turns physical robot trials into reusable programs through a researcher agent that selects experiments and an engineer agent that writes, tests, and revises code, retaining useful experience for later tasks. Frozen programs use live observations and recovery routines without runtime language-model calls. They average 84.42% success across 42 RoboDojo tasks, but this is a separate evaluation, not a matched-seed comparison.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.