Reproduce “Pro$^2$ Agent: Scientific Problem Probing for Autonomous AI R&D” 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

Yixiong Chen, Liqian Peng, Tian Tan, Ziqian Lin, Blake Jianhang Chen, Chao Zhao, Yaojie Liu, Alec Go and 1 more

Pro² Agent helps solve machine-learning competitions by tracking suspected performance bottlenecks, evidence, candidate fixes, and confidence in a revisable Problem State. It tests hypotheses through experiments and updates that state as results arrive. On MLE-Bench, it earns medals on 70.7% of tasks in 24 hours and 76.0% in 48; these comparisons are not controlled tests of the state alone.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.