Reproduce “DiscoveryAgent: A Framework for Scientist-centric Agentic AI” 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

Owen Queen, Ioana Ciuca, Emma Dann, Jean Somalwar, Carine You, Steven Dillmann, Viraj Manwadkar, Evan Gorrell and 4 more

DiscoveryAgent lets agents investigate scientific questions while scientists steer between rounds and audit agents inspect recorded work. It checkpoints actions and artifacts so claims can be traced and analyses reproduced. In a one-hour review of an intentionally corrupted paper, audit-group participants recovered 3.50 planted errors on average, versus 0.25 with paper and code only; this tests planted-error detection, not scientific correctness.

Summary by alphaXiv.

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