Reproduce “DiscoveryRSI: Recursively Learning to Discover Valuable Hypotheses” 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
Jianyuan Zhong, Chenwei Wu, Zhaochen Yu, Xixian Jiang, Xuewei Li, Hao Xiao, Xilin Dai, Junyang Wang and 15 more
DiscoveryRSI records hypotheses, evidence, decisions, and agent contributions, then uses that history to improve later research. It develops hypotheses with an evidence-grounded claim, a test plan, decision criteria, and alternatives, while compiling selected computational plans into tasks. Across 241 questions in four life-science domains, automated mean quality rises from 4.86 to 5.39 after two rounds, mainly through testability gains; scientific claims remain unvalidated.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.