Reproduce “SwarmRSI: Recursive Self-Improvement through Evolving Scientific Agent Swarm” 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

Qianru Meng, Xi Zhang, Konstantinos Papakostas, Congfeng Cao, Wenxiang Hu, Tianyi Lu, Xiao Zhang, Lun Zhou and 4 more

SwarmRSI helps scientific research agents reuse hypotheses, programs, and evidence across attempts by storing them in a shared research graph. Agents use feedback to revise hypotheses or repair implementations, while rebase and crossover adapt or combine program components. On NatureBench, it reached 44.4% Match-SOTA versus Arbor’s 36.7%, under the reported setup, not establishing general scientific improvement.

Summary by alphaXiv.

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