Reproduce “SwarmEvolve: Latent Agent Swarms for Evolutionary Discovery and Optimization” 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
Qile Jiang, George Em Karniadakis
SwarmEvolve uses many agents to search for scientific and mathematical programs, sharing internal model memories instead of lengthy text. A synthesizer writes candidate programs, which an evaluator scores. Across 21 problems, Qwen3 8B averaged 0.804, versus CodeEvolve’s 0.700, in 60-minute runs. Its median time to reach OpenEvolve’s 60-minute score was 7 minutes, a time-to-target result, not a universal win.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.