Reproduce “AutoFyn Technical Report: Non-Parametric Expert Iteration for Long-Horizon Agents” 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
Adib Hasan, Daniel Schaffield, Akashnil Dutta, Tarik Adnan Moon
AutoFyn tackles long-running agent tasks by starting fresh model sessions, carrying selected written state forward, and using task-environment checks to evaluate candidates. It retains an accepted artifact when its comparable score improves. On six 2026 IMO problems, GLM 5.2 scored 32.0 versus 20.7 in its provider’s coding agent; differing deployments mean this comparison does not isolate AutoFyn’s effect.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.