Reproduce “automodel: A Skill for Agentic Model Discovery” 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

Marco Virgolin, Harry Desmond, Ester Bergantin, Giorgio Morales, John W. Sheppard, Luigi Romano, Alessandro Lucantonio

automodel provides reusable instructions and templates for a coding agent to search scientific model structures within researcher-defined constraints. For the galaxy stellar mass function, its A5 fit has description length 36.5 nats below Bernardi’s on the Sérsic dataset, using 19 operators rather than 23. This is a fit to full binned data, not a held-out predictive test.

Summary by alphaXiv.

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