Reproduce “Algebraic World Model: From Exploration to Proofs” 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
Guancheng Pan, Qixin Tan, Yang Yuan, Andrew Chi-Chih Yao
AWM learns unfamiliar game rules through play, organizes them into typed objects and shared rules, then formalizes the model in Lean. In six sealed-level runs, it proved no win existed under the tested budget in every case. These proofs concern learned models: engine checks extended five to the games, while one did not transfer.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.