Reproduce “FormaTheoria: Constructing Large-Scale Lean Theories from Mathematical Literature $-$ Toward the Formalization of the Classification of Finite Simple Groups” 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
Tianjiao Nie, Ao Zhang, Yusen Tang, Damiano Testa, Shing-Tung Yau, Peng Li, Yuan Zhou
The FORMATHEORIA workflow provides an AI-assisted method for constructing large-scale Lean mathematical theories from distributed literature. It successfully formalized major components of the Classification of Finite Simple Groups (CFSG), resulting in over 994,000 lines of machine-checked Lean code for theorems like Feit–Thompson Odd Order and Glauberman’s Z*.
Key insights
- Large-scale formalization from distributed literature requires explicit mechanisms for discovering evolving dependency contexts and reconciling misaligned definitions and conventions across heterogeneous sources.
- Iterative, diagnostic review processes are crucial for ensuring semantic fidelity in translating natural language mathematics to formal systems, as kernel verification alone is insufficient.
- Persistent graph-based state tracking and context compaction are essential for managing long-horizon AI tasks involving numerous interactions and exceeding single LLM context windows, as demonstrated by trajectories spanning over 9 days.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.