NeurIPS Paper Proposes Theoretical Framework for Prover-Conjecturer Self-Play in LLM Theorem Proving

Stanford AI Lab highlighted a NeurIPS 2026 poster paper by Thomas Chen and Zhiyuan Li that proposes a theoretical framework for understanding when prover-conjecturer self-play enables LLMs to self-improve in theorem proving, inspired by STP (Dong & Ma, 2025).

2026-11-09 ~ 2026-11-09 · 2 related posts

• New theory shows prover-conjecturer self-play can grow proved theorems exponentially — StanfordAILab · 2026-11-09
• Theory framework explains when prover-conjecturer self-play lets LLMs self-improve at theorem proving — stanfordnlp · 2026-11-09