HyperTree Proof Search for Neural Theorem Proving

Guillaume Lample, Timothee Lacroix, Marie-anne Lachaux, Aurelien Rodriguez, Amaury Hayat, Thibaut Lavril, Gabriel Ebner, Xavier Martinet

Keywords: theorem proving, automated theorem proving, MCTS, reasoning, AI for math

TL;DR: An AlphaZero-inspired method for automated theorem proving.

Abstract: We propose an online training procedure for a transformer-based automated theorem prover. Our approach leverages a new search algorithm, HyperTree Proof Search (HTPS), that learns from previous proof searches through online training, allowing it to generalize to domains far from the training distribution. We report detailed ablations of our pipeline’s main components by studying performance on three environments of increasing complexity. In particular, we show that with HTPS alone, a model trained on annotated proofs manages to prove 65.4% of a held-out set of Metamath theorems, significantly outperforming the previous state of the art of 56.5% by GPT-f. Online training on these unproved theorems increases accuracy to 82.6%. With a similar computational budget, we improve the state of the art on the Lean-based miniF2F-curriculum dataset from 31% to 42% proving accuracy.

Community Implementations: [ 4 code implementations](https://www.catalyzex.com/paper/hypertree-proof-search-for-neural-theorem/code)