# Formal Mathematics Statement Curriculum Learning

**Authors:** Stanislas Polu, Jesse Michael Han, Kunhao Zheng, Mantas Baksys, Igor Babuschkin, Ilya Sutskever  
**arXiv:** [2202.01344](https://arxiv.org/abs/2202.01344)  
**Date:** 2022-02-03  
**Summary page:** https://www.summarizepaper.com/en/paper/2202.01344v1/

## Key points

- Authors explore the use of expert iteration in language modeling for formal mathematics
- Expert iteration significantly outperforms traditional proof search methods with the same compute budget
- Expert iteration can autonomously navigate through a range of mathematical concepts and complexities
- State-of-the-art performance achieved on challenging high school olympiad problems using a manually curated set of problem statements
- Expert iteration enhances language modeling for formal mathematics by combining proof search and learning
- Potential impact in advancing mathematical reasoning and problem-solving capabilities

## Summary

In their paper "Formal Mathematics Statement Curriculum Learning," authors Stanislas Polu, Jesse Michael Han, Kunhao Zheng, Mantas Baksys, Igor Babuschkin, and Ilya Sutskever explore the use of expert iteration in language modeling for formal mathematics. They demonstrate that this approach significantly outperforms traditional proof search methods when given the same compute budget. By applying expert iteration to a diverse collection of formal statements with varying levels of difficulty, the authors show its ability to autonomously navigate through a range of mathematical concepts and complexities. Additionally, they achieve state-of-the-art performance on challenging high school olympiad problems using a manually curated set of problem statements. This research highlights how expert iteration can enhance language modeling for formal mathematics by combining proof search and learning while also showcasing its potential impact in advancing mathematical reasoning and problem-solving capabilities. is enhanced by , which combines  and . The paper titled "Formal Mathematics Statement Curriculum Learning" explores this approach's effectiveness in solving increasingly difficult problem sets without relying on pre-existing proofs. The authors demonstrate its superiority over traditional proof search methods when given the same compute budget and achieve state-of-the-art performance on challenging high school olympiad problems using a manually curated set of problem statements.