Meta’s Fundamental AI Research (FAIR) team, supported by Superintelligence Labs, has utilized AI models to resolve six open research problems in mathematics. The company released these findings in a series of six research papers, emphasizing the role of AI as a collaborative tool for mathematicians.
A key achievement occurred on October 2, 2026, when AI helped disprove a conjecture by researchers García-Martínez and Pérez-Rodríguez regarding solvable evolution algebras. Additionally, on September 23, 2026, Meta published a preprint titled “Learning to Discover Interesting Mathematics,” focusing on using large language models to generate theorems.
In May 2026, Meta introduced AutoformBot, a system that converted 26 mathematics textbooks into the Lean 4 proof assistant language. This initiative resulted in a library containing over 45,000 declarations and hundreds of thousands of lines of verified code, aimed at automating formalization.
The broader field of AI-assisted mathematics has seen significant growth in 2026, with contributions from institutions like OpenAI and Google DeepMind. Many of these projects utilize formal verification systems like Lean to ensure the accuracy of AI-generated results, a practice Meta also employs in its ongoing research.
