LLMs and Mathematics

With the recent advanced in Large Language Models, it has become increasingly clear that they will play a pivotal role in the future of mathematics research. With this in mind, we intend to explore new ways of aiding mathematics researchers using LLMs.

We are currently focused on generalizing alpha geometry to other fields of mathematics like analysis and algebra, improving the Funsearch algorithm by DeepMind, and also fine tuning LLMs on specialized Mathematica packages like Ricci.

For more information on our team members, see our members page.

News

• The Funsearch group has been able to completely replicate the results of DeepMind's Funsearch algorithm. They are now working on a project that they've called "Darwin" to further improve the efficiency of the algorithm. Here is the GitHub repository.
• The Alpha Geometry group, along with a Google software engineer, have also been able to completely replicate the results of DeepMind's Alpha Geometry algorithm. They are now trying to generalize it to other mathematical fields like algebra and analysis.