About

The 2025 FrontierMath Symposium is the capstone of Epoch AI’s FrontierMath Tier 4 problem set project. Over the course of this two-day retreat, several dozen mathematicians will gather to discuss AI’s role in the future of mathematics, and to finalize a collection of problems that will serve as a yardstick on AI’s progress towards autonomous research capability across the major fields of mathematics.

Accepted participants will be allotted a generous stipend to cover travel expenses for the weekend of the symposium, with the expectation that they prepare in advance a problem which tests models’ mastery of a particular technique or method from their own research. At the venue, they will discuss their problem design with the colleagues nearest to them in research area and collaborate to produce strengthened versions of their problems resistant to current AI problem-solving strategies.

A panel of judges representing several domains of mathematics will assess the problems within their field, including both those submitted by retreat participants as well as those that have been written over the course of the Tier 4 project as a whole. They will address the participants regarding the rigor and diversity of the problem set we produce, and award the authors of the best problem, factoring originality, depth of solution, and their expectation of which problems will resist AI solution the longest.

Contact information

Mathematical inquiries: Elliot Glazer

elliot@epoch.ai

Logistical inquiries: Tzu Kit Chan

tzu@epoch.ai

Supported by OpenAI

Epoch AI gratefully acknowledges OpenAI for financially supporting FrontierMath. OpenAI has access to the FrontierMath datasets for evaluations, except for a set of held-out problems to allow for Epoch AI’s independent evaluations. Epoch AI runs frequent public evaluations on a wide range of frontier models from across the AI ecosystem to determine the state of the art of mathematical capabilities.