OpenAI quietly funded independent math benchmark before setting record with o3
OpenAI's involvement in funding FrontierMath, a leading AI math benchmark, only came to light when the company announced its record-breaking performance on the test. Now, the benchmark's developer Epoch AI acknowledges they should have been more transparent about the relationship.
FrontierMath, introduced in November 2024, tests how well AI systems can tackle complex mathematical problems that require advanced reasoning and problem-solving skills - the kind of tasks that typically stump even the most sophisticated AI systems. The benchmark's problems were created by a team of over 60 leading mathematicians.
The connection between OpenAI and FrontierMath emerged on December 20, the same day OpenAI unveiled its new o3 model. The system achieved an unprecedented 25.2 percent success rate on the benchmark's challenging math and logic problems - a massive jump from previous models that couldn't solve more than two percent of the questions.
Epoch AI, which developed the benchmark, had signed an agreement preventing them from revealing OpenAI's financial support until o3's announcement. They acknowledged the connection in a footnote after updating their research paper for the fifth time, simply stating: "We gratefully acknowledge OpenAI for their support in creating the benchmark."
