OpenAI Acquires Access to FrontierMath Validator for AI Math Problem Verification

ME News reports that on April 15 (UTC+8), according to monitoring by Beating, the AI research institute Epoch AI disclosed that OpenAI has acquired access to its FrontierMath: Open Problems validator. FrontierMath: Open Problems is a set of advanced mathematical problems that have yet to be solved by professional mathematicians; each problem is accompanied by a dedicated computer program (validator). Although no one currently knows the correct answers, once a potential solution is provided—by a human or an AI—the validator can automatically verify its correctness. After acquiring this access, OpenAI can use the validator to check whether mathematical solutions generated by its own models are valid. Access to the validator is available for purchase by any organization, with the primary costs going toward compensating mathematicians, as developing the problems and writing the validators is highly labor-intensive. Epoch AI has also set a condition: any party that discovers a valid solution through the validator must notify Epoch AI, and the original problem author and the discoverer will share joint authorship rights. Epoch AI emphasized that OpenAI previously funded the creation of the original FrontierMath benchmark (Tiers 1–4), but the Open Problems component was independently developed and is owned by Epoch AI, with initial funding provided by Schmidt Sciences. (Source: BlockBeats)