{"paper":{"title":"FrontierMath: A Benchmark for Evaluating Advanced Mathematical Reasoning in AI","license":"http://creativecommons.org/licenses/by/4.0/","headline":"FrontierMath shows that current AI models solve under 2% of hundreds of original expert-level mathematics problems.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Alex Gunning, Anson Ho, Bogdan Grechuk, Caroline Falkman Olsson, Diego Chicharro, Ege Erdil, Elizabeth Pratt, Elliot Glazer, Emily de Oliveira Santos, Evan Chen, Grant Barkley, Jaime Sevilla, Jean-Stanislas Denain, Lionel Levine, Mark Wildon, Matej Vrzala, Matthew Barnett, Natalie Stewart, Olli J\\\"arviniemi, Qiuyu Ren, Robert Sandler, Shreepranav Varma Enugandla, Tamay Besiroglu, Tetiana Grechuk","submitted_at":"2024-11-07T17:07:35Z","abstract_excerpt":"We introduce FrontierMath, a benchmark of hundreds of original, exceptionally challenging mathematics problems crafted and vetted by expert mathematicians. The questions cover most major branches of modern mathematics -- from computationally intensive problems in number theory and real analysis to abstract questions in algebraic geometry and category theory. Solving a typical problem requires multiple hours of effort from a researcher in the relevant branch of mathematics, and for the upper end questions, multiple days. FrontierMath uses new, unpublished problems and automated verification to "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Current state-of-the-art AI models solve under 2% of problems, revealing a vast gap between AI capabilities and the prowess of the mathematical community.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The problems are genuinely original and unpublished with no data contamination risk, and automated verification reliably measures true mathematical reasoning ability.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"FrontierMath is a new benchmark of hundreds of original hard math problems that current AI models solve less than 2% of.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"FrontierMath shows that current AI models solve under 2% of hundreds of original expert-level mathematics problems.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"d1533291bb17f4fcdd10470af2ef410e3853ceefb5e3d0db3efc952c906a1845"},"source":{"id":"2411.04872","kind":"arxiv","version":7},"verdict":{"id":"04ce5a15-2b12-4a1a-bc37-462ab30495b4","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-17T00:37:28.338908Z","strongest_claim":"Current state-of-the-art AI models solve under 2% of problems, revealing a vast gap between AI capabilities and the prowess of the mathematical community.","one_line_summary":"FrontierMath is a new benchmark of hundreds of original hard math problems that current AI models solve less than 2% of.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The problems are genuinely original and unpublished with no data contamination risk, and automated verification reliably measures true mathematical reasoning ability.","pith_extraction_headline":"FrontierMath shows that current AI models solve under 2% of hundreds of original expert-level mathematics problems."},"references":{"count":32,"sample":[{"doi":"","year":null,"title":"MSC2020 Mathematics Subject Classification System , author =","work_id":"5e2dc496-fa18-47d5-a4ac-19f19351c418","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"Training verifiers to solve math word problems, 2021 , author =","work_id":"877a76df-63bd-4ed4-af21-d24219973188","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Advances in neural information processing systems , volume=","work_id":"d428ff68-61a7-4024-83b4-d1b47a2ad468","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Measuring mathematical problem solving with the math dataset , author =","work_id":"743e28cf-6d1f-4301-a1ce-15d28821dd87","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Math Olympiad Hardness Scale (MOHS) , author =","work_id":"b6a79b50-e6c0-412d-b3e0-21e76a92e058","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":32,"snapshot_sha256":"e78a26e92f3862c6e21b005e21fbe7111a5555f6855b742f0ac72d69ddaaadc9","internal_anchors":2},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}