benchmark:frontier-math

{
  "id": "benchmark:frontier-math",
  "_kind": "Benchmark",
  "_file": "benchmarks/benchmarks/benchmarks-reasoning.yaml",
  "_cluster": "benchmarks",
  "attributes": {
    "displayName": "FrontierMath",
    "homepageUrl": "https://epoch.ai/frontiermath",
    "kind": "math",
    "targetsKind": "ModelVersion",
    "description": "Epoch AI benchmark of original, expert-vetted research-level math\nproblems requiring deep mathematical reasoning, designed to be\nhard for frontier models even with tool use.\n"
  },
  "outgoingEdges": [
    {
      "from": "benchmark:frontier-math",
      "to": "skill-area:mathematical-reasoning",
      "kind": "covers",
      "attributes": {
        "attributes": {
          "coverage": "full",
          "weight": 0.6
        }
      }
    },
    {
      "from": "benchmark:frontier-math",
      "to": "skill-area:closed-book-frontier-reasoning",
      "kind": "covers",
      "attributes": {
        "attributes": {
          "coverage": "partial",
          "weight": 0.3
        }
      }
    }
  ],
  "incomingEdges": []
}