iiRecord
Agentic AI Atlas · Humanity's Last Exam (HLE)
benchmark:hlea5c.ai
II.
Benchmark JSON

benchmark:hle

Structured · live

Humanity's Last Exam (HLE) json

Inspect the normalized record payload exactly as the atlas UI reads it.

File · benchmarks/benchmarks/benchmarks-reasoning.yamlCluster · benchmarks
Record JSON
{
  "id": "benchmark:hle",
  "_kind": "Benchmark",
  "_file": "benchmarks/benchmarks/benchmarks-reasoning.yaml",
  "_cluster": "benchmarks",
  "attributes": {
    "displayName": "Humanity's Last Exam (HLE)",
    "homepageUrl": "https://lastexam.ai/",
    "kind": "knowledge",
    "targetsKind": "ModelVersion",
    "description": "Closed-book frontier-knowledge benchmark of expert-authored\nquestions across math, sciences, and humanities, designed to\nremain unsaturated by current frontier models.\n"
  },
  "outgoingEdges": [
    {
      "from": "benchmark:hle",
      "to": "skill-area:closed-book-frontier-reasoning",
      "kind": "covers",
      "attributes": {
        "attributes": {
          "coverage": "full",
          "weight": 0.7
        }
      }
    },
    {
      "from": "benchmark:hle",
      "to": "skill-area:general-knowledge-reasoning",
      "kind": "covers",
      "attributes": {
        "attributes": {
          "coverage": "partial",
          "weight": 0.2
        }
      }
    },
    {
      "from": "benchmark:hle",
      "to": "domain:computer-science",
      "kind": "applies_to",
      "attributes": {}
    },
    {
      "from": "benchmark:hle",
      "to": "domain:mathematics",
      "kind": "applies_to",
      "attributes": {}
    }
  ],
  "incomingEdges": []
}