{
  "id": "test-set:truthful-qa-mc",
  "_kind": "TestSet",
  "_file": "benchmarks/test-sets/test-sets.yaml",
  "_cluster": "benchmarks",
  "attributes": {
    "displayName": "TruthfulQA — multiple-choice",
    "benchmarkId": "benchmark:truthful-qa",
    "caseCount": 817,
    "releasedAt": "2021-09-08",
    "composition": "The 817-question multiple-choice split of TruthfulQA, covering\n38 categories of common human misconceptions. Used in MC1 (single\ncorrect answer) and MC2 (normalized probability mass over all\ncorrect answers) scoring modes.\n",
    "homepageUrl": "https://github.com/sylinrl/TruthfulQA",
    "description": "The MC split is the standard reported number for vendor\nleaderboards on TruthfulQA.\n"
  },
  "outgoingEdges": [],
  "incomingEdges": [
    {
      "from": "eval-run:truthful-qa.claude-opus-4-5.2025-09",
      "to": "test-set:truthful-qa-mc",
      "kind": "uses_test_set"
    }
  ]
}