{
  "id": "test-set:re-bench-ai-rd-tasks",
  "_kind": "TestSet",
  "_file": "benchmarks/test-sets/source-backed-benchmark-test-sets.yaml",
  "_cluster": "benchmarks",
  "attributes": {
    "displayName": "RE-Bench AI R&D task suite",
    "benchmarkId": "benchmark:re-bench",
    "caseCount": 7,
    "composition": "RE-Bench's staged task suite measures autonomous AI research and\ndevelopment work against human-expert baselines using realistic\nmachine-learning research-engineering tasks.\n",
    "homepageUrl": "https://github.com/METR/RE-Bench",
    "description": "Canonical RE-Bench artifact for frontier AI R&D agent evaluation.\n"
  },
  "outgoingEdges": [
    {
      "from": "test-set:re-bench-ai-rd-tasks",
      "to": "benchmark:re-bench",
      "kind": "belongs_to_benchmark"
    }
  ],
  "incomingEdges": []
}