iiRecord
Agentic AI Atlas · HellaSwag validation
test-set:hellaswag-validationa5c.ai
II.
TestSet JSON

test-set:hellaswag-validation

Structured · live

HellaSwag validation json

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

File · benchmarks/test-sets/hellaswag-validation.yamlCluster · benchmarks
Record JSON
{
  "id": "test-set:hellaswag-validation",
  "_kind": "TestSet",
  "_file": "benchmarks/test-sets/hellaswag-validation.yaml",
  "_cluster": "benchmarks",
  "attributes": {
    "displayName": "HellaSwag validation",
    "benchmarkId": "benchmark:hellaswag",
    "caseCount": 10042,
    "releasedAt": "2019-05-19",
    "composition": "The validation split of HellaSwag (10,042 multiple-choice\nsentence-completion items adversarially filtered against an\nELMO/BERT discriminator). The test split labels are not public,\nso most evaluations report on validation.\n",
    "homepageUrl": "https://rowanzellers.com/hellaswag/",
    "description": "Standard HellaSwag validation split. Nearly all reported HellaSwag\nnumbers in modern model cards refer to this set.\n"
  },
  "outgoingEdges": [
    {
      "from": "test-set:hellaswag-validation",
      "to": "benchmark:hellaswag",
      "kind": "belongs_to_benchmark"
    }
  ],
  "incomingEdges": []
}