{
  "aliases": [],
  "authors": [],
  "details": {
    "benchmark_protocol": {
      "actions": [
        "browser interaction"
      ],
      "counts": [
        {
          "basis": "reported",
          "evidence_url": "https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md",
          "population": "main V1/V2 suite (shared identities)",
          "qualifier": "exact",
          "unit": "task identities",
          "value": 609
        },
        {
          "basis": "reported",
          "evidence_url": "https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md",
          "population": "refusal suite",
          "qualifier": "exact",
          "unit": "tasks",
          "value": 111
        }
      ],
      "evaluation": "Task-specific precomputed rubrics; refusal tasks are a separate evaluation population.",
      "evidence_urls": [
        "https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md"
      ],
      "focus": [
        "live web task completion",
        "unsafe-request refusal"
      ],
      "notes": "V2 refreshes task strings and rubrics while preserving V1 task IDs. Counts are task definitions, not independently verified runnable website states.",
      "observations": [
        "web pages"
      ],
      "setting": "interactive"
    },
    "disabled": false,
    "download_url": "https://huggingface.co/datasets/microsoft/WebTailBench/tree/main",
    "gated": false,
    "hub_id": "microsoft/WebTailBench",
    "license_review": {
      "data": {
        "evidence_urls": [
          "https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md"
        ],
        "reason": "The benchmark's Hub dataset card declares MIT in its metadata header (license: mit)."
      }
    }
  },
  "environments": [
    "web"
  ],
  "evidence": [
    {
      "method": "editorial_primary_review",
      "retrieved_at": "2026-10-17T02:42:32Z",
      "role": "primary",
      "sha256": "c4cfc361d1d47d7855269af4110483363964e23f9c65244a1182830eccccbb7f",
      "supports": [
        "name",
        "kind",
        "summary",
        "environments",
        "topics",
        "scope",
        "relations",
        "details",
        "licenses.data"
      ],
      "url": "https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md"
    },
    {
      "method": "editorial_primary_review",
      "retrieved_at": "2026-10-17T02:42:32Z",
      "role": "primary",
      "sha256": "93469a20d333219dde89a225891a54ee716120d559eb8ac9f69a1a7ee84f57a3",
      "supports": [
        "identity",
        "release_status",
        "details",
        "licenses.data"
      ],
      "url": "https://huggingface.co/api/datasets/microsoft/WebTailBench"
    }
  ],
  "id": "url:3b6dfd52f98bfe76bc7f",
  "identifiers": {},
  "kind": "benchmark",
  "last_verified_at": "2026-10-17T02:42:32Z",
  "licenses": {
    "code": null,
    "data": {
      "basis": "publisher model/dataset card metadata; third-party assets retain their own terms",
      "declared": "mit",
      "retrieved_at": "2026-10-17T02:42:32Z",
      "url": "https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md"
    },
    "document": null,
    "weights": null
  },
  "links": [
    {
      "rel": "primary",
      "url": "https://huggingface.co/datasets/microsoft/WebTailBench"
    }
  ],
  "name": "WebTailBench",
  "notes": "Public TSV files contain V1, V2 and refusal task definitions. The card declares MIT for the dataset. V1 and V2 share the same 609 task identities and must not be summed as distinct tasks. The May 2026 refresh supersedes calendar-bound task wording; live-site availability is not established by this metadata review.",
  "published_at": null,
  "relations": [
    {
      "evidence_url": "https://huggingface.co/datasets/microsoft/WebTailBench/raw/main/README.md",
      "target_id": "arxiv:2511.19663",
      "type": "documented_by"
    }
  ],
  "review_status": "curated",
  "scope": "core",
  "summary": "Live-web task definitions and rubrics for evaluating computer-use agents, with preserved original tasks, a refreshed V2 task suite and a separate refusal set.",
  "topics": [
    "evaluation",
    "web-navigation",
    "safety-security"
  ],
  "updated_at": null,
  "url": "https://huggingface.co/datasets/microsoft/WebTailBench"
}