{
  "aliases": [],
  "authors": [],
  "details": {
    "benchmark_protocol": {
      "actions": [
        "device actions through AgentEnv"
      ],
      "counts": [
        {
          "basis": "reported",
          "evidence_url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md",
          "population": "complete task collection",
          "qualifier": "exact",
          "unit": "tasks",
          "value": 495
        },
        {
          "basis": "reported",
          "evidence_url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md",
          "population": "AITW-derived subset",
          "qualifier": "exact",
          "unit": "tasks",
          "value": 102
        },
        {
          "basis": "reported",
          "evidence_url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md",
          "population": "self-constructed subset",
          "qualifier": "exact",
          "unit": "tasks",
          "value": 393
        },
        {
          "basis": "reported",
          "evidence_url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md",
          "population": "self-constructed subset",
          "qualifier": "exact",
          "unit": "Android applications",
          "value": 46
        }
      ],
      "evaluation": "Checks whether the recorded execution visits all annotated essential states rather than requiring an identical action sequence.",
      "evidence_urls": [
        "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md"
      ],
      "focus": [
        "mobile task completion",
        "essential-state evaluation"
      ],
      "notes": "Application coverage refers to the self-constructed subset. Task execution and essential-state annotation are separate components; counts are publisher-reported, not a fresh task-file recount.",
      "observations": [
        "screenshots",
        "view hierarchies",
        "Android activity names"
      ],
      "setting": "interactive"
    }
  },
  "environments": [
    "android"
  ],
  "evidence": [
    {
      "method": "editorial_primary_review",
      "retrieved_at": "2026-10-16T06:35:58Z",
      "role": "primary",
      "sha256": "4c4fe0fa92961f7286b6553b13117b2942e8482a48b7120a79e57dce2f8d42b9",
      "supports": [
        "name",
        "kind",
        "summary",
        "environments",
        "topics",
        "scope",
        "relations",
        "details"
      ],
      "url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md"
    },
    {
      "method": "editorial_primary_review",
      "retrieved_at": "2026-10-16T06:35:58Z",
      "role": "primary",
      "sha256": "4c4fe0fa92961f7286b6553b13117b2942e8482a48b7120a79e57dce2f8d42b9",
      "supports": [
        "details.benchmark_protocol"
      ],
      "url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md"
    }
  ],
  "id": "github:llamatouch/llamatouch",
  "identifiers": {},
  "kind": "benchmark",
  "last_verified_at": "2026-10-16T06:35:58Z",
  "licenses": {
    "code": null,
    "data": null,
    "document": null,
    "weights": null
  },
  "links": [
    {
      "rel": "primary",
      "url": "https://github.com/LlamaTouch/LlamaTouch"
    }
  ],
  "name": "LlamaTouch",
  "notes": "",
  "published_at": null,
  "relations": [
    {
      "evidence_url": "https://raw.githubusercontent.com/LlamaTouch/LlamaTouch/HEAD/README.md",
      "target_id": "arxiv:2404.16054",
      "type": "documented_by"
    }
  ],
  "review_status": "curated",
  "scope": "core",
  "summary": "Mobile UI task evaluation using application state information.",
  "topics": [
    "evaluation"
  ],
  "updated_at": null,
  "url": "https://github.com/LlamaTouch/LlamaTouch"
}