{
  "aliases": [],
  "authors": [],
  "details": {
    "dataset_profile": {
      "counts": [
        {
          "basis": "reported",
          "evidence_url": "https://raw.githubusercontent.com/neulab/multiui/HEAD/README.md",
          "population": "reported MultiUI training collection",
          "qualifier": "approximately",
          "unit": "samples",
          "value": 7300000
        }
      ],
      "evidence_urls": [
        "https://raw.githubusercontent.com/neulab/multiui/HEAD/README.md"
      ],
      "formats": [],
      "modalities": [
        "webpage UI content",
        "text",
        "accessibility-derived structure"
      ],
      "notes": "The README provides a rounded sample count. It does not establish a unique-screenshot inventory or a normalized split breakdown. Synthetic examples and publicly sourced web content have provenance and quality limits; data terms must be checked separately from repository code.",
      "splits": [
        "training collection"
      ],
      "supervision": [
        "UI understanding",
        "text-rich visual tasks"
      ]
    },
    "modalities": [
      "screenshot",
      "instruction"
    ],
    "reported_scale": {
      "comparison": "approximately",
      "unit": "samples",
      "value": 7300000
    },
    "sample_count": null,
    "sample_unit": "instructions"
  },
  "environments": [
    "web"
  ],
  "evidence": [
    {
      "method": "editorial_primary_review",
      "retrieved_at": "2026-10-16T06:36:01Z",
      "role": "primary",
      "sha256": "fe0c03fe73529a29788d559103c4aed48dca6ed9ba3bfd5f151f902fa8858b38",
      "supports": [
        "name",
        "kind",
        "summary",
        "environments",
        "topics",
        "scope",
        "relations",
        "details"
      ],
      "url": "https://raw.githubusercontent.com/neulab/multiui/HEAD/README.md"
    },
    {
      "method": "editorial_primary_review",
      "retrieved_at": "2026-10-16T06:36:01Z",
      "role": "primary",
      "sha256": "fe0c03fe73529a29788d559103c4aed48dca6ed9ba3bfd5f151f902fa8858b38",
      "supports": [
        "details.dataset_profile"
      ],
      "url": "https://raw.githubusercontent.com/neulab/multiui/HEAD/README.md"
    }
  ],
  "id": "github:neulab/multiui",
  "identifiers": {},
  "kind": "dataset",
  "last_verified_at": "2026-10-16T06:36:01Z",
  "licenses": {
    "code": null,
    "data": null,
    "document": null,
    "weights": null
  },
  "links": [
    {
      "rel": "primary",
      "url": "https://github.com/neulab/multiui"
    }
  ],
  "name": "MultiUI",
  "notes": "",
  "published_at": null,
  "relations": [],
  "review_status": "curated",
  "scope": "core",
  "summary": "7.3 million training samples built from webpage UIs and structured with enhanced accessibility trees, for text-rich visual understanding, released with UIX-Qwen2 models.",
  "topics": [
    "screen-understanding",
    "data-generation"
  ],
  "updated_at": null,
  "url": "https://github.com/neulab/multiui"
}