{
  "aliases": [],
  "authors": [],
  "details": {
    "benchmark_protocol": {
      "actions": [
        "click point or box center",
        "refusal represented by negative coordinates in the evaluator"
      ],
      "counts": [],
      "evaluation": "The inspected Jedi runner maps predictions from resized-image coordinates back to the original image, then checks the predicted center against rectangle/polygon targets or the refusal rule. It reports overall and classification-group accuracy. This is offline grounding, not OSWorld task-execution success.",
      "evidence_urls": [
        "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/README.md",
        "https://api.github.com/repos/xlang-ai/OSWorld-G/git/trees/main?recursive=1",
        "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/evaluation/qwen25_vllm_osworld_g_jedi.py",
        "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/evaluation/eval.py"
      ],
      "focus": [
        "desktop GUI grounding",
        "infeasible-target refusal"
      ],
      "notes": "The repository retains original OSWorld-G.json and refined OSWorld-G_refined.json instructions; treat them as alternative benchmark versions, not additional disjoint tasks. Refined instructions reduce requirements beyond grounding. The same repository also contains Jedi training/rendering material, which is excluded from benchmark counts. The standalone generic eval.py entry point is a stub; the named model runner supplies image_size to its geometry helper. No task count or successful execution is inferred from repository file totals. Apache-2.0 is recorded for code only.",
      "observations": [
        "screenshot",
        "instruction"
      ],
      "setting": "offline"
    }
  },
  "environments": [
    "desktop"
  ],
  "evidence": [
    {
      "method": "editorial_primary_review",
      "retrieved_at": "2026-10-16T06:35:57Z",
      "role": "primary",
      "sha256": "e814851bc3016a2abf59358f99b80afd9c2bd5894fc34baddfb7e4d0aa03bd45",
      "supports": [
        "name",
        "kind",
        "summary",
        "environments",
        "topics",
        "scope",
        "relations",
        "details.benchmark_protocol"
      ],
      "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/README.md"
    },
    {
      "method": "editorial_primary_review",
      "retrieved_at": "2026-10-17T14:14:05Z",
      "role": "primary",
      "sha256": "48308bba4f26b02cb7b8fbbc433bb31ec81c494dabd3af437ddbb10b8d9ce667",
      "supports": [
        "details.benchmark_protocol"
      ],
      "url": "https://api.github.com/repos/xlang-ai/OSWorld-G/git/trees/main?recursive=1"
    },
    {
      "method": "editorial_primary_review",
      "retrieved_at": "2026-10-17T14:22:04Z",
      "role": "primary",
      "sha256": "64f64e0b847e0fa200aa6051e20b8035a436df00e2ef0d769688ab13fa4d3046",
      "supports": [
        "details.benchmark_protocol"
      ],
      "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/evaluation/qwen25_vllm_osworld_g_jedi.py"
    },
    {
      "method": "editorial_primary_review",
      "retrieved_at": "2026-10-17T14:21:09Z",
      "role": "primary",
      "sha256": "b85f5685ce6c200ea648471666177859693d7f40cf50c12d08765ae68580de4a",
      "supports": [
        "details.benchmark_protocol"
      ],
      "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/evaluation/eval.py"
    },
    {
      "method": "license_header",
      "retrieved_at": "2026-10-16T06:41:57Z",
      "role": "primary",
      "sha256": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
      "supports": [
        "licenses.code"
      ],
      "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/LICENSE"
    }
  ],
  "id": "github:xlang-ai/osworld-g",
  "identifiers": {},
  "kind": "benchmark",
  "last_verified_at": "2026-10-16T06:35:57Z",
  "licenses": {
    "code": {
      "basis": "repository LICENSE header; not a weights/data license",
      "declared": "Apache-2.0",
      "retrieved_at": "2026-10-16T06:41:57Z",
      "url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/LICENSE"
    },
    "data": null,
    "document": null,
    "weights": null
  },
  "links": [
    {
      "rel": "primary",
      "url": "https://github.com/xlang-ai/OSWorld-G"
    }
  ],
  "name": "OSWorld-G",
  "notes": "",
  "published_at": null,
  "relations": [
    {
      "evidence_url": "https://raw.githubusercontent.com/xlang-ai/OSWorld-G/HEAD/README.md",
      "target_id": "arxiv:2505.13227",
      "type": "documented_by"
    }
  ],
  "review_status": "curated",
  "scope": "core",
  "summary": "Computer-use grounding benchmark with original and refined instructions, released with the Jedi synthetic data pipeline, a 4-million-example Jedi dataset and Jedi-3B and Jedi-7B models.",
  "topics": [
    "evaluation",
    "grounding"
  ],
  "updated_at": null,
  "url": "https://github.com/xlang-ai/OSWorld-G"
}