{
  "repository": "google-research/geospatial-reasoning-benchmarks",
  "revision": "c21a6d95c3fc446cba9205855785bc5ae31c0814",
  "retrieved_utc": "2026-11-06",
  "sources": [
    {
      "source_path": "geoktools/dev_queries.json",
      "url": "https://raw.githubusercontent.com/google-research/geospatial-reasoning-benchmarks/c21a6d95c3fc446cba9205855785bc5ae31c0814/geoktools/dev_queries.json",
      "git_blob_sha1": "b6866b1ad3a8c706eeced017bcf389c4b3d05130",
      "sha256": "8b90c0fe778e1196b8c10c0a58136b655e60ed1d988168506b466b7ca5af29bd",
      "bytes": 128517
    },
    {
      "source_path": "geoexpertqueries/dev_queries.json",
      "url": "https://raw.githubusercontent.com/google-research/geospatial-reasoning-benchmarks/c21a6d95c3fc446cba9205855785bc5ae31c0814/geoexpertqueries/dev_queries.json",
      "git_blob_sha1": "c3117277ba0d38af521a6a601baac7ea0a06e52e",
      "sha256": "1d3fb538594f49a55f3c26edd03072adbc10daa6a146bdcd60d8e8af639cb15f",
      "bytes": 499216
    },
    {
      "source_path": "geoktools/scoring/scoring_functions.py",
      "url": "https://raw.githubusercontent.com/google-research/geospatial-reasoning-benchmarks/c21a6d95c3fc446cba9205855785bc5ae31c0814/geoktools/scoring/scoring_functions.py",
      "git_blob_sha1": "f9a45a2ad8a6eed4e53f60e321d9f712c74fce0d",
      "sha256": "06f02b3af9220232137788184249426f6dca8db9d5e3fa65fc39139bd1fa1876",
      "bytes": 9769
    },
    {
      "source_path": "docs/evaluation_protocol.md",
      "url": "https://raw.githubusercontent.com/google-research/geospatial-reasoning-benchmarks/c21a6d95c3fc446cba9205855785bc5ae31c0814/docs/evaluation_protocol.md",
      "git_blob_sha1": "7a37f42f5943c42211093196e53ca3f912b5cde0",
      "sha256": "1d953586193aa6fadc848a7b4a38ddd9eee8d4e522c7cd6970738a17211d4d0c",
      "bytes": 3935
    },
    {
      "source_path": "docs/PROVENANCE.md",
      "url": "https://raw.githubusercontent.com/google-research/geospatial-reasoning-benchmarks/c21a6d95c3fc446cba9205855785bc5ae31c0814/docs/PROVENANCE.md",
      "git_blob_sha1": "896676e8a52aba09ca79c17f264ad2658df700de",
      "sha256": "3d6aeaabc501227c46bfa93d68396d6381055c1bdccadccd95f513d90a6527f5",
      "bytes": 5667
    },
    {
      "source_path": "geoexpertqueries/domain_index.md",
      "url": "https://raw.githubusercontent.com/google-research/geospatial-reasoning-benchmarks/c21a6d95c3fc446cba9205855785bc5ae31c0814/geoexpertqueries/domain_index.md",
      "git_blob_sha1": "9a5ca94eb654a6ead8f2afce73828df472428e10",
      "sha256": "e7d83df9c13796c1aba57e500d5ff3554bfcd6cae90b258ee5adf8b241b8a841",
      "bytes": 6316
    },
    {
      "source_path": "README.md",
      "url": "https://raw.githubusercontent.com/google-research/geospatial-reasoning-benchmarks/c21a6d95c3fc446cba9205855785bc5ae31c0814/README.md",
      "git_blob_sha1": "0c7f2d6b6963cd06242bf18015425f00f40bfbf0",
      "sha256": "019dbc5f794e8635ce1373e61ec837621ad1e79862052bfb0e4a2f8ce1cdd6ad",
      "bytes": 16210
    }
  ],
  "transformations": [
    "Full census of 85 GeoKTools and 25 GeoExpertQueries dev rows, not a sample.",
    "metadata.json strips all benchmark prose and all answer values. Numeric/zero/negative counts summarize only answer types.",
    "Held-out K/domain counts transcribed from official README/domain_index; held-out questions not inspected.",
    "One numeric scoring function extracted unchanged. All scoring probes use new hand-authored numeric examples, not agent predictions."
  ]
}