{
  "benchmark_version": "0.1.0",
  "released": "2026-07-11",
  "engine": {
    "name": "Tesseract",
    "version": "5.5.2",
    "language": "eng",
    "oem": 1,
    "page_segmentation_modes": [
      6,
      11
    ]
  },
  "interpretation_warning": "Engine confidence is not ground-truth accuracy.",
  "results": [
    {
      "case_id": "nist-sd19",
      "title": "NIST SD19 handwriting sample form",
      "psm": 6,
      "recognized_words": 67,
      "text_characters": 414,
      "nonempty_lines": 14,
      "mean_confidence": 45.81,
      "median_confidence": 41.21,
      "words_below_50": 37,
      "words_below_50_share": 0.5522,
      "source_sha256": "19077d062910638da0aa39e2a4344da68374ec745231074dbf9f46f07b72a1bc",
      "observation": "PSM 6 emphasizes the handwritten constitutional passage while PSM 11 surfaces more printed labels and isolated number rows; neither plain-text output preserves the form as fields."
    },
    {
      "case_id": "nist-sd19",
      "title": "NIST SD19 handwriting sample form",
      "psm": 11,
      "recognized_words": 218,
      "text_characters": 1418,
      "nonempty_lines": 72,
      "mean_confidence": 65.15,
      "median_confidence": 83.88,
      "words_below_50": 74,
      "words_below_50_share": 0.3394,
      "source_sha256": "19077d062910638da0aa39e2a4344da68374ec745231074dbf9f46f07b72a1bc",
      "observation": "PSM 6 emphasizes the handwritten constitutional passage while PSM 11 surfaces more printed labels and isolated number rows; neither plain-text output preserves the form as fields."
    },
    {
      "case_id": "nara-1930-census",
      "title": "1930 Census population schedule",
      "psm": 6,
      "recognized_words": 343,
      "text_characters": 1248,
      "nonempty_lines": 29,
      "mean_confidence": 24.65,
      "median_confidence": 23.36,
      "words_below_50": 317,
      "words_below_50_share": 0.9242,
      "source_sha256": "489eb0a3acbed2bfe7c9452d10e140cd8c090af2add702e1f8f8b9881aa1088c",
      "observation": "Both outputs contain fragments, but the plain text does not retain row identity or column attachment, which are necessary to interpret a census record."
    },
    {
      "case_id": "nara-1930-census",
      "title": "1930 Census population schedule",
      "psm": 11,
      "recognized_words": 531,
      "text_characters": 2394,
      "nonempty_lines": 355,
      "mean_confidence": 29.32,
      "median_confidence": 29.17,
      "words_below_50": 446,
      "words_below_50_share": 0.8399,
      "source_sha256": "489eb0a3acbed2bfe7c9452d10e140cd8c090af2add702e1f8f8b9881aa1088c",
      "observation": "Both outputs contain fragments, but the plain text does not retain row identity or column attachment, which are necessary to interpret a census record."
    },
    {
      "case_id": "loc-bell-notebook",
      "title": "Alexander Graham Bell notebook page",
      "psm": 6,
      "recognized_words": 151,
      "text_characters": 648,
      "nonempty_lines": 20,
      "mean_confidence": 25.84,
      "median_confidence": 22.61,
      "words_below_50": 132,
      "words_below_50_share": 0.8742,
      "source_sha256": "4d387dd0ba9d38af0ac7d2751a7db10c822f360d6d0fa2d2efbd7c632a6c48a5",
      "observation": "Both profiles produce many plausible-looking fragments, illustrating why fluent output still requires line-by-line comparison with the source."
    },
    {
      "case_id": "loc-bell-notebook",
      "title": "Alexander Graham Bell notebook page",
      "psm": 11,
      "recognized_words": 143,
      "text_characters": 577,
      "nonempty_lines": 47,
      "mean_confidence": 26.31,
      "median_confidence": 23.42,
      "words_below_50": 122,
      "words_below_50_share": 0.8531,
      "source_sha256": "4d387dd0ba9d38af0ac7d2751a7db10c822f360d6d0fa2d2efbd7c632a6c48a5",
      "observation": "Both profiles produce many plausible-looking fragments, illustrating why fluent output still requires line-by-line comparison with the source."
    },
    {
      "case_id": "irs-w4",
      "title": "IRS Form W-4",
      "psm": 6,
      "recognized_words": 784,
      "text_characters": 4458,
      "nonempty_lines": 55,
      "mean_confidence": 86.75,
      "median_confidence": 95.25,
      "words_below_50": 56,
      "words_below_50_share": 0.0714,
      "source_sha256": "904b53ad9c31444e1db36e6dcfc830155616a48e97a26ac46f52e2039c774192",
      "observation": "Large amounts of text are surfaced, but spelling substitutions appear and plain text does not preserve checkbox state or reliable label-to-field relationships."
    },
    {
      "case_id": "irs-w4",
      "title": "IRS Form W-4",
      "psm": 11,
      "recognized_words": 792,
      "text_characters": 4685,
      "nonempty_lines": 103,
      "mean_confidence": 90.04,
      "median_confidence": 95.7,
      "words_below_50": 27,
      "words_below_50_share": 0.0341,
      "source_sha256": "904b53ad9c31444e1db36e6dcfc830155616a48e97a26ac46f52e2039c774192",
      "observation": "Large amounts of text are surfaced, but spelling substitutions appear and plain text does not preserve checkbox state or reliable label-to-field relationships."
    }
  ]
}