{
  "project": "KAMIWAZA",
  "kind": "retained-llm-verdict-recount",
  "evaluation_date_reported_by_source": "2026-07-06",
  "recounted_on": "2026-09-09",
  "case_id_rule": "K01 corresponds to source index 0; K50 to source index 49.",
  "source_article": "https://zenn.dev/hocky3/articles/2b3b638a8d6885",
  "scope": "The same 50 documents evaluated in two recorded passes. Recount of stored LLM verdicts; no image regrading or extraction rerun.",
  "limitations": [
    "Original evaluation images are unavailable; the same-set evaluation cannot be rerun.",
    "Grader model identifier was not recorded; agreement with human graders was not measured.",
    "Success is the stored rubric verdict and allows minor auxiliary omissions; it does not mean every value is correct.",
    "Value denominators count evaluated nonempty extracted values; they do not measure all information omitted from an image.",
    "One v1 document fell back to demo and was counted as an automatic failure, with zero evaluated values.",
    "v1 and v2 used the same documents; 2,225 is the sum of two passes, not independent documents or unique values.",
    "The historical evaluation used the direct model-provider path; its results do not establish current routed-path performance.",
    "Multiple implementation and prompt changes occurred between versions; these records do not isolate the effect of the DSL change."
  ],
  "summary": {
    "v1": {
      "documents": 50,
      "verdict_counts": {
        "success": 38,
        "partial": 11,
        "fail": 1
      },
      "values_total": 707,
      "values_correct": 704,
      "classified_fabrications": 0
    },
    "v2": {
      "documents": 50,
      "verdict_counts": {
        "success": 48,
        "partial": 2,
        "fail": 0
      },
      "values_total": 1518,
      "values_correct": 1516,
      "classified_fabrications": 0,
      "lines_expected": 196,
      "lines_extracted": 196
    },
    "transitions": [
      {
        "from": "fail",
        "to": "success",
        "documents": 1
      },
      {
        "from": "partial",
        "to": "partial",
        "documents": 1
      },
      {
        "from": "partial",
        "to": "success",
        "documents": 10
      },
      {
        "from": "success",
        "to": "partial",
        "documents": 1
      },
      {
        "from": "success",
        "to": "success",
        "documents": 37
      }
    ]
  },
  "cases": [
    {
      "case_id": "K01",
      "v1_verdict": "partial",
      "v1_values_correct": 17,
      "v1_values_total": 17,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 36,
      "v2_values_total": 36,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 2,
      "v2_lines_extracted": 2
    },
    {
      "case_id": "K02",
      "v1_verdict": "success",
      "v1_values_correct": 9,
      "v1_values_total": 9,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 33,
      "v2_values_total": 33,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 5,
      "v2_lines_extracted": 5
    },
    {
      "case_id": "K03",
      "v1_verdict": "partial",
      "v1_values_correct": 11,
      "v1_values_total": 11,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 48,
      "v2_values_total": 48,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 10,
      "v2_lines_extracted": 10
    },
    {
      "case_id": "K04",
      "v1_verdict": "success",
      "v1_values_correct": 10,
      "v1_values_total": 10,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 13,
      "v2_values_total": 13,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K05",
      "v1_verdict": "partial",
      "v1_values_correct": 9,
      "v1_values_total": 9,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 20,
      "v2_values_total": 20,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 4,
      "v2_lines_extracted": 4
    },
    {
      "case_id": "K06",
      "v1_verdict": "partial",
      "v1_values_correct": 14,
      "v1_values_total": 14,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 35,
      "v2_values_total": 35,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 9,
      "v2_lines_extracted": 9
    },
    {
      "case_id": "K07",
      "v1_verdict": "success",
      "v1_values_correct": 13,
      "v1_values_total": 13,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 15,
      "v2_values_total": 15,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 0,
      "v2_lines_extracted": 0
    },
    {
      "case_id": "K08",
      "v1_verdict": "partial",
      "v1_values_correct": 19,
      "v1_values_total": 19,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 112,
      "v2_values_total": 112,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 16,
      "v2_lines_extracted": 16
    },
    {
      "case_id": "K09",
      "v1_verdict": "success",
      "v1_values_correct": 4,
      "v1_values_total": 4,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 4,
      "v2_values_total": 4,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 0,
      "v2_lines_extracted": 0
    },
    {
      "case_id": "K10",
      "v1_verdict": "partial",
      "v1_values_correct": 16,
      "v1_values_total": 16,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 124,
      "v2_values_total": 124,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 18,
      "v2_lines_extracted": 18
    },
    {
      "case_id": "K11",
      "v1_verdict": "success",
      "v1_values_correct": 24,
      "v1_values_total": 24,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 51,
      "v2_values_total": 51,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 4,
      "v2_lines_extracted": 4
    },
    {
      "case_id": "K12",
      "v1_verdict": "success",
      "v1_values_correct": 26,
      "v1_values_total": 26,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 68,
      "v2_values_total": 69,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 6,
      "v2_lines_extracted": 6
    },
    {
      "case_id": "K13",
      "v1_verdict": "success",
      "v1_values_correct": 21,
      "v1_values_total": 21,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 59,
      "v2_values_total": 59,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 10,
      "v2_lines_extracted": 10
    },
    {
      "case_id": "K14",
      "v1_verdict": "success",
      "v1_values_correct": 23,
      "v1_values_total": 23,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 35,
      "v2_values_total": 35,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 3,
      "v2_lines_extracted": 3
    },
    {
      "case_id": "K15",
      "v1_verdict": "success",
      "v1_values_correct": 22,
      "v1_values_total": 22,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 41,
      "v2_values_total": 41,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 5,
      "v2_lines_extracted": 5
    },
    {
      "case_id": "K16",
      "v1_verdict": "partial",
      "v1_values_correct": 23,
      "v1_values_total": 24,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 65,
      "v2_values_total": 65,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 10,
      "v2_lines_extracted": 10
    },
    {
      "case_id": "K17",
      "v1_verdict": "success",
      "v1_values_correct": 21,
      "v1_values_total": 21,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 56,
      "v2_values_total": 56,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 7,
      "v2_lines_extracted": 7
    },
    {
      "case_id": "K18",
      "v1_verdict": "success",
      "v1_values_correct": 26,
      "v1_values_total": 26,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 35,
      "v2_values_total": 35,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 8,
      "v2_lines_extracted": 8
    },
    {
      "case_id": "K19",
      "v1_verdict": "success",
      "v1_values_correct": 23,
      "v1_values_total": 23,
      "v1_classified_fabrications": 0,
      "v2_verdict": "partial",
      "v2_values_correct": 41,
      "v2_values_total": 41,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 5,
      "v2_lines_extracted": 5
    },
    {
      "case_id": "K20",
      "v1_verdict": "success",
      "v1_values_correct": 22,
      "v1_values_total": 22,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 26,
      "v2_values_total": 26,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 2,
      "v2_lines_extracted": 2
    },
    {
      "case_id": "K21",
      "v1_verdict": "success",
      "v1_values_correct": 14,
      "v1_values_total": 14,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 16,
      "v2_values_total": 16,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K22",
      "v1_verdict": "success",
      "v1_values_correct": 13,
      "v1_values_total": 13,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 15,
      "v2_values_total": 15,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K23",
      "v1_verdict": "success",
      "v1_values_correct": 14,
      "v1_values_total": 14,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 18,
      "v2_values_total": 18,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K24",
      "v1_verdict": "success",
      "v1_values_correct": 14,
      "v1_values_total": 14,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 16,
      "v2_values_total": 16,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K25",
      "v1_verdict": "success",
      "v1_values_correct": 14,
      "v1_values_total": 14,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 16,
      "v2_values_total": 16,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K26",
      "v1_verdict": "success",
      "v1_values_correct": 10,
      "v1_values_total": 10,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 33,
      "v2_values_total": 33,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 8,
      "v2_lines_extracted": 8
    },
    {
      "case_id": "K27",
      "v1_verdict": "success",
      "v1_values_correct": 14,
      "v1_values_total": 14,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 20,
      "v2_values_total": 20,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K28",
      "v1_verdict": "success",
      "v1_values_correct": 13,
      "v1_values_total": 13,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 15,
      "v2_values_total": 15,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K29",
      "v1_verdict": "success",
      "v1_values_correct": 15,
      "v1_values_total": 15,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 20,
      "v2_values_total": 20,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K30",
      "v1_verdict": "success",
      "v1_values_correct": 15,
      "v1_values_total": 15,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 20,
      "v2_values_total": 20,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K31",
      "v1_verdict": "success",
      "v1_values_correct": 9,
      "v1_values_total": 9,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 13,
      "v2_values_total": 13,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 2,
      "v2_lines_extracted": 2
    },
    {
      "case_id": "K32",
      "v1_verdict": "success",
      "v1_values_correct": 9,
      "v1_values_total": 9,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 10,
      "v2_values_total": 10,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K33",
      "v1_verdict": "success",
      "v1_values_correct": 9,
      "v1_values_total": 9,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 13,
      "v2_values_total": 13,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 2,
      "v2_lines_extracted": 2
    },
    {
      "case_id": "K34",
      "v1_verdict": "success",
      "v1_values_correct": 9,
      "v1_values_total": 9,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 13,
      "v2_values_total": 13,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 2,
      "v2_lines_extracted": 2
    },
    {
      "case_id": "K35",
      "v1_verdict": "success",
      "v1_values_correct": 9,
      "v1_values_total": 9,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 14,
      "v2_values_total": 14,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 2,
      "v2_lines_extracted": 2
    },
    {
      "case_id": "K36",
      "v1_verdict": "success",
      "v1_values_correct": 9,
      "v1_values_total": 9,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 11,
      "v2_values_total": 11,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 2,
      "v2_lines_extracted": 2
    },
    {
      "case_id": "K37",
      "v1_verdict": "partial",
      "v1_values_correct": 8,
      "v1_values_total": 9,
      "v1_classified_fabrications": 0,
      "v2_verdict": "partial",
      "v2_values_correct": 10,
      "v2_values_total": 11,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 2,
      "v2_lines_extracted": 2
    },
    {
      "case_id": "K38",
      "v1_verdict": "success",
      "v1_values_correct": 9,
      "v1_values_total": 9,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 12,
      "v2_values_total": 12,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 2,
      "v2_lines_extracted": 2
    },
    {
      "case_id": "K39",
      "v1_verdict": "success",
      "v1_values_correct": 9,
      "v1_values_total": 9,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 10,
      "v2_values_total": 10,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K40",
      "v1_verdict": "partial",
      "v1_values_correct": 8,
      "v1_values_total": 9,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 12,
      "v2_values_total": 12,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 2,
      "v2_lines_extracted": 2
    },
    {
      "case_id": "K41",
      "v1_verdict": "partial",
      "v1_values_correct": 5,
      "v1_values_total": 5,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 60,
      "v2_values_total": 60,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 18,
      "v2_lines_extracted": 18
    },
    {
      "case_id": "K42",
      "v1_verdict": "success",
      "v1_values_correct": 14,
      "v1_values_total": 14,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 44,
      "v2_values_total": 44,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 4,
      "v2_lines_extracted": 4
    },
    {
      "case_id": "K43",
      "v1_verdict": "fail",
      "v1_values_correct": 0,
      "v1_values_total": 0,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 17,
      "v2_values_total": 17,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 0,
      "v2_lines_extracted": 0
    },
    {
      "case_id": "K44",
      "v1_verdict": "success",
      "v1_values_correct": 15,
      "v1_values_total": 15,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 22,
      "v2_values_total": 22,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K45",
      "v1_verdict": "success",
      "v1_values_correct": 18,
      "v1_values_total": 18,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 35,
      "v2_values_total": 35,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 4,
      "v2_lines_extracted": 4
    },
    {
      "case_id": "K46",
      "v1_verdict": "success",
      "v1_values_correct": 0,
      "v1_values_total": 0,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 0,
      "v2_values_total": 0,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 0,
      "v2_lines_extracted": 0
    },
    {
      "case_id": "K47",
      "v1_verdict": "success",
      "v1_values_correct": 20,
      "v1_values_total": 20,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 22,
      "v2_values_total": 22,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K48",
      "v1_verdict": "partial",
      "v1_values_correct": 16,
      "v1_values_total": 16,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 36,
      "v2_values_total": 36,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 4,
      "v2_lines_extracted": 4
    },
    {
      "case_id": "K49",
      "v1_verdict": "success",
      "v1_values_correct": 21,
      "v1_values_total": 21,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 22,
      "v2_values_total": 22,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 1,
      "v2_lines_extracted": 1
    },
    {
      "case_id": "K50",
      "v1_verdict": "success",
      "v1_values_correct": 18,
      "v1_values_total": 18,
      "v1_classified_fabrications": 0,
      "v2_verdict": "success",
      "v2_values_correct": 34,
      "v2_values_total": 34,
      "v2_classified_fabrications": 0,
      "v2_lines_expected": 3,
      "v2_lines_extracted": 3
    }
  ]
}
