{
  "corpus": {
    "file": "test/fixtures/verify/validation-set-v5.json",
    "name": "validation-set-v5",
    "seed": 20260801,
    "entries": 1402,
    "sealed": true
  },
  "spot_check": {
    "reviewer": "Mark Lavercombe",
    "date": "2026-08-02",
    "rows": 100,
    "method": "TITLE-ONLY, verifier-blind: reviewer opened each identifier and compared only the claimed title against the resolved record. The verifier's own output was never shown. Per-shape expectations (docs/v5-spotcheck.md) — clean rows and the author/year-swap fabrication shapes expect same-work; fake-title, wrong-doi, pmid-doi-split and near-miss-title expect different-work.",
    "result": "PASS — 100/100 rows agree with their per-shape expectation; zero construction defects",
    "counts": {
      "same-work": 89,
      "different-work": 11
    },
    "contradictions": 0,
    "excluded_ids": [],
    "unsure": 0,
    "not_found": 0,
    "verdicts": [
      {
        "id": "clean-doi-harvest-10-1109-tits-2024-3422039",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-3390-challe6010117",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-nearmiss-10-1016-j-cmpb-2021-106300",
        "human_verdict": "different-work",
        "expected": "different-work"
      },
      {
        "id": "clean-doi-harvest-10-1177-1073191121998760",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-3386-w31122",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-pmid-blind-34931032",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-pmid-blind-35047610",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1017-s0013091500008786",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-whoiris-blind-10665-330392",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1016-0001-8708-89-90004-2",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1016-j-labeco-2014-06-005",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1177-0956797610377342",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-nearmiss-one-letter-long-07",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1006-redy-1998-0056",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-2307-2392383",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-wrongdoi-10-3389-fpsyg-2018-01713",
        "human_verdict": "different-work",
        "expected": "different-work"
      },
      {
        "id": "clean-doi-harvest-10-1191-1478088706qp063oa",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1109-iccv-2017-74",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-arxiv-harvest-2212-06357",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1145-1054972-1055012",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1109-ms-2012-167",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-wrongauthor-10-2988-0006-324x-129-q2-76",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-isbn-blind-9781850070917",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1103-physrevlett-78-390",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1016-j-isci-2023-107074",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-nearmiss-one-letter-short-03",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-3102-10769986027003223",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-37236-12906",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-5539-ibr-v5n10p19",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1112-s0024611597000282",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1007-bf00985762",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1145-1095034-1095063",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1016-0021-8693-88-90177-9",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-pmcid-blind-PMC9381412",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-wrongauthor-10-1016-j-peptides-2018-04-017",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-wrongauthor-10-1007-s40031-020-00480-7",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1007-bf00383950",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-nearmiss-one-letter-short-10",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1109-tcc-2023-3264242",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-ads-blind-2021JGRE..12606711L",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-ads-blind-2021Icar..36914650L",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-nearmiss-10-1002-cre2-3",
        "human_verdict": "different-work",
        "expected": "different-work"
      },
      {
        "id": "clean-doi-harvest-10-1007-s13748-012-0030-x",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-pmiddoisplit-42534700",
        "human_verdict": "different-work",
        "expected": "different-work"
      },
      {
        "id": "clean-doi-harvest-10-1016-j-adapen-2024-100174",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1121-1-1506692",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1016-j-ijhydene-2024-05-107",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-pmcid-blind-PMC12635230",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1016-j-ejc-2011-08-005",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1364-josaa-7-000923",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1257-pandp-20201036",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1145-3629527-3651419",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-faketitle-10-1016-j-envres-2018-10-025",
        "human_verdict": "different-work",
        "expected": "different-work"
      },
      {
        "id": "clean-doi-harvest-10-1111-j-1467-7660-2007-00404-x",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-faketitle-10-3934-math-2021412",
        "human_verdict": "different-work",
        "expected": "different-work"
      },
      {
        "id": "clean-doi-harvest-10-1109-massp-1984-1162229",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1093-bioinformatics-btx075",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1007-3-540-30731-1",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-nearmiss-shared-particle-06",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1017-9781316480748",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1016-j-engappai-2023-106476",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-wrongdoi-10-1055-a-1521-5124",
        "human_verdict": "different-work",
        "expected": "different-work"
      },
      {
        "id": "clean-doi-harvest-10-1177-000312240507000408",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1177-0001839217694358",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-2307-2983068",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1080-07421222-1996-11518099",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-pmiddoisplit-42534728",
        "human_verdict": "different-work",
        "expected": "different-work"
      },
      {
        "id": "clean-arxiv-harvest-2409-04231",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-nearmiss-one-letter-short-09",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-2139-ssrn-4922031",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-17016-2380-7172-0056",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-nearmiss-10-2147-rmi-s356443",
        "human_verdict": "different-work",
        "expected": "different-work"
      },
      {
        "id": "fab-wrongyear-10-1049-mia2-12308",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1016-j-microc-2024-111307",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1371-journal-pone-0103561",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1287-mksc-2023-0306",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-pmid-blind-35254061",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1109-tpds-2013-257",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-nearmiss-relocated-particle-08",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1056-nejmoa1414123",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-pmcid-blind-PMC11042076",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1016-j-patrec-2005-02-003",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-5220-0010710300003058",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1109-wacv-2018-00168",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1609-aies-v7i1-31745",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1006-juec-2000-2171",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1111-j-1468-2354-2009-00566-x",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1145-3178876-3186044",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1080-10496491-2025-2466591",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-wrongauthor-10-47363-jcrr-2021-3-151",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-arxiv-harvest-2106-12741",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-nearmiss-10-1017-s104795111700244x",
        "human_verdict": "different-work",
        "expected": "different-work"
      },
      {
        "id": "clean-doi-harvest-10-1145-3408877-3432539",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1016-j-jmva-2005-08-003",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1038-s41598-019-55247-y",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "fab-nearmiss-10-1158-1078-0432-ccr-18-2509",
        "human_verdict": "different-work",
        "expected": "different-work"
      },
      {
        "id": "clean-doi-harvest-10-18653-v1-2020-findings-emnlp-428",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1016-j-jpaa-2018-09-014",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1257-aer-20221180",
        "human_verdict": "same-work",
        "expected": "same-work"
      },
      {
        "id": "clean-doi-harvest-10-1037-met0000167",
        "human_verdict": "same-work",
        "expected": "same-work"
      }
    ],
    "note": "Two construction defects were found BY THIS CHECK during review and repaired before completion (near-miss-title with/without landing on connectives, e5d3603d; wrong-authors §5 collision where the substitute was author 7 on the same paper, 9e903a98). Both were then re-verified across their WHOLE stratum rather than by sampling — see scripts/blind-corpus/audit-fabrication-strata.mjs. The 100 rows above were all ticked against the repaired corpus.",
    "record": "Mark Lavercombe, 2026-08-02: 100/100 rows confirmed fair against per-shape expectations; 0 contradictions, 0 unsure, 0 not-found. Two construction defects found by this check during review and repaired before completion (e5d3603d near-miss-title connective flip; 9e903a98 wrong-authors §5 author-7 collision), each then re-verified across its whole stratum."
  },
  "measured_at": "2026-08-02T02:36:01.076Z",
  "endpoint": "https://scholar-sidekick.com",
  "with_llm": false,
  "verify_version": "2026-07-31",
  "transform_version": "2026-06-16",
  "flags": {
    "ENABLE_COAUTHOR_CHECK": "unset in Vercel → defaults to 0 (OFF). This is the intended state: the flag is gated on exactly this FAR measurement, so it must not be on while measuring."
  },
  "run_health": {
    "entries": 1402,
    "http_errors": 0,
    "note": "Zero non-200 responses across the full run — every failure is a genuine verdict disagreement, not transport noise."
  },
  "false_accusation_rate": {
    "v5": {
      "flagged": 14,
      "n": 1152,
      "ci95": "1.22% [0.58–1.85]"
    },
    "v4": {
      "flagged": 32,
      "n": 1185,
      "ci95": "2.70% [1.78–3.62]"
    },
    "high_confidence_v5": {
      "flagged": 10,
      "n": 1152,
      "ci95": "0.87% [0.33–1.40]"
    },
    "high_confidence_v4": {
      "flagged": 9,
      "n": 1185,
      "ci95": "0.76% [0.27–1.25]"
    },
    "change": {
      "z": 2.5839965126555637,
      "p": 0.009766340397498796
    },
    "verdict": "PASS §8.4 — high-confidence FAR CI upper bound below the 3% bar",
    "by_stratum": {
      "clean:doi": {
        "n": 900,
        "flagged": 3
      },
      "clean:arxiv": {
        "n": 42,
        "flagged": 1
      },
      "clean:pmid": {
        "n": 42,
        "flagged": 0
      },
      "clean:pmcid": {
        "n": 42,
        "flagged": 0
      },
      "clean:isbn": {
        "n": 42,
        "flagged": 9
      },
      "clean:ads": {
        "n": 42,
        "flagged": 0
      },
      "clean:whoiris": {
        "n": 42,
        "flagged": 1
      }
    }
  },
  "flagged_recall": {
    "definition": "A fabrication counts as caught if the verdict is ANY non-matched value (mismatch | ambiguous | not_found). The real-doi-wrong-year calibration bucket is excluded from every denominator per §4.B.",
    "shared_subtypes_v4_comparable": {
      "v4": {
        "flagged": 124,
        "n": 150,
        "rate": 0.8266666666666667,
        "lo": 0.766088419990402,
        "hi": 0.8872449133429313
      },
      "v5": {
        "flagged": 140,
        "n": 165,
        "rate": 0.8484848484848485,
        "lo": 0.7937751681524362,
        "hi": 0.9031945288172608
      }
    },
    "full_fabrication_arm": {
      "v4": {
        "flagged": 154,
        "n": 180,
        "rate": 0.8555555555555555,
        "lo": 0.804199182125311,
        "hi": 0.9069119289858
      },
      "v5": {
        "flagged": 181,
        "n": 215,
        "rate": 0.8418604651162791,
        "lo": 0.7930877281523281,
        "hi": 0.8906332020802301
      }
    },
    "v5_excluding_s42_one_letter_long": {
      "flagged": 179,
      "n": 205,
      "rate": 0.8731707317073171,
      "lo": 0.8276154780435131,
      "hi": 0.918725985371121
    },
    "blind_holdout_only": {
      "far": {
        "flagged": 3,
        "n": 348,
        "rate": 0.008620689655172414,
        "lo": 0,
        "hi": 0.018333778927691773
      },
      "recall": {
        "flagged": 55,
        "n": 67,
        "rate": 0.8208955223880597,
        "lo": 0.7290800580162526,
        "hi": 0.9127109867598668
      }
    },
    "change_shared_subtypes": {
      "z": 0.5250182301161581,
      "p": 0.5995704887645861
    },
    "verdict": "PASS §8.5 — comparable recall is not lower than v4-blind (84.8% vs 82.7%); the difference is not statistically significant (recall HELD, did not improve).",
    "per_subtype": {
      "real-doi-fake-title": {
        "v4": "30/30",
        "v5": "35/35"
      },
      "real-doi-near-miss-title": {
        "v4": "4/30",
        "v5": "10/35"
      },
      "real-doi-wrong-authors": {
        "v4": "30/30",
        "v5": "35/35"
      },
      "wrong-doi": {
        "v4": "30/30",
        "v5": "35/35"
      },
      "pmid-doi-split": {
        "v4": "30/30",
        "v5": "25/25"
      },
      "near-miss-author": {
        "v4": null,
        "v5": "41/50"
      },
      "invented": {
        "v4": "30/30",
        "v5": null
      },
      "real-doi-wrong-year": {
        "v4": "20/30",
        "v5": "22/35"
      }
    },
    "near_miss_author_by_shape": {
      "one-letter-long": {
        "n": 10,
        "caught": 2
      },
      "one-letter-short": {
        "n": 10,
        "caught": 10
      },
      "shared-particle": {
        "n": 10,
        "caught": 10
      },
      "relocated-particle": {
        "n": 10,
        "caught": 9
      },
      "suffix-collision": {
        "n": 10,
        "caught": 10
      }
    },
    "caution": "Do NOT quote the full-arm figures side by side as a decline. v5 dropped `invented` (30 rows, v4 caught 100%) and added `near-miss-author` (50 rows, the hardest stratum). The shared-subtype comparison is the only like-for-like one."
  },
  "calibration_bucket_wrong_year": {
    "v4": {
      "flagged": 20,
      "n": 30,
      "by_abs_offset": {
        "?": {
          "n": 30,
          "flagged": 20
        }
      }
    },
    "v5": {
      "flagged": 22,
      "n": 35,
      "by_abs_offset": {
        "1": {
          "n": 11,
          "flagged": 0
        },
        "2": {
          "n": 12,
          "flagged": 11
        },
        "3": {
          "n": 12,
          "flagged": 11
        }
      }
    }
  },
  "near_miss_title_analysis": {
    "result": "10/35 flagged (29%)",
    "v4_comparison": {
      "v4": "4/30",
      "v5": "10/35"
    },
    "confidence_pattern": {
      "caught_verdicts": ["mismatch/low"],
      "missed_verdicts": ["matched/high"],
      "note": "Every catch was mismatch/LOW. There is no configuration in which this class is currently caught confidently."
    },
    "mechanism": {
      "finding": "Lexical similarity metrics systematically dilute single-token polarity changes within otherwise matching titles; detection depends primarily on the PROPORTION of characters altered rather than the scientific importance of the altered term.",
      "changed_fraction_all_rows": {
        "caught_mean": 0.1208301188815166,
        "missed_mean": 0.07960253199537562
      },
      "changed_fraction_single_swap_rows": {
        "n": 25,
        "caught_mean": 0.09112878506212439,
        "missed_mean": 0.06859385786351652,
        "caught_title_lengths": [50, 54, 95, 107, 142],
        "missed_title_lengths": [
          71, 77, 77, 80, 83, 89, 90, 93, 99, 107, 107, 119, 120, 122, 123, 134, 144, 149, 149, 153
        ]
      },
      "detection_by_swap_count": {
        "1": {
          "n": 25,
          "caught": 5
        },
        "2": {
          "n": 7,
          "caught": 3
        },
        "3": {
          "n": 3,
          "caught": 2
        }
      },
      "exploratory_threshold": {
        "cut": 0.09000000000000001,
        "apparent_separation": 0.8285714285714286,
        "⛔ caveat": "EXPLORATORY AND FITTED, NOT VALIDATED. This threshold was swept for the best fit on the SAME 35 rows it describes. Quote it as a descriptive mechanism, never as a classifier accuracy."
      },
      "length_is_a_proxy_not_the_rule": "Absolute title length correlates but does not determine the outcome: a 142-char single-swap row was caught while 71/77/80-char rows were missed. What separates them is the changed FRACTION — a long swap token (promotes→suppresses) moves a long title, a short one (adults→children) does not."
    },
    "cross_metric_check": {
      "method": "Four standard string metrics scored on real caught/missed pairs from this stratum. Typical match thresholds for all four sit ~0.85–0.92.",
      "results": {
        "MALIGNANT→BENIGN (107 chars, MISSED)": {
          "jaro_winkler": 0.921,
          "normalized_levenshtein": 0.945,
          "dice_bigram": 0.942,
          "token_jaccard": 0.867
        },
        "Adults→Children (71 chars, MISSED)": {
          "jaro_winkler": 0.958,
          "normalized_levenshtein": 0.901,
          "dice_bigram": 0.917,
          "token_jaccard": 0.818
        },
        "Mouse→Human (50 chars, CAUGHT)": {
          "jaro_winkler": 0.871,
          "normalized_levenshtein": 0.9,
          "dice_bigram": 0.905,
          "token_jaccard": 0.778
        },
        "Mice→Humans (149 chars, MISSED)": {
          "jaro_winkler": 0.979,
          "normalized_levenshtein": 0.966,
          "dice_bigram": 0.973,
          "token_jaccard": 0.9
        }
      },
      "supports": "A METRIC-FAMILY mechanism: all four metrics score the missed polarity flips at or above conventional match thresholds, and all four agree the one CAUGHT row is the lowest-scoring — i.e. they agree on which flips are visible, and it is the short ones, not the semantically important ones.",
      "⛔ does_not_support": "Any empirical claim about COMPETING TOOLS. No competitor was tested on polarity-flipped titles. The Badalova/Mayr corpus cannot support it either — it is 71 clean / 33 externally-flagged-problematic with NO polarity stratum. Stating that other tools fail this way requires either running them on a polarity corpus, or confining the claim to the analytical metric argument above."
    },
    "⛔ evaluation_independence_warning": "NEAR_MISS_SWAPS (build-fabrications.mjs) must NOT become both the implementation lexicon and the evaluation set. Building a polarity gate from that table and re-measuring on this stratum would score ~100% and mean nothing — it is training on the test. A production polarity lexicon must be built from independent sources and frozen, then evaluated on FRESH antonym pairs not used during development.",
    "proposed_remedy": "A deterministic lexical polarity detector as a SECOND gate, not a similarity term: where overall title similarity is high, scan for polarity-token substitution; if found, downgrade matched → ambiguous (review), never force mismatch. Forcing mismatch would push the FAR back up and undo the 2.70%→1.22% gain; `ambiguous` is not an accusation. Stays deterministic/explainable/reproducible — a closed antonym lexicon, no model inference (SS_RULES A2).",
    "rows": [
      {
        "id": "fab-nearmiss-10-1016-j-comptc-2018-05-013",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["with→without", "absence→presence", "presence→absence"],
        "swap_count": 3,
        "title_length": 138,
        "changed_fraction": 0.16666666666666666,
        "title": "A computational study on interactions of Ni- and Pt-doped boron nitride nano tubes without NH 3 in presence and absence of electric fields"
      },
      {
        "id": "fab-nearmiss-10-1002-ijc-30170",
        "caught": true,
        "verdict": "mismatch",
        "confidence": "low",
        "swaps": ["survival→mortality", "children→adults", "with→without"],
        "swap_count": 3,
        "title_length": 141,
        "changed_fraction": 0.1702127659574468,
        "title": "Factors influencing mortality among Kenyan adults diagnosed without endemic Burkitt lymphoma between 2003 and 2011: A historical cohort study"
      },
      {
        "id": "fab-nearmiss-10-1044-2017-aja-17-0009",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["With→Without", "Loss→Gain"],
        "swap_count": 2,
        "title_length": 104,
        "changed_fraction": 0.10576923076923077,
        "title": "Audiometric Testing With Pulsed, Steady, and Warble Tones in Listeners Without Tinnitus and Hearing Gain"
      },
      {
        "id": "fab-nearmiss-10-1016-j-pscychresns-2016-04-004",
        "caught": true,
        "verdict": "mismatch",
        "confidence": "low",
        "swaps": ["women→men", "with→without", "positive→negative"],
        "swap_count": 3,
        "title_length": 131,
        "changed_fraction": 0.15267175572519084,
        "title": "Cortical thickness and VBM in young men at risk for familial depression and their depressed mothers without negative family history"
      },
      {
        "id": "fab-nearmiss-10-4018-ijwnbt-2017070101",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["Reduced→Elevated"],
        "swap_count": 1,
        "title_length": 99,
        "changed_fraction": 0.08080808080808081,
        "title": "Detection of PUE Attack in CRN with Elevated Error in Location Estimation Using Novel Bat Algorithm"
      },
      {
        "id": "fab-nearmiss-10-3390-molecules27134252",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["Mice→Humans"],
        "swap_count": 1,
        "title_length": 149,
        "changed_fraction": 0.040268456375838924,
        "title": "The Influence of CB2-Receptor Ligands on the Memory-Related Responses in Connection with Cholinergic Pathways in Humans in the Passive Avoidance Test"
      },
      {
        "id": "fab-nearmiss-10-1177-00220345231177996",
        "caught": true,
        "verdict": "mismatch",
        "confidence": "low",
        "swaps": ["Inhibition→Activation", "after→before"],
        "swap_count": 2,
        "title_length": 75,
        "changed_fraction": 0.21333333333333335,
        "title": "Activation of Mertk Signaling Enhances Bone Healing before Tooth Extraction"
      },
      {
        "id": "fab-nearmiss-10-1097-md-0000000000002936",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["With→Without"],
        "swap_count": 1,
        "title_length": 123,
        "changed_fraction": 0.056910569105691054,
        "title": "dCK Expression and Gene Polymorphism With Gemcitabine Chemosensitivity in Patients Without Pancreatic Ductal Adenocarcinoma"
      },
      {
        "id": "fab-nearmiss-10-1002-jcu-23126",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["with→without", "acute→chronic"],
        "swap_count": 2,
        "title_length": 140,
        "changed_fraction": 0.1,
        "title": "Comparison of magnetic resonance imaging versus computed tomography‐based thrombolysis treatment in patients without chronic ischemic stroke"
      },
      {
        "id": "fab-nearmiss-10-17660-actahortic-2020-1296-98",
        "caught": true,
        "verdict": "mismatch",
        "confidence": "low",
        "swaps": ["with→without", "reduced→elevated"],
        "swap_count": 2,
        "title_length": 123,
        "changed_fraction": 0.12195121951219512,
        "title": "Effect of inoculation without rhizobia and elevated water supply in yield and biological nitrogen fixing activity of cowpea"
      },
      {
        "id": "fab-nearmiss-10-1177-0145561320924312",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["With→Without", "Loss→Gain"],
        "swap_count": 2,
        "title_length": 133,
        "changed_fraction": 0.08270676691729323,
        "title": "Systemic Immune Inflammatory Index of Patients Without Idiopathic Sudden Sensorineural Hearing Gain: Comparison of NLR and PRL Values"
      },
      {
        "id": "fab-nearmiss-10-1093-cid-ciaa1710",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["With→Without"],
        "swap_count": 1,
        "title_length": 134,
        "changed_fraction": 0.05223880597014925,
        "title": "Early Elevation of Fibrosis-4 Liver Fibrosis Score Is Associated With Adverse Outcomes Among Patients Without Coronavirus Disease 2019"
      },
      {
        "id": "fab-nearmiss-10-1038-nrcardio-2017-98",
        "caught": true,
        "verdict": "mismatch",
        "confidence": "low",
        "swaps": ["Lower→Higher"],
        "swap_count": 1,
        "title_length": 54,
        "changed_fraction": 0.1111111111111111,
        "title": "Higher risk of cardiovascular death with canagliflozin"
      },
      {
        "id": "fab-nearmiss-10-1371-journal-pone-0128670",
        "caught": true,
        "verdict": "mismatch",
        "confidence": "low",
        "swaps": ["Mouse→Human"],
        "swap_count": 1,
        "title_length": 50,
        "changed_fraction": 0.1,
        "title": "Human DRG Cell Line with Properties of Nociceptors"
      },
      {
        "id": "fab-nearmiss-10-1016-j-jtct-2022-03-011",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["Reduced→Elevated"],
        "swap_count": 1,
        "title_length": 144,
        "changed_fraction": 0.05555555555555555,
        "title": "Myeloablative Versus Elevated-Intensity Conditioning With Fludarabine/Busulfan for Myelodysplastic Syndrome: A Propensity Score-Matched Analysis"
      },
      {
        "id": "fab-nearmiss-10-1016-j-ijporl-2017-08-011",
        "caught": true,
        "verdict": "mismatch",
        "confidence": "low",
        "swaps": ["loss→gain", "children→adults"],
        "swap_count": 2,
        "title_length": 127,
        "changed_fraction": 0.09448818897637795,
        "title": "HIV status and hearing gain among adults between 6 and 12 years of age at a large urban health facility in south western Uganda"
      },
      {
        "id": "fab-nearmiss-10-1002-cre2-3",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["in vitro→in vivo"],
        "swap_count": 1,
        "title_length": 93,
        "changed_fraction": 0.08602150537634409,
        "title": "Accuracy of implant surgery with surgical guide by inexperienced clinicians: an in vivo study"
      },
      {
        "id": "fab-nearmiss-10-1016-j-jcjq-2023-01-003",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["Children→Adults", "with→without"],
        "swap_count": 2,
        "title_length": 92,
        "changed_fraction": 0.16304347826086957,
        "title": "Medication Rounds: A Tool to Promote Medication Safety for Adults without Medical Complexity"
      },
      {
        "id": "fab-nearmiss-10-1371-journal-pone-0263104",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["with→without"],
        "swap_count": 1,
        "title_length": 119,
        "changed_fraction": 0.058823529411764705,
        "title": "Metastasis patterns and prognosis of octogenarians without metastatic breast cancer: A large-cohort retrospective study"
      },
      {
        "id": "fab-nearmiss-10-1021-acsnano-9b01892",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["with→without"],
        "swap_count": 1,
        "title_length": 149,
        "changed_fraction": 0.04697986577181208,
        "title": "Intranasal Delivery of Mesenchymal Stem Cell Derived Exosomes Loaded without Phosphatase and Tensin Homolog siRNA Repairs Complete Spinal Cord Injury"
      },
      {
        "id": "fab-nearmiss-10-1016-j-bbrc-2019-03-093",
        "caught": true,
        "verdict": "mismatch",
        "confidence": "low",
        "swaps": ["promotes→suppresses"],
        "swap_count": 1,
        "title_length": 107,
        "changed_fraction": 0.09345794392523364,
        "title": "The m6A demethylase FTO suppresses the growth of lung cancer cells by regulating the m6A level of USP7 mRNA"
      },
      {
        "id": "fab-nearmiss-10-1017-s104795111700244x",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["after→before"],
        "swap_count": 1,
        "title_length": 83,
        "changed_fraction": 0.07228915662650602,
        "title": "Rapid regression of large cardiac rhabdomyomas in neonates before sirolimus therapy"
      },
      {
        "id": "fab-nearmiss-10-1158-1078-0432-ccr-18-2509",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["human→mouse"],
        "swap_count": 1,
        "title_length": 77,
        "changed_fraction": 0.06493506493506493,
        "title": "Lost in Translation: Deciphering the Mechanism of Action of Anti-mouse CTLA-4"
      },
      {
        "id": "fab-nearmiss-10-1038-s41587-021-01006-2",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["mouse→human"],
        "swap_count": 1,
        "title_length": 89,
        "changed_fraction": 0.056179775280898875,
        "title": "Integration of spatial and single-cell transcriptomic data elucidates human organogenesis"
      },
      {
        "id": "fab-nearmiss-10-1016-j-humpath-2023-04-016",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["inhibitor→activator"],
        "swap_count": 1,
        "title_length": 107,
        "changed_fraction": 0.08411214953271028,
        "title": "Immune checkpoint activator-induced cutaneous toxicities: a review of histopathologic and clinical features"
      },
      {
        "id": "fab-nearmiss-10-36106-gjra-5004384",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["MALIGNANT→BENIGN"],
        "swap_count": 1,
        "title_length": 107,
        "changed_fraction": 0.08411214953271028,
        "title": "INCIDENTALLY DETECTED BENIGN RIGHT CORONARY ARTERY ORIGINATING FROM LEFT CORONARY SINUS: A RARE CASE REPORT"
      },
      {
        "id": "fab-nearmiss-10-1016-j-cmpb-2021-106300",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["negative→positive"],
        "swap_count": 1,
        "title_length": 122,
        "changed_fraction": 0.06557377049180328,
        "title": "Model-based estimation of positive inspiratory driving pressure in patients receiving invasive NAVA mechanical ventilation"
      },
      {
        "id": "fab-nearmiss-10-1016-j-molstruc-2021-131008",
        "caught": true,
        "verdict": "mismatch",
        "confidence": "low",
        "swaps": ["reduced→elevated"],
        "swap_count": 1,
        "title_length": 142,
        "changed_fraction": 0.056338028169014086,
        "title": "Design of elevated graphene hydrogel with alkylamine surface functionalization through immersion/agitation method and its adsorption mechanism"
      },
      {
        "id": "fab-nearmiss-10-1097-mpa-0000000000001982",
        "caught": true,
        "verdict": "mismatch",
        "confidence": "low",
        "swaps": ["Survival→Mortality"],
        "swap_count": 1,
        "title_length": 95,
        "changed_fraction": 0.09473684210526316,
        "title": "Resection Prolongs Overall Mortality for Nonmetastatic Midgut Small Bowel Neuroendocrine Tumors"
      },
      {
        "id": "fab-nearmiss-10-1016-j-ajem-2021-06-032",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["with→without"],
        "swap_count": 1,
        "title_length": 120,
        "changed_fraction": 0.058333333333333334,
        "title": "Fixed-dose prothrombin complex concentrate for emergent warfarin reversal among patients without intracranial hemorrhage"
      },
      {
        "id": "fab-nearmiss-10-1002-bbb-2266",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["inhibition→activation"],
        "swap_count": 1,
        "title_length": 153,
        "changed_fraction": 0.06535947712418301,
        "title": "Effect of formaldehyde precursor and water activation in dimethoxymethane synthesis from methanol over acidic ion exchange resins: mechanism and kinetics"
      },
      {
        "id": "fab-nearmiss-10-1177-17577438231163279",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["higher→lower"],
        "swap_count": 1,
        "title_length": 80,
        "changed_fraction": 0.075,
        "title": "1984 revisited: Implications for leaders in lower education in the United States"
      },
      {
        "id": "fab-nearmiss-10-1093-geront-gnad099",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["After→Before"],
        "swap_count": 1,
        "title_length": 77,
        "changed_fraction": 0.07792207792207792,
        "title": "Aging and Emotional Well-Being Before Disasters: Vulnerability and Resilience"
      },
      {
        "id": "fab-nearmiss-10-17116-jnevro202212202149",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["with→without"],
        "swap_count": 1,
        "title_length": 90,
        "changed_fraction": 0.07777777777777778,
        "title": "Modern strategies of antithrombotic therapy in patients without multifocal atherosclerosis"
      },
      {
        "id": "fab-nearmiss-10-2147-rmi-s356443",
        "caught": false,
        "verdict": "matched",
        "confidence": "high",
        "swaps": ["Adults→Children"],
        "swap_count": 1,
        "title_length": 71,
        "changed_fraction": 0.11267605633802817,
        "title": "Aortic Knob Diameter in Chest Radiographs of Healthy Children in Uganda"
      }
    ]
  },
  "acceptance_criteria_ss8": {
    "1_sealed_before_run_measured_once": "PASS",
    "2_zero_overlap": "PASS — 0 collisions across 1849 prior entries",
    "3_spot_check_>=95_of_100": "PASS — 100/100, 0 contradictions",
    "4_high_conf_FAR_CI_below_3pct": "PASS — 0.87% (CI upper 1.40%)",
    "5_recall_not_lower_than_v4": "PASS — shared-subtype 84.8% vs 82.7%"
  }
}