{
  "schemaVersion": "kingy-document-independent-final-audit/1.0.0",
  "studyId": "document-comparison-v24-20261018.1",
  "reviewedAtUtc": "2026-10-18T17:58:00.955061+00:00",
  "attemptsCovered": 240,
  "all240ActualOutputsSemanticallyReviewed": true,
  "originalFrozenResultsChanged": false,
  "humanValidation": false,
  "scoreValidityPassed": true,
  "uniqueBundlesReviewed": 80,
  "summary": "All 240 actual outputs were covered by review of 80 exact semantic/source bundles, each mapped to its three original attempts. No unsupported frozen pass or wrong factual value/role was found. Kingy supplied complete cited support on all answerable tasks. Native supplied correct requested factual values throughout; its Q11/Q29 claimed chunks omitted the requested support, and its complete, supported Q23 paraphrase remains frozen unresolved. Both systems cleanly declined all ten wholly absent requests. The frozen conservative bounds remain valid within their stated finite-recognition scope; these results do not establish an overall semantic winner.",
  "scoreValidityScope": "Passed only for the unchanged prospective scorer results interpreted as conservative finite-recognition lower/upper bounds for complete supported answers, citation precision and clean abstention in this five-PDF/forty-task study. A frozen unresolved outcome is not a demonstrated factual error. Native Q23 is semantically complete and sourced despite remaining unresolved; no alias or result was changed. Native Q11/Q29 failures concern emitted citation support, not their correct answer values. This same-Mac independent-context AI audit is not human adjudication, second-hardware reproduction, timing/privacy validation, release qualification or a broad workflow/model superiority claim.",
  "inputSha256": {
    "kingy-common.jsonl": "cf85176126d315b58160d3cbb60d57f868afb6ed9e84b87cd85df86f5917b065",
    "lmstudio-common.jsonl": "f2e2baa5187b9a4cea5b8c3810381e0780e95bdb253558cb5f423ca3c5bc8424"
  },
  "analysisCompletionSha256": "aaadd02fde4486dc2720f43034bc95913a2f69e9fcfc8eedc3609cb549fb8e7d",
  "comparisonFreezeSha256": "f968a685190b843d121167b1031130e79a8daaf5a86da13effef37cb7af45936",
  "honestCaseLevelDifferences": [
    {
      "advantage": "Kingy supplied the requested source support on two cross-page tasks where the native answer values were correct but their claimed chunks omitted both requested facts.",
      "questionIds": [
        "V24-Q11",
        "V24-Q29"
      ],
      "examples": "Q11: Kingy cited the page-3 sleeve sentence and page-4 reset/deadline sentence. Native cited a later chunk beginning after those facts and continuing through pages 5/6. Q29: Kingy cited the page-3 muster location and page-4 shutter deadline. Native cited a chunk that ended early on page 3 before either fact.",
      "scope": "Same two exact outputs recur in each of the three repetitions; this is a supporting-evidence advantage on two unique questions, not a factual-answer-accuracy win or an overall model win."
    },
    {
      "advantage": "Native delivered shorter complete answers on specific tasks while retaining the required facts and supporting native citations.",
      "questionIds": [
        "V24-Q09",
        "V24-Q13",
        "V24-Q15",
        "V24-Q19",
        "V24-Q23"
      ],
      "examples": "Q09 and Q15: native states both quantities, their physical units and the correct roles in one sentence; Kingy also copies neighboring preparation/inspection prose. Q13 and Q19: native gives only the requested current value/date; Kingy additionally copies a contact/owner and an explicitly archived value/date. Neither Kingy answer falsely treats the archived value as current. Q23: native gives the rack and a complete event-based deadline paraphrase with a chunk containing both actual source anchors. Kingy copies the location plus an extended reweighing procedure.",
      "scope": "A bounded presentation and concision advantage, based on the actual field text, not a speed, privacy or general quality advantage. Q23 remains unresolved in the untouched frozen score."
    },
    {
      "advantage": "Tie in the factual values and requested role bindings supplied on all thirty answerable tasks, after contextual semantic review.",
      "questionIds": [
        "V24-Q01",
        "V24-Q02",
        "V24-Q03",
        "V24-Q04",
        "V24-Q05",
        "V24-Q06",
        "V24-Q07",
        "V24-Q08",
        "V24-Q09",
        "V24-Q10",
        "V24-Q11",
        "V24-Q12",
        "V24-Q13",
        "V24-Q14",
        "V24-Q15",
        "V24-Q16",
        "V24-Q17",
        "V24-Q18",
        "V24-Q19",
        "V24-Q20",
        "V24-Q21",
        "V24-Q22",
        "V24-Q23",
        "V24-Q24",
        "V24-Q25",
        "V24-Q26",
        "V24-Q27",
        "V24-Q28",
        "V24-Q29",
        "V24-Q30"
      ],
      "examples": "Native Q11/Q29 values are correct even though the cited chunks are insufficient. Native Q23 after the end of blotting conveys the same requested trigger as after blotting ends; its original unresolved status is preserved.",
      "scope": "Independent same-Mac AI semantic finding, distinct from the frozen supported-answer metric and finite-matcher fact labels; no original grade is replaced."
    },
    {
      "advantage": "Tie in clean abstention on all ten wholly absent requests.",
      "questionIds": [
        "V24-Q31",
        "V24-Q32",
        "V24-Q33",
        "V24-Q34",
        "V24-Q35",
        "V24-Q36",
        "V24-Q37",
        "V24-Q38",
        "V24-Q39",
        "V24-Q40"
      ],
      "examples": "Kingy emits an explicit normalized abstention with no value/citation; native emits only the required field label with UNKNOWN. Both avoid guessing the absent insurance, founder, budget, supplier, permit, serial-number and instrument-specification values.",
      "scope": "Ten unique absent requests repeated three times, not thirty independent absence trials."
    }
  ],
  "coverage": {
    "systems": 2,
    "repetitionsPerSystem": 3,
    "attemptsPerSystem": 120,
    "uniqueQuestions": 40,
    "answerableUniqueQuestions": 30,
    "whollyAbsentUniqueQuestions": 10,
    "exactBundlesPerSystem": 40,
    "attemptsMappedPerBundle": 3,
    "groupingDefinition": "Exact raw final output, normalized answer fields, all emitted citation content/IDs/pages, requested task, frozen grade/response-only grade, errors and additional-output state. Runtime identity and timing are excluded because this is a semantic/source audit; every original row is individually mapped. Automatically displayed unclaimed source cards are not emitted citations.",
    "all240Mapping": {
      "artifact": "[redacted metadata reference; sha256 67f46292ff456b2c908b91777ee709576aee37e5afc391557356e4b78ae1e8e9]",
      "sha256": "a7703206b02c01afd377e0ce4837a902a60e0aa6dea5c867906ad92be1166919"
    },
    "actualBundleContentAndMapping": {
      "artifact": "[redacted metadata reference; sha256 11d6620bc7dcced6e15940fb04dcd9758eaaf439ac862390bc86feee41e4818a]",
      "sha256": "e54179d8b9484d5ab36aac47d9c5ba900792dfb742b3a44fd44454c7c073e18d"
    }
  },
  "independentSemanticFindings": {
    "all30AnswerableQuestionValuesAndRoleBindingsCorrectForBothSystems": true,
    "nativeCitationSupportMissingQuestionIds": [
      "V24-Q11",
      "V24-Q29"
    ],
    "nativeSemanticallyCorrectSupportedButFrozenUnresolvedQuestionIds": [
      "V24-Q23"
    ],
    "all10WhollyAbsentRequestsCleanlyDeclinedByBoth": true,
    "factualErrorNotInferredFromUnresolvedOrWrongCitation": true,
    "semanticAssessmentIsSeparateFromFrozenResults": true,
    "all80ContextualAssessments": {
      "artifact": "[redacted metadata reference; sha256 f6d47caba7f6efae9cd4a659b69164fe4c9ce146a34fe1f7580ffa7f63195c1c]",
      "sha256": "078a06fc7ccbc40f835cebc9f1fec458676ef092066478430cc792d6d0f04f66"
    }
  },
  "sourceChecks": {
    "all30ActualPdfPagesReextractedAndMatchedPreviouslyReviewedUnchangedPdfBytes": true,
    "uniqueEmittedQuotesIndependentlyLocated": 50,
    "allEmittedQuotesGenuineUniqueFullTokenConsecutiveSourceSpans": true,
    "originalBoundariesChecked": true,
    "eachRequestedObligationCheckedAgainstItsAssociatedEmittedCitation": true,
    "allFiveCrossPageTasksReviewedForBothSystems": true,
    "sourceAndAdapterAudit": {
      "artifact": "[redacted metadata reference; sha256 a65b5ac0651e922d68b7cf8ae67cc37acbf42e51f628d5be6ce0d61c0e0426a4]",
      "sha256": "00da7aca2595e8475c9c9e34e66cf6afed54f1a168cac6a714702629c34b0f3a"
    }
  },
  "frozenResultsPreserved": {
    "KingyEachRepetition": {
      "completeSupportedAnswers": {
        "pass": 30,
        "total": 30,
        "unresolved": 0
      },
      "citationPrecision": {
        "pass": 35,
        "total": 35,
        "unresolved": 0
      },
      "cleanAbstentions": {
        "pass": 10,
        "total": 10,
        "unresolved": 0
      }
    },
    "nativeEachRepetition": {
      "completeSupportedAnswers": {
        "pass": 27,
        "total": 30,
        "unresolved": 1
      },
      "citationPrecision": {
        "pass": 31,
        "total": 35,
        "unresolved": 2
      },
      "cleanAbstentions": {
        "pass": 10,
        "total": 10,
        "unresolved": 0
      }
    },
    "all80UniqueBundleFrozenGradeRecomputationsMatch": true,
    "bothOriginalInputHashesReverifiedAfterAudit": true,
    "noPostOutputAliasesQuestionsSourcesOrGradesChanged": true
  },
  "casePresentationLengthObservations": {
    "definition": "Decoded answer-field characters only, excluding labels/citation markers; not a latency measure.",
    "observations": [
      {
        "questionId": "V24-Q09",
        "KingyAnswerFieldCharacters": 334,
        "nativeAnswerFieldCharacters": 78,
        "definition": "Exact decoded answer-field values joined by a single newline, excluding field labels and emitted citation markers; source chunks are assessed separately."
      },
      {
        "questionId": "V24-Q13",
        "KingyAnswerFieldCharacters": 162,
        "nativeAnswerFieldCharacters": 18,
        "definition": "Exact decoded answer-field values joined by a single newline, excluding field labels and emitted citation markers; source chunks are assessed separately."
      },
      {
        "questionId": "V24-Q15",
        "KingyAnswerFieldCharacters": 321,
        "nativeAnswerFieldCharacters": 83,
        "definition": "Exact decoded answer-field values joined by a single newline, excluding field labels and emitted citation markers; source chunks are assessed separately."
      },
      {
        "questionId": "V24-Q19",
        "KingyAnswerFieldCharacters": 138,
        "nativeAnswerFieldCharacters": 14,
        "definition": "Exact decoded answer-field values joined by a single newline, excluding field labels and emitted citation markers; source chunks are assessed separately."
      },
      {
        "questionId": "V24-Q23",
        "KingyAnswerFieldCharacters": 460,
        "nativeAnswerFieldCharacters": 49,
        "definition": "Exact decoded answer-field values joined by a single newline, excluding field labels and emitted citation markers; source chunks are assessed separately."
      }
    ],
    "evidence": {
      "artifact": "[redacted metadata reference; sha256 65d431998650e8f305e4e23a852917b1c6afc14cc58787f0b159726da5aae2fc]",
      "sha256": "8c16af2d27b07de49809e0d35fa6adb04db14280fea8cc3ffaf3151c67a59482"
    }
  },
  "independence": {
    "reviewType": "same-Mac independent-context AI audit",
    "reviewedAfterCoordinatorReportedAll240StoppedAndPregradeSealed": true,
    "noInferenceModelGuiBrowserNetworkOrSpendUsedDuringAudit": true,
    "noHumanAdjudication": true,
    "noSecondHardwareReproduction": true,
    "priorStudyResultsOrPacksNotRead": true
  },
  "limits": [
    "Forty unique task clusters repeated three times per system; 240 attempts are not 240 independent questions.",
    "The native Q23 recognition limitation and native Q11/Q29 citation-support defects must remain explicit in any publication.",
    "Conciseness observations concern these actual decoded answers only; no general usability, privacy, speed or model-quality superiority is established.",
    "Semantic/source validity alone does not pass separate installation, security, prospective-method, publication or launch gates."
  ]
}