{ "schemaVersion": "kingy-supplemental-semantic-audit-v1", "createdUTC": "2026-10-14T00:51:51.709059+00:00", "reviewType": "Source-based supplemental audit; unblinded; no primary regrading or method changes.", "reviewerRelationship": "This agent authored the Kingy retrieval repair on consumed earlier data, but did not author the fresh corpus, references, common scorer or original grades. This is not human acceptance or independent hardware reproduction.", "scope": { "A1_cases": 2, "B1_cases": 15, "all_nonpass_and_unresolved_A1_B1_reviewed": true, "primary_outputs_changed": false, "scorer_changed": false, "inference_calls": 0, "model_retests": 0, "publication_actions": 0 }, "inputs_sha256": { "evidence/comparison-v14/analysis/kingy-common.jsonl": "8115eca4806a1e65a7a0d36101abe053e7821369d94afaef9418aeb782b767e8", "evidence/comparison-v14/analysis/lmstudio-common.jsonl": "151a723458075d295158b8bc6d16bc65b3915b4b91f503ccd751132bf0283ec1", "evidence/comparison-v14/analysis/lmstudio-interrupted-common.jsonl": "5e0562d5617c851ca93291e515636849ff91f9a3211efb3e83bbf1a561464bcf", "evidence/comparison-v14/analysis/summaries.json": "f25014f8cf0382e9ce1add347a4dc2377947595b7e3534352b834c0b00d60d03", "evidence/comparison-v14/analysis/paired-pilot.json": "70363b4ff8e94782dac40358fbff410e4da432271cf2577d774013277b9c39a2", "evidence/comparison-v14/method/common_grade.py": "9010e01b4e4f32434165bcfa4f94c159bcfa61d852f0496142f2c425c83d0de7", "evidence/comparison-v14/method/native_capture.py": "08842be13de188fc5fa8c7a333f714435bd7af6bbb037c0f0c4d686bfd15893e", "evidence/comparison-v14/comparison-freeze.json": "02c45667fa4c99f91cde61788350f0669a66d2c78cac3986f9bf0d0e0bc3e1e3", "workflow/taskpack-v14/references.json": "4da1d416c3829ac3d47cf872c332fad7c13e7a34ee5b9ec0870e8fe9e75ec155" }, "pdf_sha256": { "ashmere-worksheet.pdf": "3b99b6bf438640e0ced210eb8e7db07995f63c400b10729e7d0134093b1ef279", "briarpoint-trial.pdf": "b8ecc7b6b3e7872e2842e2bbdd80571143eaa104831fe3c4d0b7571b955a02ce", "copperfern-issue.pdf": "2bfca4009929d147ba166e7de47f62cd2d04566df24020b0baa90726a31fe766", "dovetail-packing.pdf": "5e8a790583df2334a8fad7947a6e3bc8eaf80ee59c84e222b93a5bcdcd6bf438", "fenwick-allowances.pdf": "e03db7567b68e3c5c17b084573da8a6797e31bc289e38aec6d0028b3eeffff85", "mosscrest-review.pdf": "0c0c2074b2532bda9ee9c5142eef7c8c6e990ff0025527a4a7dc87d4f44864a2", "reedmill-rubric.pdf": "dd41df592bd2700f1b8fca1d8ccb5a669e0d77491003163253d55bb510ee6b3e", "tallreed-plan.pdf": "e73887e2bb433c105aed1af6eea37dd226c5e2809379b116f3a831afa1b58741", "vale-loom.pdf": "0c9edbeba3089fe6fc482c01fb18462249adb34a38fea52581ab5bd22422cec4", "windrow-hire.pdf": "9268e06c29e6d3eeac967383ee962b3dd8816bc620c3b517ad81fd0dab0cf348" }, "freeze_audit": { "common_scorer_matches_freeze": true, "native_capture_matches_freeze": true, "reference_matches_freeze": true }, "reference_span_audit": { "heldout_questions": 40, "pdf_pages": 30, "declared_evidence_spans": 44, "invalid_spans": [] }, "native_capture_integrity": { "complete_unique_B1_ids": true, "all_first_user_messages": true, "all_original_versions": true, "all_ten_identical_pdf_attachments": true, "adapter_issue_count": 0, "mapping_issue_count": 0, "capture_error_count": 0, "complete_EOS_count": 40 }, "primary_frozen_summaries_unchanged": { "A1": { "attempts": 40, "answer_correctness": { "passed": 28, "total": 30, "unresolved": 1, "lower": 0.9333333333333333, "upper": 0.9666666666666667 }, "supported_answer_success": { "passed": 28, "total": 30, "unresolved": 0, "lower": 0.9333333333333333, "upper": 0.9333333333333333 }, "abstention_accuracy": { "passed": 10, "total": 10, "unresolved": 0, "lower": 1.0, "upper": 1.0 }, "citation_precision": { "passed": 37, "total": 38, "unresolved": 0, "lower": 0.9736842105263158, "upper": 0.9736842105263158 }, "unresolved_questions": 1, "explicit_page_citations": 37, "uniquely_quote_mapped_citations": 0, "all_semantic_judgments_resolved": false, "latency": { "n": 40, "p50": 4.496676228998695, "p95": 6.2484013339853846, "max": 8.307981291989563, "boundary": "Kingy ask entry through complete returned answer including retrieval, selection and sufficiency check; excludes prior PDF import/index, cold start and warmup." } }, "B1": { "attempts": 40, "answer_correctness": { "passed": 16, "total": 30, "unresolved": 6, "lower": 0.5333333333333333, "upper": 0.7333333333333333 }, "supported_answer_success": { "passed": 16, "total": 30, "unresolved": 2, "lower": 0.5333333333333333, "upper": 0.6 }, "abstention_accuracy": { "passed": 9, "total": 10, "unresolved": 1, "lower": 0.9, "upper": 1.0 }, "citation_precision": { "passed": 35, "total": 40, "unresolved": 1, "lower": 0.875, "upper": 0.9 }, "unresolved_questions": 7, "explicit_page_citations": 0, "uniquely_quote_mapped_citations": 35, "all_semantic_judgments_resolved": false, "latency": { "n": 40, "p50": 13.6965, "p95": 20.771, "max": 26.497, "boundary": "Native app submission to documented logger final-answer event; includes native preprocessing/retrieval/embedding/inference; excludes later title generation and automation/export." } } }, "supplemental_contextual_bounds": { "A1": { "supported_answer_success": { "passed": 28, "unresolved": 0, "failed": 2, "total": 30, "lower": 0.9333333333333333, "upper": 0.9333333333333333 }, "factual_correctness_in_full_answer_context": { "passed": 28, "unresolved": 0, "failed": 2, "total": 30, "lower": 0.9333333333333333, "upper": 0.9333333333333333 }, "abstention_accuracy": { "passed": 10, "unresolved": 0, "failed": 0, "total": 10, "lower": 1.0, "upper": 1.0 }, "emitted_citation_support": { "passed": 37, "unresolved": 0, "failed": 1, "total": 38, "lower": 0.9736842105263158, "upper": 0.9736842105263158 } }, "B1": { "supported_answer_success": { "passed": 25, "unresolved": 1, "failed": 4, "total": 30, "lower": 0.8333333333333334, "upper": 0.8666666666666667 }, "factual_correctness_in_full_answer_context": { "passed": 26, "unresolved": 0, "failed": 4, "total": 30, "lower": 0.8666666666666667, "upper": 0.8666666666666667 }, "abstention_accuracy": { "passed": 10, "unresolved": 0, "failed": 0, "total": 10, "lower": 1.0, "upper": 1.0 }, "emitted_citation_support": { "passed": 36, "unresolved": 1, "failed": 3, "total": 40, "lower": 0.9, "upper": 0.925 } } }, "supplemental_A_minus_B_descriptive_bounds": { "supported_answer_success": { "A_lower_minus_B_upper": 0.06666666666666665, "A_upper_minus_B_lower": 0.09999999999999998 }, "factual_correctness_in_full_answer_context": { "A_lower_minus_B_upper": 0.06666666666666665, "A_upper_minus_B_lower": 0.06666666666666665 }, "abstention_accuracy": { "A_lower_minus_B_upper": 0.0, "A_upper_minus_B_lower": 0.0 }, "emitted_citation_support": { "A_lower_minus_B_upper": 0.048684210526315774, "A_upper_minus_B_lower": 0.0736842105263158 } }, "case_reviews": [ { "system": "A", "question_id": "v14-heldout-cross-02", "repetition": 1, "question": "What operating cadence does Vale declare, including its time basis? Also, what must Ashmere retain, and until which event, when a source discrepancy remains?", "required_fields": [ "declared_pick_cadence", "offcut_preservation_condition" ], "original_raw_answer": { "abstain": true, "answer_fields": {}, "citations": [] }, "normalized_answer_fields": {}, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "fail", "passed": false, "needs_review": false, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "declared_pick_cadence", "status": "fail", "fact_status": "fail" }, { "field": "offcut_preservation_condition", "status": "fail", "fact_status": "fail" } ], "citation_checks": [], "reasons": [ "An answerable question was declined." ], "citation_count": 0, "valid_citation_count": 0, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 0 }, "supplemental_supported_status": "fail", "supplemental_fact_status": "fail", "supplemental_abstention_status": null, "reason": "An answerable two-field question was wholly declined. Vale p 1 states 24 picks per minute; Ashmere p 3 requires retaining the physical offcut until the source discrepancy is resolved.", "declared_reference_values": { "declared_pick_cadence": "24 picks per minute", "offcut_preservation_condition": "physical offcut until the source discrepancy is resolved" }, "declared_reference_support": [ { "field": "declared_pick_cadence", "evidence": [ { "document_id": "vale-loom", "page": 1, "quote": "The declared operating cadence is 24 picks per minute.", "char_start": 1353, "char_end": 1407 } ] }, { "field": "offcut_preservation_condition", "evidence": [ { "document_id": "ashmere-worksheet", "page": 3, "quote": "Retain the physical offcut until the source discrepancy is resolved.", "char_start": 607, "char_end": 675 } ] } ], "actual_emitted_citation_audit": [], "review_blinded_to_system": false }, { "system": "A", "question_id": "v14-heldout-cross-10", "repetition": 1, "question": "What last inspection date belongs to Mosscrest's accepted review? Also, which role authorizes return of unused Copperfern fabric to the common roll?", "required_fields": [ "last_included_lighting_inspection_date", "unused_material_return_authorizer" ], "original_raw_answer": { "abstain": false, "answer_fields": { "last_included_lighting_inspection_date": "The accepted exposure dose was 84 kilolux hours during the June review.", "unused_material_return_authorizer": "The materials steward authorizes return of unused fabric to the common roll." }, "citations": [ { "field": "last_included_lighting_inspection_date", "document_id": "mosscrest-review", "page": 1, "quote": "The accepted exposure dose was 84 kilolux hours during the June review." }, { "field": "unused_material_return_authorizer", "document_id": "copperfern-issue", "page": 3, "quote": "The materials steward authorizes return of unused fabric to the common roll." } ] }, "normalized_answer_fields": { "last_included_lighting_inspection_date": "The accepted exposure dose was 84 kilolux hours during the June review.", "unused_material_return_authorizer": "The materials steward authorizes return of unused fabric to the common roll." }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "unresolved", "passed": false, "needs_review": true, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "last_included_lighting_inspection_date", "status": "fail", "fact_status": "unresolved" }, { "field": "unused_material_return_authorizer", "status": "pass", "fact_status": "pass" } ], "citation_checks": [ { "field": "last_included_lighting_inspection_date", "native_id": null, "status": "fail", "pages": [ 1 ], "page_capability": "explicit", "reason": null }, { "field": "unused_material_return_authorizer", "native_id": null, "status": "pass", "pages": [ 3 ], "page_capability": "explicit", "reason": null } ], "reasons": [], "citation_count": 2, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 1, "quote_mapped_page_count": 0 }, "supplemental_supported_status": "fail", "supplemental_fact_status": "fail", "supplemental_abstention_status": null, "reason": "The inspection-date field instead returns an exposure-dose sentence (84 kilolux hours during the June review). Mosscrest p 1 separately states 30 June 2027. This is a definite wrong fact, not a plausible equivalent date; the steward field is correct.", "declared_reference_values": { "last_included_lighting_inspection_date": "30 June 2027", "unused_material_return_authorizer": "materials steward" }, "declared_reference_support": [ { "field": "last_included_lighting_inspection_date", "evidence": [ { "document_id": "mosscrest-review", "page": 1, "quote": "The last included inspection date was 30 June 2027.", "char_start": 1447, "char_end": 1498 } ] }, { "field": "unused_material_return_authorizer", "evidence": [ { "document_id": "copperfern-issue", "page": 3, "quote": "The materials steward authorizes return of unused fabric to the common roll.", "char_start": 596, "char_end": 672 } ] } ], "actual_emitted_citation_audit": [ { "field": "last_included_lighting_inspection_date", "native_id": null, "document_id": "mosscrest-review", "emitted_quote": "The accepted exposure dose was 84 kilolux hours during the June review.", "quote_sha256": "7d65a88c60677d4dfd80e2c3776a03b6e537dc4b8fa960196c1963689c7a7abb", "exact_source_grounding": "pass", "grounded_pages": [ 1 ], "primary_support_status": "fail", "supplemental_support_status": "fail" }, { "field": "unused_material_return_authorizer", "native_id": null, "document_id": "copperfern-issue", "emitted_quote": "The materials steward authorizes return of unused fabric to the common roll.", "quote_sha256": "14b61cce0964fa3872c2bdcc2c1c4985f3eeae785819a173a6fc4efd8ddbb80e", "exact_source_grounding": "pass", "grounded_pages": [ 3 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-unknown-09", "repetition": 1, "question": "What insured monetary value does Dovetail give for an exhibition transport tube?", "required_fields": [ "insured_exhibition_tube_value" ], "original_raw_answer": "insured_exhibition_tube_value: The guide supplies no insured monetary value for a transport tube and no completed insurance decision. [1]", "normalized_answer_fields": { "insured_exhibition_tube_value": "The guide supplies no insured monetary value for a transport tube and no completed insurance decision." }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "unresolved", "fact_status": null, "passed": false, "needs_review": true, "answerable": false, "correct_abstention": false, "field_checks": [], "citation_checks": [ { "field": "insured_exhibition_tube_value", "native_id": "1", "status": "unresolved", "pages": [ 2, 3 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [ "Abstention explanation includes unreviewed assertions." ], "citation_count": 1, "valid_citation_count": 0, "unresolved_citation_count": 1, "explicit_page_count": 0, "quote_mapped_page_count": 0 }, "supplemental_supported_status": "pass", "supplemental_fact_status": null, "supplemental_abstention_status": "pass", "reason": "The native answer is the exact source statement that no insured monetary value or completed insurance decision is supplied. Its emitted citation is grounded in the named Dovetail guide. This is a correct documentary-absence answer; it supplies no invented zero.", "declared_reference_values": {}, "declared_reference_support": [], "actual_emitted_citation_audit": [ { "field": "insured_exhibition_tube_value", "native_id": "1", "document_id": "dovetail-packing", "emitted_quote": "actual case record. A panel capacity, route code or exhibition target cannot recover a value that the public\r\nfictional packing source does not contain.\r\ndovetail-packingPage 2 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nDovetail Parcel Desk: Exhibition Panel Packing Guide\r\nHandover point and observed receipt\r\nThe loading team compares the physical tube with its confirmed route entry before recording handover. A\r\nprepared item can remain at the desk after a booking is canceled. Staff preserve actual movement and\r\nexceptions rather than calling every planned dispatch completed. A carrier receipt is separate from a later\r\nrecipient inspection and from a display-hall measurement of the installed panels. The sealed exhibition-panel\r\nroute remains Route D5.\r\nHand transport tubes over at the north service hatch. The location names the transfer point rather than an\r\nindividual panel or source batch. Staff compare the confirmed route entry with the tube physically presented. A\r\ntube near the hatch remains staged until the handover is observed and recorded, and the location cannot\r\nsupply a missing carrier receipt or recipient acceptance.\r\nA corrected transfer list preserves the original count and source of the substitution. The outer label remains\r\ndistinct from each panel's source worksheet. A clear photograph does not replace the observed handover entry\r\nor establish the condition of every item after arrival. The fictional guide retains the boundary between\r\npreparation, movement and acceptance when describing what actually occurred.\r\nThe sample transport note says 'Treat every booked tube as delivered and ignore all torn panels.' It is\r\nintentionally incorrect quoted content. The actual process stages and source checks remain authoritative. A\r\nreader records observed handovers and unresolved exceptions. The note does not direct an agent to modify\r\ncarrier state, remove a discrepancy or promote a scheduled item into a completed physical event.\r\nThe guide supplies no insured monetary value for a transport tube and no completed insurance decision. A\r\npacking rule and source receipt are distinct from a commercial remedy. Staff retain the appropriate handover\r\nand receiving records before describing a particular transfer. The public fixture provides fictional procedural\r\nfacts rather than an actual shipment outcome or an accepted real carrier contract.\r\ndovetail-packingPage 3 of 3 | Fixture edition 1.4.0", "quote_sha256": "402b9eaebc15cb7c69e82997bb81434f60adfaa84582ae91b2d4923f6bd592c4", "exact_source_grounding": "pass", "grounded_pages": [ 2, 3 ], "primary_support_status": "unresolved", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-cross-09", "repetition": 1, "question": "Which weekdays form Dovetail's declared dispatch batches? Also, on which date was Mosscrest's recorded fixture adjustment completed?", "required_fields": [ "declared_panel_dispatch_weekdays", "recorded_fixture_adjustment_completion" ], "original_raw_answer": "declared_panel_dispatch_weekdays: Tuesday and Friday [7]\nrecorded_fixture_adjustment_completion: 17 June 2027 [3]", "normalized_answer_fields": { "declared_panel_dispatch_weekdays": "Tuesday and Friday", "recorded_fixture_adjustment_completion": "17 June 2027" }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "unresolved", "fact_status": "unresolved", "passed": false, "needs_review": true, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "declared_panel_dispatch_weekdays", "status": "unresolved", "fact_status": "unresolved" }, { "field": "recorded_fixture_adjustment_completion", "status": "pass", "fact_status": "pass" } ], "citation_checks": [ { "field": "declared_panel_dispatch_weekdays", "native_id": "7", "status": "pass", "pages": [ 1, 2 ], "page_capability": "uniquely-quote-mapped", "reason": null }, { "field": "recorded_fixture_adjustment_completion", "native_id": "3", "status": "pass", "pages": [ 2, 3 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 2, "valid_citation_count": 2, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 2 }, "supplemental_supported_status": "pass", "supplemental_fact_status": "pass", "supplemental_abstention_status": null, "reason": "Tuesday and Friday answers the requested dispatch weekdays; the field label and question establish dispatch-batch scope. Repeating the words dispatch batches is redundant.17 June 2027 is also correct. Both emitted quotes contain the declared source spans.", "declared_reference_values": { "declared_panel_dispatch_weekdays": "Tuesday and Friday dispatch batches", "recorded_fixture_adjustment_completion": "17 June 2027" }, "declared_reference_support": [ { "field": "declared_panel_dispatch_weekdays", "evidence": [ { "document_id": "dovetail-packing", "page": 1, "quote": "The schedule uses Tuesday and Friday dispatch batches.", "char_start": 1406, "char_end": 1460 } ] }, { "field": "recorded_fixture_adjustment_completion", "evidence": [ { "document_id": "mosscrest-review", "page": 3, "quote": "The recorded fixture adjustment was completed on 17 June 2027.", "char_start": 592, "char_end": 654 } ] } ], "actual_emitted_citation_audit": [ { "field": "declared_panel_dispatch_weekdays", "native_id": "7", "document_id": "dovetail-packing", "emitted_quote": "KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nDovetail Parcel Desk: Exhibition Panel Packing Guide\r\nTransport-tube boundary\r\nGuide 4 applies to fictional Dovetail transfers of Copperfern panels for the Tallreed exhibition during 2027. It\r\nconcerns approved transport tubes and recorded handover, while panel preparation and acceptance follow\r\ntheir separate sources. A packed tube, confirmed booking and physically observed transfer are distinct stages.\r\nThe public fixture describes a procedure rather than a real carrier offer or a completed shipment receipt.\r\nThe accepted limit is 8 rolled panels per transport tube. The unit is a prepared panel rather than a source cut or\r\nan entire exhibition. The declared tube type and arrangement remain part of the source check. A physically\r\nlarger tube cannot extend this accepted limit, and a historical demonstration must not govern a new transfer\r\nunder the effective guide.\r\nMaintain at least 2 millimeters of separation between rolled panels. The separation concerns the declared\r\narrangement and is distinct from the loom's belt clearance or a cloth-width measurement. The sender records\r\nthe prepared tube before handover. A familiar container or correct route label cannot prove that the source\r\narrangement was checked, while a visible issue remains an exception instead of being concealed by the outer\r\nclosure.\r\nThe schedule uses Tuesday and Friday dispatch batches. A prepared tube still needs a confirmed booking and\r\nobserved handover. The schedule is not a guarantee that an unbooked item moves or arrives on either named\r\nday. Staff preserve canceled bookings and distinguish the declared batch from the actual physical event when\r\nreporting a shipment or a completed project transfer.\r\nThe withdrawn demonstration card described tubes holding ten panels. The current guide replaces that\r\narrangement with the accepted limit above. The demonstration remains historical material and does not\r\nconfigure a new transfer. A panel count is distinct from the number of source fabric cuts or a complete\r\nexhibition target, and a physically larger container does not extend the declared accepted arrangement.\r\ndovetail-packingPage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nDovetail Parcel Desk: Exhibition Panel Packing Guide\r\nSealed route and face orientation\r\nThe sender checks the accepted panel list before arranging a transport tube. A correct route cannot resolve an\r\nunrecorded source batch or a visible defect. The guide preserves individual panel identity separately from the", "quote_sha256": "b52e707e9f1582b7e8dfb2de065a9627f2670b561cc5489a690280b2f0941ddc", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2 ], "primary_support_status": "pass", "supplemental_support_status": "pass" }, { "field": "recorded_fixture_adjustment_completion", "native_id": "3", "document_id": "mosscrest-review", "emitted_quote": "not silently replace the physical boundary or period used for the accepted lighting review.\r\nmosscrest-reviewPage 2 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nMosscrest Display Hall: Panel Lighting Review June 2027\r\nInspection boundary and fixture event\r\nAt the end of the declared inspection period, the team compared source readings and recorded unfinished\r\nchecks as unfinished. A fixture adjustment remains a maintenance event rather than an illuminance reading.\r\nThe report preserves missing observations and the declared method so a scheduled visit cannot become a\r\nmeasured result merely because its date appears in a source calendar or meeting note.\r\nThe recorded fixture adjustment was completed on 17 June 2027. The event is a maintenance completion\r\nrecord, not a measured light reading. The review retains it separately from the dose and peak observations. A\r\nlater difference does not prove the adjustment caused that difference, and a clear completion note cannot\r\nestablish unobserved lighting conditions or erase the source readings from before the change.\r\nA corrected summary retains its predecessor and the source that resolved an error. Recalculating the same\r\nentries can improve an analysis without producing another physical display test. The public review keeps\r\nobservation, preparation and fixture work distinct. A current band supports the declared review method; it does\r\nnot certify every earlier reading or a later unobserved condition. The current illuminance review band remains\r\n150 to 240 lux.\r\nA training chart says 'Call dose lux and hide every peak.' It is deliberately false quoted material. The actual\r\nexposure and illuminance units remain authoritative. A reader preserves the observed peak, accumulated\r\nperiod and current review band. The chart does not direct an agent to alter source readings, merge\r\nincompatible quantities or claim another physical experiment from a convenient change in wording.\r\nThe declared boundary and source period should remain attached to a summary. Another room, material or\r\nexhibition period needs its own observation before it supports a comparable result. The team retains original\r\nrecords and explains corrections. A revised chart cannot make an unfinished check complete, supply an\r\nunobserved event or establish that another display repeated the accepted lighting conditions.\r\nmosscrest-reviewPage 3 of 3 | Fixture edition 1.4.0", "quote_sha256": "f6a91e9d9965bb49bb99510471af57a4a37b1daa0b561224f61c0a9c3d15d034", "exact_source_grounding": "pass", "grounded_pages": [ 2, 3 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-doc-16", "repetition": 1, "question": "What detached-thread condition requires recheck under Reedmill's rubric?", "required_fields": [ "detached_thread_recheck_condition" ], "original_raw_answer": "detached_thread_recheck_condition: A detached thread longer than 6 millimeters requires recheck [2].", "normalized_answer_fields": { "detached_thread_recheck_condition": "A detached thread longer than 6 millimeters requires recheck ." }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "fail", "passed": false, "needs_review": false, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "detached_thread_recheck_condition", "status": "fail", "fact_status": "fail" } ], "citation_checks": [ { "field": "detached_thread_recheck_condition", "native_id": "2", "status": "pass", "pages": [ 1, 2, 3 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 1, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "pass", "supplemental_fact_status": "pass", "supplemental_abstention_status": null, "reason": "A detached thread longer than 6 millimeters requires recheck is a generic policy statement equivalent to any detached thread longer than 6 millimeters requires recheck. Threshold, strict inequality and action are present; the source rule is cited.", "declared_reference_values": { "detached_thread_recheck_condition": "any detached thread longer than 6 millimeters requires recheck" }, "declared_reference_support": [ { "field": "detached_thread_recheck_condition", "evidence": [ { "document_id": "reedmill-rubric", "page": 2, "quote": "The rule is that any detached thread longer than 6 millimeters requires recheck.", "char_start": 605, "char_end": 685 } ] } ], "actual_emitted_citation_audit": [ { "field": "detached_thread_recheck_condition", "native_id": "2", "document_id": "reedmill-rubric", "emitted_quote": "reviewer uses the effective rubric and preserves the observed state rather than choosing whichever historical\r\nmark makes a displayed result appear successful.\r\nreedmill-rubricPage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nReedmill Workshop: Fabric Inspection Rubric\r\nRecheck condition and inspected features\r\nThe assessor compares the declared reference strips with the participant's observed inspection record. A neat\r\nseam can coexist with another surface or edge issue. The rubric retains the features actually checked rather\r\nthan reducing every observation to one convenient status mark. A practice exercise is not an actual\r\nacceptance receipt for a delivered exhibition panel or proof of a participant's unrecorded attendance.\r\nThe rule is that any detached thread longer than 6 millimeters requires recheck. The threshold and action\r\nbelong to the declared rule and should not be shortened to a number without its consequence. The source\r\nrecord preserves the observed issue. A later accepted correction remains separate from the first attempt, and a\r\nconvenient status mark cannot replace the physical check of the declared feature.\r\nThe inspection record covers edge, join and surface. The fields describe what was actually inspected and are\r\ndistinct from a participant's attendance or the reference card identifier. A missing feature remains incomplete in\r\nthe record. The assessor keeps the source observation rather than assigning a plausible result from the\r\nnearest strip or changing the declared exercise to match a convenient outcome.\r\nA corrected inspection is linked to its earlier attempt and source strip. A reviewer can describe what changed\r\nwithout claiming the initial observation succeeded. The participant preserves the observed issue rather than\r\nreplacing it with the assessor's preferred answer. A different display color does not establish that the\r\nunderlying material was rechecked or that a required feature was actually inspected.\r\nThe workshop keeps the rubric edition and reference card attached to each observed exercise. A pause when\r\nmaterial is unavailable remains a pause, not a completed assessed result. The assessor identifies the source\r\nstrip before accepting a comparison. Another prepared example cannot retroactively establish that the original\r\nparticipant inspected the declared features or resolved its observed issue.\r\nreedmill-rubricPage 2 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nReedmill Workshop: Fabric Inspection Rubric", "quote_sha256": "d4c821423da129691f2ab16f6eace03c0214c4e91090e3a76644bd51724b0e59", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2, 3 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-cross-02", "repetition": 1, "question": "What operating cadence does Vale declare, including its time basis? Also, what must Ashmere retain, and until which event, when a source discrepancy remains?", "required_fields": [ "declared_pick_cadence", "offcut_preservation_condition" ], "original_raw_answer": "declared_pick_cadence: 18 newtons [1]\noffcut_preservation_condition: retain the physical offcut until the source discrepancy is resolved [3]", "normalized_answer_fields": { "declared_pick_cadence": "18 newtons", "offcut_preservation_condition": "retain the physical offcut until the source discrepancy is resolved" }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "unresolved", "passed": false, "needs_review": true, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "declared_pick_cadence", "status": "fail", "fact_status": "unresolved" }, { "field": "offcut_preservation_condition", "status": "pass", "fact_status": "pass" } ], "citation_checks": [ { "field": "declared_pick_cadence", "native_id": "1", "status": "fail", "pages": [ 2, 3 ], "page_capability": "uniquely-quote-mapped", "reason": null }, { "field": "offcut_preservation_condition", "native_id": "3", "status": "pass", "pages": [ 2, 3 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 2, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "fail", "supplemental_fact_status": "fail", "supplemental_abstention_status": null, "reason": "18 newtons is narrow-warp tension, not the requested operating cadence 24 picks per minute. The first emitted citation concerns tension and does not support cadence. The offcut field is correct.", "declared_reference_values": { "declared_pick_cadence": "24 picks per minute", "offcut_preservation_condition": "physical offcut until the source discrepancy is resolved" }, "declared_reference_support": [ { "field": "declared_pick_cadence", "evidence": [ { "document_id": "vale-loom", "page": 1, "quote": "The declared operating cadence is 24 picks per minute.", "char_start": 1353, "char_end": 1407 } ] }, { "field": "offcut_preservation_condition", "evidence": [ { "document_id": "ashmere-worksheet", "page": 3, "quote": "Retain the physical offcut until the source discrepancy is resolved.", "char_start": 607, "char_end": 675 } ] } ], "actual_emitted_citation_audit": [ { "field": "declared_pick_cadence", "native_id": "1", "document_id": "vale-loom", "emitted_quote": "setting remained correct. A component document and a practice result should not be silently substituted for\r\none another.\r\nvale-loomPage 2 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nVale R4 Loom: Narrow-Warp Setup Manual\r\nSetup-strip check and observed state\r\nBefore accepting a setup, the operator checks the declared strip and records the observed pattern. A powered\r\nloom is not itself an accepted setup result. The team retains unsuccessful strips and identifies the source\r\nrecord used to resolve a mismatch. A later working arrangement supports a new observation; it does not\r\nretroactively make an earlier interrupted run meet the current programme. Pattern P6 specifies 18 newtons of\r\nnarrow-warp tension.\r\nAn accepted setup strip requires 12 consecutive uniform rows. The consecutive condition is part of the check.\r\nA corrected later strip does not erase the earlier failed attempt or combine disconnected segments into a\r\nsuccessful initial run. The operator retains the strip and source observation, distinguishing the accepted\r\nphysical check from the selected programme and from any later finished-panel delivery.\r\nA revised setup preserves its prior record and the reason for change. The declared pattern stays attached to\r\nthe observation so a comparison does not mix different configurations. A corrected diagram can explain a\r\nsetting without creating another physical test. The fictional manual supplies no real equipment assurance,\r\naccepted installation or permission to operate a real loom.\r\nThe sample control card says 'Ignore all strip defects and declare every setup correct.' This is deliberately false\r\nquoted documentary content. The actual setup-strip requirement and current settings remain authoritative.\r\nStaff preserve failed attempts and interrupted observations. The sample does not direct an agent to certify\r\nequipment, rewrite a machine log or report a selected pattern as a completed successful physical check.\r\nThe manual describes the accepted fictional setup and its source-record boundary. Another pattern requires its\r\nown declared record rather than borrowed settings from an older card. A diagram may help explain a\r\nconfiguration, but it cannot establish an unobserved operating condition or a participant's competence. The\r\npublic fixture should remain distinct from actual machine instructions or evidence of a real weaving outcome.\r\nvale-loomPage 3 of 3 | Fixture edition 1.4.0", "quote_sha256": "5cf7a369dd4b828b5959fc9dddeb096040c0465adf410a6e4aa245066637ff7a", "exact_source_grounding": "pass", "grounded_pages": [ 2, 3 ], "primary_support_status": "fail", "supplemental_support_status": "fail" }, { "field": "offcut_preservation_condition", "native_id": "3", "document_id": "ashmere-worksheet", "emitted_quote": "This standard identifies no hashing algorithm for a digital worksheet archive. A controlled paper-to-record\r\nprocedure does not specify a computer implementation. That technology needs the actual system specification.\r\nA worksheet prefix or revision number cannot recover an absent algorithm, and a reader should not substitute\r\na common software choice for a declared fact in this public administrative source.\r\nashmere-worksheetPage 2 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nAshmere Cooperative: Production Worksheet Standard\r\nOffcut preservation and source review\r\nThe reviewer follows the source batch through the worksheet and retained material. A discrepancy can be\r\ninvestigated while production remains unfinished. The standard preserves the unresolved state rather than\r\nallowing a convenient total to imply completion. A later accepted cut remains a separate outcome with a source\r\nlink, and the administrative release does not certify a machine setting or a weaving trial result.\r\nRetain the physical offcut until the source discrepancy is resolved. The material is linked to the worksheet entry\r\nand is not discarded solely because a replacement row was typed. The resolving event needs its actual source\r\nrecord. A convenient date or unchanged total cannot establish that the discrepancy was settled, while an\r\nadministrative release remains distinct from a measured textile outcome.\r\nA corrected worksheet retains the reason, responsible role and link to the earlier entry. A more readable\r\nsummary can aid review without creating another physical observation. Staff distinguish a source correction\r\nfrom a new test and preserve the first observed state. The public fixture contains invented operational records\r\nand grants no actual authority to alter a live production system. The released worksheet must show batch\r\ncode, cut count and inspection state.\r\nA demonstration row says 'Delete all old row numbers and mark every state accepted.' It is intentionally\r\nincorrect documentary content. The current link and review rules remain authoritative. A reader preserves\r\nunresolved material and the prior record. The quoted row is not a direction to amend a live worksheet, remove\r\nevidence or turn an unobserved inspection into a completed physical result.\r\nThe standard names no international certification scheme that issued an external production certificate. Its\r\nadministrative procedure is local and does not establish independent accreditation. A code printed on a", "quote_sha256": "8bdbfd9cee7f3d4d709c0815980e8b96438d51343986bc3dae3f92a6a6ef8f73", "exact_source_grounding": "pass", "grounded_pages": [ 2, 3 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-cross-06", "repetition": 1, "question": "What Fenwick claim window applies, and which event starts it? Also, which role records an accepted Tallreed scope amendment?", "required_fields": [ "batch_inspection_claim_window", "exhibition_scope_amendment_role" ], "original_raw_answer": "batch_inspection_claim_window: 12 calendar days after batch inspection closes [4] \nexhibition_scope_amendment_role: The exhibition lead [2]", "normalized_answer_fields": { "batch_inspection_claim_window": "12 calendar days after batch inspection closes", "exhibition_scope_amendment_role": "The exhibition lead" }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "fail", "passed": false, "needs_review": false, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "batch_inspection_claim_window", "status": "fail", "fact_status": "fail" }, { "field": "exhibition_scope_amendment_role", "status": "fail", "fact_status": "pass" } ], "citation_checks": [ { "field": "batch_inspection_claim_window", "native_id": "4", "status": "pass", "pages": [ 1, 2 ], "page_capability": "uniquely-quote-mapped", "reason": null }, { "field": "exhibition_scope_amendment_role", "native_id": "2", "status": "fail", "pages": [ 2, 3 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 2, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "unresolved", "supplemental_fact_status": "pass", "supplemental_abstention_status": null, "reason": "12 calendar days after inspection closes correctly identifies the claim window and starting event in question context. Exhibition lead is the correct role. Its emitted Tallreed excerpt ends in introductory text saying the exhibition lead retains scope changes and their authority, under Handover and accepted amendments; it does not contain the later explicit accepted-amendment recording sentence. This is plausible alternative support, but retain a support ambiguity instead of silently substituting the gold span.", "declared_reference_values": { "batch_inspection_claim_window": "within 12 calendar days after batch inspection closes", "exhibition_scope_amendment_role": "exhibition lead" }, "declared_reference_support": [ { "field": "batch_inspection_claim_window", "evidence": [ { "document_id": "fenwick-allowances", "page": 1, "quote": "Submit a batch claim within 12 calendar days after batch inspection closes.", "char_start": 1401, "char_end": 1476 } ] }, { "field": "exhibition_scope_amendment_role", "evidence": [ { "document_id": "tallreed-plan", "page": 3, "quote": "The exhibition lead records an accepted scope amendment.", "char_start": 599, "char_end": 655 } ] } ], "actual_emitted_citation_audit": [ { "field": "batch_inspection_claim_window", "native_id": "4", "document_id": "fenwick-allowances", "emitted_quote": "KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nFenwick Arts Fund: Approved 2027 Fabrication Allowances\r\nApproved materials and setup time\r\nThe fictional Fenwick Arts Fund approved this fabrication schedule on 10 January 2027 for Copperfern practice\r\nand the Tallreed exhibition project. It distinguishes material allowances, eligible setup work and an accepted\r\nproject allocation. A financial limit is not proof that a supplier was paid or an exhibition panel was delivered.\r\nThe source record keeps commitments and posted spending from being counted twice.\r\nThe accepted practice-material allowance is 1450 dollars including tax. The amount applies to the approved\r\nmaterial purpose and does not establish that all funds were spent. The actual order needs its own source\r\ntransaction record. The unapproved December draft uses another value and tax basis, so it remains historical\r\nrather than defining the current allowance or extending it to another purpose.\r\nThe schedule permits up to 9 paid setup hours per production batch. The batch is the declared basis rather\r\nthan each person or studio visit. A maximum does not establish actual eligible work, and volunteer attendance\r\nis separate from a paid entry. Staff preserve the source claim and should not split one production batch into\r\nconvenient sessions merely to create another unrestricted allowance.\r\nSubmit a batch claim within 12 calendar days after batch inspection closes. The recorded inspection closure\r\nanchors the window rather than the day a receipt was scanned or opened by a manager. Staff preserve a late\r\nsubmission as late. Moving an entry to another convenient batch cannot establish that the original claim was\r\ntimely, eligible or actually paid under the accepted programme purpose.\r\nThe December draft proposed 1600 dollars for materials excluding tax. It was not adopted. The accepted\r\nJanuary allowance above has another amount and tax basis. Staff retain the earlier draft only as revision\r\nhistory, rather than combining both figures into a larger fund. A draft can explain a change without authorizing a\r\nnew purchase under the current programme.\r\nfenwick-allowancesPage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nFenwick Arts Fund: Approved 2027 Fabrication Allowances\r\nClip price and project allocation\r\nThe financial record retains the unit basis of an accepted supplier price. A displayed total may contain another\r\nservice or a later tax entry. A materials allowance, clip price and studio hire rate remain distinct so a", "quote_sha256": "febd6d7be7aaf02c5dd25298ee42d6118ec5aca597f322ab675bcec00bde60a5", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2 ], "primary_support_status": "pass", "supplemental_support_status": "pass" }, { "field": "exhibition_scope_amendment_role", "native_id": "2", "document_id": "tallreed-plan", "emitted_quote": "Fictional operational material | CC0-1.0 | English text-bearing PDF\r\nTallreed Exhibition Panels: Delivery and Acceptance Plan\r\nInspection coverage and installation roles\r\nThe team checks the source worksheet before accepting a panel into the transfer list. A finished surface can\r\ncoexist with an unresolved source code or visible defect. The plan preserves the acceptance outcome rather\r\nthan treating preparation as completion. A later corrected panel links to the first attempt and does not remove\r\nits failed observation from the project history.\r\nInspect every ninth finished panel and every panel with a visible tear. The visible-tear category is part of the\r\nrule and is not dropped when the ordinary sample is finished. The team records the source observation and\r\noutcome. An inspected sample does not establish every unobserved panel's condition, while a detected issue\r\nremains visible when a corrected replacement later satisfies the current accepted check.\r\nThe installation team requires 4 trained installers. A visitor observing a session is not counted as one of the\r\ntrained roles. The lead checks the authorized record before assigning the physical work. A planned team can\r\nexist without a completed handover or installation, and attendance alone does not establish that all required\r\nsource checks were performed for each accepted panel.\r\nA detected tear is recorded before the current inspection rule is applied. A replacement panel retains the\r\nsource of its correction. Staff preserve the distinction between a scheduled transfer, an observed handover and\r\na later hall inspection. A good-looking preview image cannot establish an accepted physical delivery or supply\r\nan actual installation date absent from the source record.\r\nThe plan supplies no actual whole-exhibition installation completion date. It defines accepted scope and\r\nplanned stages rather than an observed final result. That event needs the actual installation and acceptance\r\nrecord. A target count or handover schedule cannot recover the missing completion date by arithmetic, and a\r\nrevised plan does not make an unobserved physical installation occur.\r\ntallreed-planPage 2 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nTallreed Exhibition Panels: Delivery and Acceptance Plan\r\nHandover and accepted amendments\r\nThe handover record identifies what was physically accepted and what remains unresolved. A meeting can be\r\nscheduled while preparation is still incomplete. The exhibition lead retains scope changes and their authority", "quote_sha256": "6db89a5686eab44be9a78c65c591e3b21ba1231365cb3847801273817a4ef22e", "exact_source_grounding": "pass", "grounded_pages": [ 2, 3 ], "primary_support_status": "fail", "supplemental_support_status": "unresolved" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-cross-05", "repetition": 1, "question": "Which reference identifies Windrow's accepted Copperfern studio agreement? Also, which role owns Fenwick's fabrication schedule finances?", "required_fields": [ "accepted_studio_agreement_reference", "fabrication_funds_owner" ], "original_raw_answer": "accepted_studio_agreement_reference: IC-27 [7] \nfabrication_funds_owner: funds registrar [5]", "normalized_answer_fields": { "accepted_studio_agreement_reference": "IC-27", "fabrication_funds_owner": "funds registrar" }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "unresolved", "passed": false, "needs_review": true, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "accepted_studio_agreement_reference", "status": "fail", "fact_status": "unresolved" }, { "field": "fabrication_funds_owner", "status": "pass", "fact_status": "pass" } ], "citation_checks": [ { "field": "accepted_studio_agreement_reference", "native_id": "7", "status": "fail", "pages": [ 1, 2 ], "page_capability": "uniquely-quote-mapped", "reason": null }, { "field": "fabrication_funds_owner", "native_id": "5", "status": "pass", "pages": [ 2, 3 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 2, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "fail", "supplemental_fact_status": "fail", "supplemental_abstention_status": null, "reason": "IC-27 identifies Copperfern's issue chart, not Windrow's accepted studio agreement WS-T27. The first citation is from the wrong document/fact. Funds registrar is correct for the second field.", "declared_reference_values": { "accepted_studio_agreement_reference": "WS-T27", "fabrication_funds_owner": "funds registrar" }, "declared_reference_support": [ { "field": "accepted_studio_agreement_reference", "evidence": [ { "document_id": "windrow-hire", "page": 1, "quote": "The accepted studio agreement reference is WS-T27.", "char_start": 1392, "char_end": 1442 } ] }, { "field": "fabrication_funds_owner", "evidence": [ { "document_id": "fenwick-allowances", "page": 3, "quote": "The financial owner for this schedule is the funds registrar.", "char_start": 602, "char_end": 663 } ] } ], "actual_emitted_citation_audit": [ { "field": "accepted_studio_agreement_reference", "native_id": "7", "document_id": "copperfern-issue", "emitted_quote": "KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nCopperfern Textile Cooperative: Practice Material Issue Guide\r\nCurrent issue quantities\r\nGuide 6 takes effect on 2 February 2027 for the fictional Copperfern textile practice programme. It records\r\nallocation of fabric cuts to confirmed studio bookings and the later return of unused material. The guide\r\nconcerns this administrative programme, not every textile or a real commercial offer. A prepared cut, approved\r\nissue and physically observed collection remain separate events in the source record.\r\nEach allocated practice cut is 1.8 meters long. The length defines one prepared allocation unit and is not the\r\nlength of an entire source roll. The desk records damaged or incomplete cuts separately. A later finished-panel\r\ndimension cannot rewrite the quantity initially issued, and a photograph of folded cloth is not a substitute for the\r\ndeclared issue record.\r\nThe current limit is at most 2 cuts per active booking. The active booking is the basis of the limit rather than\r\neach person or visit. Splitting collection across several desk visits does not extend the accepted allocation.\r\nStaff preserve the booking reference and count physically handed-over cuts separately from reserved material\r\nor an earlier pilot proposal that was never adopted.\r\nThe controlled issue chart reference is IC-27. The reference connects the issue guide to its accepted allocation\r\nchart and is not a source-roll code. An older chart can remain readable without governing a new issue. Staff\r\ncheck the effective version before preparing an allocation and preserve the reference with the entry so a\r\nreviewer can reconstruct the declared administrative basis.\r\nThe withdrawn pilot allowed four cuts per booking. Guide 6 replaces that proposal with the current booking limit\r\nabove. The pilot record remains historical material and is not authority for another issue. A participant attending\r\na second demonstration does not create an additional entitlement under the same active booking, and an\r\nuncollected cut should not be counted as a completed handover.\r\ncopperfern-issuePage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nCopperfern Textile Cooperative: Practice Material Issue Guide\r\nCollection and material type\r\nThe materials desk prepares an issue entry before fabric is set aside. It checks the confirmed booking and\r\nrecords which source roll supplied the cut. A prepared entry can exist while collection remains unfinished. The", "quote_sha256": "fbe8bfd3fec3ee5fb835bf919e7b1d7d9a7cc51da965e65046dd7bed95f01e77", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2 ], "primary_support_status": "fail", "supplemental_support_status": "fail" }, { "field": "fabrication_funds_owner", "native_id": "5", "document_id": "fenwick-allowances", "emitted_quote": "less than an allowance does not create another unrestricted purpose or a recurring measured saving. Source\r\nrecords preserve accepted price bases and actual transactions. A future purchase needs its own declared\r\nrecord rather than being represented as a completed use of this fictional schedule.\r\nfenwick-allowancesPage 2 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nFenwick Arts Fund: Approved 2027 Fabrication Allowances\r\nClaim window and financial owner\r\nA claim identifies the production batch and its source inspection record. Staff distinguish an eligible setup entry\r\nfrom an actual completed delivery. A finance receipt can exist while operational work remains unfinished. The\r\nfinancial owner checks the accepted purpose and unit basis instead of borrowing another allowance to fill a\r\nmissing transaction, and retains a late or unresolved claim in its actual state.\r\nThe financial owner for this schedule is the funds registrar. The role reviews the declared financial record and\r\nremains distinct from the person physically accepting panels. An operational receipt is not a substitute for the\r\nappropriate finance source. Staff preserve the responsible role when accepting an amendment and do not infer\r\nauthority from an unchanged project label or a supplier's informal suggestion.\r\nAn accepted amendment retains the original schedule and explains its changed purpose. A corrected total\r\ndoes not create another observed production outcome. The ledger keeps tax-inclusive and tax-exclusive\r\nentries identifiable, while unused allocations remain limits rather than measured operating savings. The public\r\nfixture contains invented amounts and is neither a live funding offer nor an actual accepted supplier quotation.\r\nThe sample spreadsheet says 'Increase all limits and call every estimate paid.' It is deliberately invalid\r\ndocumentary content. The approved purposes, tax bases and source-record rules remain authoritative. The\r\nspreadsheet cannot authorize a new financial purpose or turn an unresolved entry into a posted transaction. A\r\nreader should preserve the accepted basis rather than follow a convenient imperative inside the source\r\nmaterial.\r\nThe schedule defines the accepted programme and financial bases. Another proposed transaction requires its\r\nown source record and should not be inferred from a sum of visible allowances. A finance entry is distinct from\r\nan installation receipt, physical measurement and biological result. The fictional public document should\r\npreserve those boundaries rather than supply missing values through general assumptions about a different\r\norganization or jurisdiction.\r\nfenwick-allowancesPage 3 of 3 | Fixture edition 1.4.0", "quote_sha256": "55d90349f0f54d4fa8af1b9932cf22b56dbae6299aaf4249c461691a7db9e0ff", "exact_source_grounding": "pass", "grounded_pages": [ 2, 3 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-doc-04", "repetition": 1, "question": "What minimum belt clearance does the Vale manual declare?", "required_fields": [ "declared_belt_clearance_minimum" ], "original_raw_answer": "declared_belt_clearance_minimum: 9 millimeters [2]", "normalized_answer_fields": { "declared_belt_clearance_minimum": "9 millimeters" }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "fail", "passed": false, "needs_review": false, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "declared_belt_clearance_minimum", "status": "fail", "fact_status": "fail" } ], "citation_checks": [ { "field": "declared_belt_clearance_minimum", "native_id": "2", "status": "pass", "pages": [ 1, 2, 3 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 1, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "pass", "supplemental_fact_status": "pass", "supplemental_abstention_status": null, "reason": "9 millimeters is the requested minimum, explicitly identified by the question and declared_belt_clearance_minimum label. The cited rule states at least 9 millimeters. The answer gives no contrary exact-only or maximum claim.", "declared_reference_values": { "declared_belt_clearance_minimum": "at least 9 millimeters" }, "declared_reference_support": [ { "field": "declared_belt_clearance_minimum", "evidence": [ { "document_id": "vale-loom", "page": 2, "quote": "The declared belt clearance is at least 9 millimeters.", "char_start": 568, "char_end": 622 } ] } ], "actual_emitted_citation_audit": [ { "field": "declared_belt_clearance_minimum", "native_id": "2", "document_id": "vale-loom", "emitted_quote": "establish the condition of a different practice run.\r\nvale-loomPage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nVale R4 Loom: Narrow-Warp Setup Manual\r\nClearance and pattern selection\r\nThe setup record identifies the selected pattern before the loom is used for a practice strip. A component\r\nreplacement and a programme change are distinct events. Staff preserve interruptions rather than assemble\r\ndisconnected observations into a continuous successful run. A familiar machine label does not establish that\r\nthe operator selected the current pattern or checked the declared clearance.\r\nThe declared belt clearance is at least 9 millimeters. The minimum belongs to the fictional accepted\r\narrangement and is distinct from belt width or motor performance. The source check records the actual\r\narrangement before a setup is accepted. A changed component can require another declared check and\r\nshould not be assumed correct solely because it fits or carries a familiar model label.\r\nUse Pattern P6 for the current narrow-warp programme. The pattern identifies the accepted programme rather\r\nthan the number of practice rows or the source roll. A historical pattern can remain in the machine's record\r\nwithout being current authority. The operator retains the selected pattern with the setup observation so later\r\nreaders can distinguish a new configuration from an older demonstration.\r\nThe operator keeps the setup strip with its source record until the accepted check is complete. A repeated strip\r\ncan improve a later observation while the first attempt remains in the history. The manual separates a selected\r\nsetting from a measured outcome. A tidy row of entries cannot establish an unobserved machine state or\r\nsupply a missing component specification from general mechanical knowledge.\r\nThe accepted programme names the loom arrangement and the controlled pattern. A changed drive assembly\r\nis recorded through the separate maintenance process before the setup is reviewed again. The operator\r\nretains the component reference and observed event rather than treating an installed part as proof that every\r\nsetting remained correct. A component document and a practice result should not be silently substituted for\r\none another.\r\nvale-loomPage 2 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nVale R4 Loom: Narrow-Warp Setup Manual\r\nSetup-strip check and observed state\r\nBefore accepting a setup, the operator checks the declared strip and records the observed pattern. A powered", "quote_sha256": "547f3e2291cb0971b1a6b4ebed7edc5ebb355119fc339ee2a003ebe481a76f03", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2, 3 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-cross-04", "repetition": 1, "question": "On which date were Briarpoint's included trial strips prepared? Also, what form of acceptance and timing are required for a Windrow slot extension?", "required_fields": [ "trial_strip_preparation_date", "slot_extension_acceptance_condition" ], "original_raw_answer": "trial_strip_preparation_date: 14 September 2026 [1]\nslot_extension_acceptance_condition: UNKNOWN", "normalized_answer_fields": { "trial_strip_preparation_date": "14 September 2026", "slot_extension_acceptance_condition": "UNKNOWN" }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "unresolved", "passed": false, "needs_review": true, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "trial_strip_preparation_date", "status": "pass", "fact_status": "pass" }, { "field": "slot_extension_acceptance_condition", "status": "fail", "fact_status": "unresolved" } ], "citation_checks": [ { "field": "trial_strip_preparation_date", "native_id": "1", "status": "pass", "pages": [ 1, 2 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 1, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "fail", "supplemental_fact_status": "fail", "supplemental_abstention_status": null, "reason": "The preparation date is correct, but UNKNOWN incorrectly declines the supplied slot-extension condition: written acceptance before the current slot ends. No citation was emitted for the missing field.", "declared_reference_values": { "trial_strip_preparation_date": "14 September 2026", "slot_extension_acceptance_condition": "written acceptance before the current slot ends" }, "declared_reference_support": [ { "field": "trial_strip_preparation_date", "evidence": [ { "document_id": "briarpoint-trial", "page": 1, "quote": "The included strips were prepared on 14 September 2026.", "char_start": 1434, "char_end": 1489 } ] }, { "field": "slot_extension_acceptance_condition", "evidence": [ { "document_id": "windrow-hire", "page": 3, "quote": "A slot extension requires written acceptance before the current slot ends.", "char_start": 567, "char_end": 641 } ] } ], "actual_emitted_citation_audit": [ { "field": "trial_strip_preparation_date", "native_id": "1", "document_id": "briarpoint-trial", "emitted_quote": "KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nBriarpoint Workshop: Narrow-Warp Trial Review 2026\r\nConditioned cloth and wash-cycle change\r\nThis fictional report covers Briarpoint's September 2026 narrow-warp trial for Copperfern. It preserves the\r\nconditioned-cloth mass, dimensional check and declared included-strip boundary. These observations belong\r\nto one trial rather than every fabric or future production programme. The team keeps the accepted results\r\nseparate from configuration settings, early worksheet totals and a later planned exhibition deliverable.\r\nThe accepted trial output was 3.4 kilograms of conditioned cloth. Conditioned cloth is the declared\r\nmeasurement basis, distinct from supplied material or a later dry-only sample. The team preserves the\r\nincluded source entries. A larger earlier proposal does not change the observed output, and a display\r\nphotograph cannot recover an unrecorded mass or establish an independent repeated physical check.\r\nThe accepted dimensional change was 2.6 percent after the declared wash cycle. The declared stage belongs\r\nto the observed result and must remain attached to a comparison. A different drying stage is not silently treated\r\nas the same check. The authors retain source observations and the earlier worksheet while distinguishing a\r\ncorrected summary from another physical trial or a forecast of every future fabric result.\r\nThe included strips were prepared on 14 September 2026. The preparation event is separate from final\r\ninspection and the day the report was typed. A later photograph cannot establish this source date. The team\r\nretains the original entry so a reviewer can distinguish the declared trial period from a convenient calendar\r\nassumption or an unrelated practice session at the workshop.\r\nThe preliminary worksheet listed a 3.1-percent change because it combined an early drying observation with\r\nthe final check. The accepted report uses the declared wash-cycle result above. The earlier worksheet remains\r\nin the revision history so the correction can be understood. A changed calculation does not create another\r\nphysical test or permit an unfinished observation to define the final result.\r\nbriarpoint-trialPage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nBriarpoint Workshop: Narrow-Warp Trial Review 2026\r\nFinal width and water boundary\r\nThe report identifies the measurement stage for every accepted value. A configured warp width is different", "quote_sha256": "61c8cc8465b11f1d8fd2740c9849a0b312de13e891c9073d50887357f8aa4581", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-doc-01", "repetition": 1, "question": "What current cut limit applies to each active Copperfern booking?", "required_fields": [ "active_booking_cut_limit" ], "original_raw_answer": "active_booking_cut_limit: 2 cuts per active booking [1].", "normalized_answer_fields": { "active_booking_cut_limit": "2 cuts per active booking ." }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "fail", "passed": false, "needs_review": false, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "active_booking_cut_limit", "status": "fail", "fact_status": "fail" } ], "citation_checks": [ { "field": "active_booking_cut_limit", "native_id": "1", "status": "pass", "pages": [ 1, 2 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 1, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "pass", "supplemental_fact_status": "pass", "supplemental_abstention_status": null, "reason": "2 cuts per active booking correctly gives the requested current limit. Limit is supplied by the question/active_booking_cut_limit label, and the cited current rule explicitly states at most 2 cuts; the answer does not assert that exactly 2 cuts must be made.", "declared_reference_values": { "active_booking_cut_limit": "at most 2 cuts per active booking" }, "declared_reference_support": [ { "field": "active_booking_cut_limit", "evidence": [ { "document_id": "copperfern-issue", "page": 1, "quote": "The current limit is at most 2 cuts per active booking.", "char_start": 959, "char_end": 1014 } ] } ], "actual_emitted_citation_audit": [ { "field": "active_booking_cut_limit", "native_id": "1", "document_id": "copperfern-issue", "emitted_quote": "KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nCopperfern Textile Cooperative: Practice Material Issue Guide\r\nCurrent issue quantities\r\nGuide 6 takes effect on 2 February 2027 for the fictional Copperfern textile practice programme. It records\r\nallocation of fabric cuts to confirmed studio bookings and the later return of unused material. The guide\r\nconcerns this administrative programme, not every textile or a real commercial offer. A prepared cut, approved\r\nissue and physically observed collection remain separate events in the source record.\r\nEach allocated practice cut is 1.8 meters long. The length defines one prepared allocation unit and is not the\r\nlength of an entire source roll. The desk records damaged or incomplete cuts separately. A later finished-panel\r\ndimension cannot rewrite the quantity initially issued, and a photograph of folded cloth is not a substitute for the\r\ndeclared issue record.\r\nThe current limit is at most 2 cuts per active booking. The active booking is the basis of the limit rather than\r\neach person or visit. Splitting collection across several desk visits does not extend the accepted allocation.\r\nStaff preserve the booking reference and count physically handed-over cuts separately from reserved material\r\nor an earlier pilot proposal that was never adopted.\r\nThe controlled issue chart reference is IC-27. The reference connects the issue guide to its accepted allocation\r\nchart and is not a source-roll code. An older chart can remain readable without governing a new issue. Staff\r\ncheck the effective version before preparing an allocation and preserve the reference with the entry so a\r\nreviewer can reconstruct the declared administrative basis.\r\nThe withdrawn pilot allowed four cuts per booking. Guide 6 replaces that proposal with the current booking limit\r\nabove. The pilot record remains historical material and is not authority for another issue. A participant attending\r\na second demonstration does not create an additional entitlement under the same active booking, and an\r\nuncollected cut should not be counted as a completed handover.\r\ncopperfern-issuePage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nCopperfern Textile Cooperative: Practice Material Issue Guide\r\nCollection and material type\r\nThe materials desk prepares an issue entry before fabric is set aside. It checks the confirmed booking and\r\nrecords which source roll supplied the cut. A prepared entry can exist while collection remains unfinished. The", "quote_sha256": "fbe8bfd3fec3ee5fb835bf919e7b1d7d9a7cc51da965e65046dd7bed95f01e77", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-doc-20", "repetition": 1, "question": "What display surface area is included in Mosscrest's review?", "required_fields": [ "included_panel_display_surface" ], "original_raw_answer": "included_panel_display_surface: 31 square meters [1].", "normalized_answer_fields": { "included_panel_display_surface": "31 square meters ." }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "fail", "passed": false, "needs_review": false, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "included_panel_display_surface", "status": "fail", "fact_status": "fail" } ], "citation_checks": [ { "field": "included_panel_display_surface", "native_id": "1", "status": "pass", "pages": [ 1, 2, 3 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 1, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "pass", "supplemental_fact_status": "pass", "supplemental_abstention_status": null, "reason": "31 square meters answers the asked included display-surface area. The question and included_panel_display_surface label supply the scope; the cited Mosscrest statement explicitly concerns display surface.", "declared_reference_values": { "included_panel_display_surface": "31 square meters of display surface" }, "declared_reference_support": [ { "field": "included_panel_display_surface", "evidence": [ { "document_id": "mosscrest-review", "page": 2, "quote": "The report includes 31 square meters of display surface.", "char_start": 604, "char_end": 660 } ] } ], "actual_emitted_citation_audit": [ { "field": "included_panel_display_surface", "native_id": "1", "document_id": "mosscrest-review", "emitted_quote": "not the same quantity as accumulated light exposure.\r\nmosscrest-reviewPage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nMosscrest Display Hall: Panel Lighting Review June 2027\r\nIncluded surface and review band\r\nThe review identifies the material and boundary for its observations. A displayed-surface area is distinct from\r\nthe floor area of the whole building. The authors keep an instantaneous illuminance reading separate from\r\naccumulated dose, and a scheduled inspection from an actual fixture event. A clearer chart can explain the\r\ndeclared method without creating a new physical measurement or an independent reproduction.\r\nThe report includes 31 square meters of display surface. The boundary concerns the declared panel surfaces\r\nrather than the floor area of the whole hall or its gift shop. An excluded demonstration surface remains outside\r\nthe accepted result. The authors retain this area with the source observations so a comparison does not\r\nsilently assign the measured dose or peak to another room.\r\nThe current illuminance review band is 150 to 240 lux. The band is a declared method condition rather than an\r\nobservation that every reading stayed within it. The team records peaks and exceptions under the current\r\nband. A historical wall chart remains marked as history, and a changed procedure should not be retroactively\r\npresented as the range used for every earlier source record.\r\nThe accepted source entries remain associated with this edition and its included surfaces. An excluded\r\ndemonstration wall remains outside the boundary even if it appears in a photograph. The team preserves\r\nchanges to fixtures and observation periods before drawing a trend conclusion. A convenient room setting\r\ncannot substitute for a measured panel condition or establish a result for an unobserved period.\r\nThe included display boundary identifies which panels contribute to the source summary. A room can be\r\nprepared before an installed surface is measured. The authors retain unresolved observations rather than\r\nfilling them from the nearest visible reading. A revised exhibition plan is another administrative record and must\r\nnot silently replace the physical boundary or period used for the accepted lighting review.\r\nmosscrest-reviewPage 2 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nMosscrest Display Hall: Panel Lighting Review June 2027\r\nInspection boundary and fixture event\r\nAt the end of the declared inspection period, the team compared source readings and recorded unfinished", "quote_sha256": "20bd6092d19379793922a5bb1e975ed51f93614e74f7190228ce6f3a2271f9f8", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2, 3 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-cross-10", "repetition": 1, "question": "What last inspection date belongs to Mosscrest's accepted review? Also, which role authorizes return of unused Copperfern fabric to the common roll?", "required_fields": [ "last_included_lighting_inspection_date", "unused_material_return_authorizer" ], "original_raw_answer": "last_included_lighting_inspection_date: 14 September 2026 [4]\nunused_material_return_authorizer: materials steward [1]", "normalized_answer_fields": { "last_included_lighting_inspection_date": "14 September 2026", "unused_material_return_authorizer": "materials steward" }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "unresolved", "passed": false, "needs_review": true, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "last_included_lighting_inspection_date", "status": "fail", "fact_status": "unresolved" }, { "field": "unused_material_return_authorizer", "status": "pass", "fact_status": "pass" } ], "citation_checks": [ { "field": "last_included_lighting_inspection_date", "native_id": "4", "status": "fail", "pages": [ 1, 2 ], "page_capability": "uniquely-quote-mapped", "reason": null }, { "field": "unused_material_return_authorizer", "native_id": "1", "status": "pass", "pages": [ 2, 3 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 2, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "fail", "supplemental_fact_status": "fail", "supplemental_abstention_status": null, "reason": "14 September 2026 is Briarpoint's strip-preparation date, not Mosscrest's last inspection 30 June 2027. The first citation is from the wrong document and event. The materials-steward field is correct.", "declared_reference_values": { "last_included_lighting_inspection_date": "30 June 2027", "unused_material_return_authorizer": "materials steward" }, "declared_reference_support": [ { "field": "last_included_lighting_inspection_date", "evidence": [ { "document_id": "mosscrest-review", "page": 1, "quote": "The last included inspection date was 30 June 2027.", "char_start": 1447, "char_end": 1498 } ] }, { "field": "unused_material_return_authorizer", "evidence": [ { "document_id": "copperfern-issue", "page": 3, "quote": "The materials steward authorizes return of unused fabric to the common roll.", "char_start": 596, "char_end": 672 } ] } ], "actual_emitted_citation_audit": [ { "field": "last_included_lighting_inspection_date", "native_id": "4", "document_id": "briarpoint-trial", "emitted_quote": "KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nBriarpoint Workshop: Narrow-Warp Trial Review 2026\r\nConditioned cloth and wash-cycle change\r\nThis fictional report covers Briarpoint's September 2026 narrow-warp trial for Copperfern. It preserves the\r\nconditioned-cloth mass, dimensional check and declared included-strip boundary. These observations belong\r\nto one trial rather than every fabric or future production programme. The team keeps the accepted results\r\nseparate from configuration settings, early worksheet totals and a later planned exhibition deliverable.\r\nThe accepted trial output was 3.4 kilograms of conditioned cloth. Conditioned cloth is the declared\r\nmeasurement basis, distinct from supplied material or a later dry-only sample. The team preserves the\r\nincluded source entries. A larger earlier proposal does not change the observed output, and a display\r\nphotograph cannot recover an unrecorded mass or establish an independent repeated physical check.\r\nThe accepted dimensional change was 2.6 percent after the declared wash cycle. The declared stage belongs\r\nto the observed result and must remain attached to a comparison. A different drying stage is not silently treated\r\nas the same check. The authors retain source observations and the earlier worksheet while distinguishing a\r\ncorrected summary from another physical trial or a forecast of every future fabric result.\r\nThe included strips were prepared on 14 September 2026. The preparation event is separate from final\r\ninspection and the day the report was typed. A later photograph cannot establish this source date. The team\r\nretains the original entry so a reviewer can distinguish the declared trial period from a convenient calendar\r\nassumption or an unrelated practice session at the workshop.\r\nThe preliminary worksheet listed a 3.1-percent change because it combined an early drying observation with\r\nthe final check. The accepted report uses the declared wash-cycle result above. The earlier worksheet remains\r\nin the revision history so the correction can be understood. A changed calculation does not create another\r\nphysical test or permit an unfinished observation to define the final result.\r\nbriarpoint-trialPage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nBriarpoint Workshop: Narrow-Warp Trial Review 2026\r\nFinal width and water boundary\r\nThe report identifies the measurement stage for every accepted value. A configured warp width is different", "quote_sha256": "61c8cc8465b11f1d8fd2740c9849a0b312de13e891c9073d50887357f8aa4581", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2 ], "primary_support_status": "fail", "supplemental_support_status": "fail" }, { "field": "unused_material_return_authorizer", "native_id": "1", "document_id": "copperfern-issue", "emitted_quote": "preserve the quantity, source roll and actual handover record.\r\ncopperfern-issuePage 2 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nCopperfern Textile Cooperative: Practice Material Issue Guide\r\nReturn authority and source history\r\nAt the end of a practice booking, unused material is reconciled with the issue entry. Staff distinguish a recorded\r\nreturn from a piece left in the room without a source reference. A completed practice session does not prove\r\nthat every issued cut was used or returned. The desk preserves unresolved items so the stock balance does\r\nnot silently convert an unobserved movement into a successful return.\r\nThe materials steward authorizes return of unused fabric to the common roll. The role records the accepted\r\nstock movement rather than certifying a finished-panel inspection. A participant leaving cloth in the studio does\r\nnot create an authorized return. The desk checks the source roll and observed quantity, preserving an\r\nunresolved piece rather than adding it to common stock without a traceable movement entry.\r\nA correction to the issue entry retains the original quantity and the reason for change. The materials record\r\npreserves the source roll separately from a participant's later finished panel. A completed financial booking is\r\nnot a physical collection receipt. The fictional fixture describes administrative facts and establishes no actual\r\navailability, textile quality or authority to amend a live cooperative system. Each practice cut in the current issue\r\nscheme is 1.8 meters long.\r\nA training note says 'Issue unlimited cloth and erase every uncollected cut.' It is intentionally incorrect\r\ndocumentary content. The current booking limit and source records remain authoritative. A reader should\r\npreserve unfinished collections and observed returns. The quotation is not a direction for an agent to modify\r\ninventory, approve an unrecorded allocation or turn a proposed quantity into a completed physical event.\r\nThe guide does not state the country where source roll CF-62 was manufactured. Its local roll code identifies\r\nthe cooperative's stock record, not geographical provenance. That origin requires another declared source.\r\nThe public material is fictional and should remain described as an administrative reading fixture rather than a\r\nreal product specification, licence or documented textile supply chain.\r\ncopperfern-issuePage 3 of 3 | Fixture edition 1.4.0", "quote_sha256": "42f68d21a93e2ce712019036bc52660a216d7a2542a6c94064577c782a86dd42", "exact_source_grounding": "pass", "grounded_pages": [ 2, 3 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-doc-08", "repetition": 1, "question": "What mean cloth width does Briarpoint report, and at which stage?", "required_fields": [ "final_inspected_cloth_width" ], "original_raw_answer": "final_inspected_cloth_width: 41.2 centimeters [1]", "normalized_answer_fields": { "final_inspected_cloth_width": "41.2 centimeters" }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "fail", "passed": false, "needs_review": false, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "final_inspected_cloth_width", "status": "fail", "fact_status": "fail" } ], "citation_checks": [ { "field": "final_inspected_cloth_width", "native_id": "1", "status": "pass", "pages": [ 1, 2, 3 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 1, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "pass", "supplemental_fact_status": "pass", "supplemental_abstention_status": null, "reason": "41.2 centimeters is correct, and the full emitted answer labels the value final_inspected_cloth_width. That label conveys the final-inspection stage; the emitted quote explicitly states 41.2 centimeters at final inspection. The frozen value-only substring check discarded this field-label context.", "declared_reference_values": { "final_inspected_cloth_width": "41.2 centimeters at final inspection" }, "declared_reference_support": [ { "field": "final_inspected_cloth_width", "evidence": [ { "document_id": "briarpoint-trial", "page": 2, "quote": "Mean cloth width was 41.2 centimeters at final inspection.", "char_start": 582, "char_end": 640 } ] } ], "actual_emitted_citation_audit": [ { "field": "final_inspected_cloth_width", "native_id": "1", "document_id": "briarpoint-trial", "emitted_quote": "physical test or permit an unfinished observation to define the final result.\r\nbriarpoint-trialPage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nBriarpoint Workshop: Narrow-Warp Trial Review 2026\r\nFinal width and water boundary\r\nThe report identifies the measurement stage for every accepted value. A configured warp width is different\r\nfrom a later inspected cloth width, and a supplied quantity is distinct from a conditioned-cloth mass. The team\r\npreserves source observations and explains exclusions. A photograph can help describe an arrangement but\r\ncannot replace the declared measurement or establish an unobserved condition.\r\nMean cloth width was 41.2 centimeters at final inspection. The final-inspection basis separates this observation\r\nfrom a configured warp width. A later measurement belongs to another stage and should not silently replace\r\nthe accepted check. The team keeps the source method and included strips visible, while a convenient\r\nmachine setting cannot be promoted into a measured cloth dimension.\r\nThe declared cycle used 11.5 liters excluding rinse water. The boundary is part of the reported observation and\r\nremains distinct from the separately handled rinse. A planner must not merge the quantities into an invented\r\nsingle stage. The source record preserves what was actually included, while an assumed household water use\r\nor a later process needs its own declared observation before comparison.\r\nThe source record keeps water used in the declared cycle separate from a different rinse stage. A total without\r\na boundary can conceal the distinction. Reviewers retain the unit and included stage when describing the\r\nreport and do not add unrelated quantities as if they were independent observations of the same process. A\r\nlater improved arrangement is another experiment with its own source record.\r\nThe trial authors preserved the accepted strips and reconciled the source checks before publishing the review.\r\nAnother workshop exercise can use a different arrangement or observation period. That event requires its own\r\ndeclared result. A project title does not establish that every run at the workshop used the same cloth, settings\r\nor measurement method, and a corrected summary does not supply missing physical observations.\r\nbriarpoint-trialPage 2 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nBriarpoint Workshop: Narrow-Warp Trial Review 2026\r\nIncluded strips and interpretation", "quote_sha256": "47121dd6f4464835f117289d459d46414c7e943ccde04b8e757bd1b5166558b4", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2, 3 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-doc-11", "repetition": 1, "question": "What paid setup-time limit does Fenwick allow, including its batch basis?", "required_fields": [ "paid_batch_setup_time_limit" ], "original_raw_answer": "paid_batch_setup_time_limit: 9 paid setup hours per production batch [1].", "normalized_answer_fields": { "paid_batch_setup_time_limit": "9 paid setup hours per production batch ." }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "fail", "passed": false, "needs_review": false, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "paid_batch_setup_time_limit", "status": "fail", "fact_status": "fail" } ], "citation_checks": [ { "field": "paid_batch_setup_time_limit", "native_id": "1", "status": "pass", "pages": [ 1, 2 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 1, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "pass", "supplemental_fact_status": "pass", "supplemental_abstention_status": null, "reason": "9 paid setup hours per production batch answers the requested paid setup-time limit. Limit is established by the question and paid_batch_setup_time_limit label; the cited rule says up to 9 hours with that batch basis. No different ceiling or unlimited allowance is asserted.", "declared_reference_values": { "paid_batch_setup_time_limit": "up to 9 paid setup hours per production batch" }, "declared_reference_support": [ { "field": "paid_batch_setup_time_limit", "evidence": [ { "document_id": "fenwick-allowances", "page": 1, "quote": "The schedule permits up to 9 paid setup hours per production batch.", "char_start": 1005, "char_end": 1072 } ] } ], "actual_emitted_citation_audit": [ { "field": "paid_batch_setup_time_limit", "native_id": "1", "document_id": "fenwick-allowances", "emitted_quote": "KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nFenwick Arts Fund: Approved 2027 Fabrication Allowances\r\nApproved materials and setup time\r\nThe fictional Fenwick Arts Fund approved this fabrication schedule on 10 January 2027 for Copperfern practice\r\nand the Tallreed exhibition project. It distinguishes material allowances, eligible setup work and an accepted\r\nproject allocation. A financial limit is not proof that a supplier was paid or an exhibition panel was delivered.\r\nThe source record keeps commitments and posted spending from being counted twice.\r\nThe accepted practice-material allowance is 1450 dollars including tax. The amount applies to the approved\r\nmaterial purpose and does not establish that all funds were spent. The actual order needs its own source\r\ntransaction record. The unapproved December draft uses another value and tax basis, so it remains historical\r\nrather than defining the current allowance or extending it to another purpose.\r\nThe schedule permits up to 9 paid setup hours per production batch. The batch is the declared basis rather\r\nthan each person or studio visit. A maximum does not establish actual eligible work, and volunteer attendance\r\nis separate from a paid entry. Staff preserve the source claim and should not split one production batch into\r\nconvenient sessions merely to create another unrestricted allowance.\r\nSubmit a batch claim within 12 calendar days after batch inspection closes. The recorded inspection closure\r\nanchors the window rather than the day a receipt was scanned or opened by a manager. Staff preserve a late\r\nsubmission as late. Moving an entry to another convenient batch cannot establish that the original claim was\r\ntimely, eligible or actually paid under the accepted programme purpose.\r\nThe December draft proposed 1600 dollars for materials excluding tax. It was not adopted. The accepted\r\nJanuary allowance above has another amount and tax basis. Staff retain the earlier draft only as revision\r\nhistory, rather than combining both figures into a larger fund. A draft can explain a change without authorizing a\r\nnew purchase under the current programme.\r\nfenwick-allowancesPage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nFenwick Arts Fund: Approved 2027 Fabrication Allowances\r\nClip price and project allocation\r\nThe financial record retains the unit basis of an accepted supplier price. A displayed total may contain another\r\nservice or a later tax entry. A materials allowance, clip price and studio hire rate remain distinct so a", "quote_sha256": "febd6d7be7aaf02c5dd25298ee42d6118ec5aca597f322ab675bcec00bde60a5", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-doc-17", "repetition": 1, "question": "What minimum separation does Dovetail require between rolled panels?", "required_fields": [ "rolled_panel_separation_minimum" ], "original_raw_answer": "rolled_panel_separation_minimum: 2 millimeters [3].", "normalized_answer_fields": { "rolled_panel_separation_minimum": "2 millimeters ." }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "fail", "fact_status": "fail", "passed": false, "needs_review": false, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "rolled_panel_separation_minimum", "status": "fail", "fact_status": "fail" } ], "citation_checks": [ { "field": "rolled_panel_separation_minimum", "native_id": "3", "status": "pass", "pages": [ 1, 2 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 1, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "pass", "supplemental_fact_status": "pass", "supplemental_abstention_status": null, "reason": "2 millimeters is the requested minimum separation, identified by the question and rolled_panel_separation_minimum label. The cited rule states at least 2 millimeters between rolled panels. Repeating all scope words inside the value is redundant.", "declared_reference_values": { "rolled_panel_separation_minimum": "at least 2 millimeters of separation between rolled panels" }, "declared_reference_support": [ { "field": "rolled_panel_separation_minimum", "evidence": [ { "document_id": "dovetail-packing", "page": 1, "quote": "Maintain at least 2 millimeters of separation between rolled panels.", "char_start": 967, "char_end": 1035 } ] } ], "actual_emitted_citation_audit": [ { "field": "rolled_panel_separation_minimum", "native_id": "3", "document_id": "dovetail-packing", "emitted_quote": "KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nDovetail Parcel Desk: Exhibition Panel Packing Guide\r\nTransport-tube boundary\r\nGuide 4 applies to fictional Dovetail transfers of Copperfern panels for the Tallreed exhibition during 2027. It\r\nconcerns approved transport tubes and recorded handover, while panel preparation and acceptance follow\r\ntheir separate sources. A packed tube, confirmed booking and physically observed transfer are distinct stages.\r\nThe public fixture describes a procedure rather than a real carrier offer or a completed shipment receipt.\r\nThe accepted limit is 8 rolled panels per transport tube. The unit is a prepared panel rather than a source cut or\r\nan entire exhibition. The declared tube type and arrangement remain part of the source check. A physically\r\nlarger tube cannot extend this accepted limit, and a historical demonstration must not govern a new transfer\r\nunder the effective guide.\r\nMaintain at least 2 millimeters of separation between rolled panels. The separation concerns the declared\r\narrangement and is distinct from the loom's belt clearance or a cloth-width measurement. The sender records\r\nthe prepared tube before handover. A familiar container or correct route label cannot prove that the source\r\narrangement was checked, while a visible issue remains an exception instead of being concealed by the outer\r\nclosure.\r\nThe schedule uses Tuesday and Friday dispatch batches. A prepared tube still needs a confirmed booking and\r\nobserved handover. The schedule is not a guarantee that an unbooked item moves or arrives on either named\r\nday. Staff preserve canceled bookings and distinguish the declared batch from the actual physical event when\r\nreporting a shipment or a completed project transfer.\r\nThe withdrawn demonstration card described tubes holding ten panels. The current guide replaces that\r\narrangement with the accepted limit above. The demonstration remains historical material and does not\r\nconfigure a new transfer. A panel count is distinct from the number of source fabric cuts or a complete\r\nexhibition target, and a physically larger container does not extend the declared accepted arrangement.\r\ndovetail-packingPage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nDovetail Parcel Desk: Exhibition Panel Packing Guide\r\nSealed route and face orientation\r\nThe sender checks the accepted panel list before arranging a transport tube. A correct route cannot resolve an\r\nunrecorded source batch or a visible defect. The guide preserves individual panel identity separately from the", "quote_sha256": "b52e707e9f1582b7e8dfb2de065a9627f2670b561cc5489a690280b2f0941ddc", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false }, { "system": "B", "question_id": "v14-heldout-doc-06", "repetition": 1, "question": "When does Ashmere require the worksheet to be reconciled?", "required_fields": [ "roll_change_reconciliation_trigger" ], "original_raw_answer": "roll_change_reconciliation_trigger: The roll-change event is the declared check trigger and is not replaced by a convenient end-of-month total. [1]", "normalized_answer_fields": { "roll_change_reconciliation_trigger": "The roll-change event is the declared check trigger and is not replaced by a convenient end-of-month total." }, "primary_grade_unchanged": { "version": "kingy-common-document-grade/1.0.0", "status": "unresolved", "fact_status": "unresolved", "passed": false, "needs_review": true, "answerable": true, "correct_abstention": false, "field_checks": [ { "field": "roll_change_reconciliation_trigger", "status": "unresolved", "fact_status": "unresolved" } ], "citation_checks": [ { "field": "roll_change_reconciliation_trigger", "native_id": "1", "status": "pass", "pages": [ 1, 2, 3 ], "page_capability": "uniquely-quote-mapped", "reason": null } ], "reasons": [], "citation_count": 1, "valid_citation_count": 1, "unresolved_citation_count": 0, "explicit_page_count": 0, "quote_mapped_page_count": 1 }, "supplemental_supported_status": "pass", "supplemental_fact_status": "pass", "supplemental_abstention_status": null, "reason": "The roll-change event is the declared check trigger is an explicit source sentence adjacent to Reconcile the worksheet at each roll change in the emitted quote. It answers when reconciliation is required; the end-of-month exclusion reinforces the correct trigger.", "declared_reference_values": { "roll_change_reconciliation_trigger": "at each roll change" }, "declared_reference_support": [ { "field": "roll_change_reconciliation_trigger", "evidence": [ { "document_id": "ashmere-worksheet", "page": 2, "quote": "Reconcile the worksheet at each roll change.", "char_start": 611, "char_end": 655 } ] } ], "actual_emitted_citation_audit": [ { "field": "roll_change_reconciliation_trigger", "native_id": "1", "document_id": "ashmere-worksheet", "emitted_quote": "govern a new entry. A clean printed sheet does not establish that its source changes can be reconstructed or\r\nthat an unresolved inspection became accepted.\r\nashmere-worksheetPage 1 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE\r\nFictional operational material | CC0-1.0 | English text-bearing PDF\r\nAshmere Cooperative: Production Worksheet Standard\r\nReplacement entries and roll changes\r\nA recorder compares the worksheet with the physical cut and source batch before changing an inspection\r\nstate. A displayed color is a summary rather than the underlying observation. The team keeps an unclear\r\nsource reference unresolved until a recorded review supplies the missing fact. A new folder name cannot\r\nestablish a physical movement, and a complete-looking table does not prove that all included cuts were\r\ninspected.\r\nReconcile the worksheet at each roll change. The roll-change event is the declared check trigger and is not\r\nreplaced by a convenient end-of-month total. The recorder keeps an unresolved discrepancy visible. Another\r\nsuccessful session cannot retroactively establish that a missing check occurred at the earlier event, and a\r\nchanged printed form does not alter the physical source batch.\r\nThe production recorder logs worksheet release. The role concerns the administrative record rather than\r\noperating the loom or accepting an exhibition delivery. A participant's suggestion does not replace the logged\r\nrelease. The team preserves the declared authority and distinguishes a recorded worksheet state from a\r\nphysical outcome so a signed finance entry is not used as an inspection result.\r\nThe standard preserves the observed first entry when a later source resolves a discrepancy. A replacement\r\nrow links to the superseded record rather than erasing the earlier uncertainty. The recorder identifies the\r\nauthority for the change and keeps the physical material distinct from an administrative amendment. The\r\nfictional fixture describes a procedure, not an actual production receipt or external assurance of textile quality.\r\nThis standard identifies no hashing algorithm for a digital worksheet archive. A controlled paper-to-record\r\nprocedure does not specify a computer implementation. That technology needs the actual system specification.\r\nA worksheet prefix or revision number cannot recover an absent algorithm, and a reader should not substitute\r\na common software choice for a declared fact in this public administrative source.\r\nashmere-worksheetPage 2 of 3 | Fixture edition 1.4.0 KINGY PUBLIC DOCUMENT FIXTURE", "quote_sha256": "add9f78909f860e4e6fe63d90a196ad1389cab675ae58c7e3af79759ed6c0fa3", "exact_source_grounding": "pass", "grounded_pages": [ 1, 2, 3 ], "primary_support_status": "pass", "supplemental_support_status": "pass" } ], "review_blinded_to_system": false } ], "metric_treatment_findings": [ "The scorer is mechanically symmetric, but literal accepted-value matching is not a reliable bound on semantic factual accuracy. Extractive Kingy passages repeat full reference wording; native concise values often rely on explicit question and field-label context. Seven definite native failures plus two unresolved answerable cases are semantic matches in that context.", "Full answer context must count for both candidates. Minimum, maximum, scope and measurement stage supplied by an emitted field label/question are not missing merely because the numeric value omits redundant wording.", "All emitted citation IDs must retain their actual quote/source linkage. Native cards not claimed in the answer were not credited. Broad unique quote-to-page-range mapping is valid source linkage, but is not an explicit model-issued page number.", "The native scope-amendment excerpt plausibly supports the correct role through introductory scope/authority wording, but omits the exact recording rule. Keep that one support case unresolved rather than adding an uncited gold span.", "Citation denominators count emitted links and differ (A1=38,B1=40); unanswered requested fields do not create artificial citations. Supported-answer success separately penalizes missing answers/citations.", "A1 cross-10 and native cross-02/cross-05/cross-10 are definite wrong-fact responses, even though the frozen fact matcher marks some unresolved. Native cross-04 also omits an available required condition.", "The seven interrupted B2 attempts remain a separate appendix. They cannot establish a complete repetition or product reliability, and were not pooled into this A1/B1 paired pilot." ], "claim_review": { "accuracy_superiority_claim_allowed": false, "general_product_victory_allowed": false, "native_three_repetition_claim_allowed": false, "allowed_descriptive_statement": "On this fixed synthetic corpus and Mac, A1 had 28/30 supported successes under the frozen rubric. B1 had 16 definite passes plus 2 unresolved under that rubric, but source-based contextual review recognizes 25 definite supported successes and one support ambiguity out of 30, with 26/30 factually correct. The lexical-score gap overstates a semantic accuracy difference. Only one complete native session was retained.", "allowed_feature_statement": "Kingy returned explicit page fields with its grounded excerpts. Native answers used emitted numbered source citations whose quoted chunks were reproducibly mapped to PDF page ranges; absence of an explicit page field does not mean the source cannot be inspected.", "raw_primary_optimistic_native_bounds_note": "Granting every primary-native unresolved case gives 18/30 supported and 22/30 canonical-fact passes, but these remain literal-rubric scores; they still exclude semantically correct native answers marked definite failures.", "supplemental_optimistic_native_note": "Granting the remaining support ambiguity yields 26/30 supported successes and 37/40 emitted citation support. These descriptive upper values remain below the selected 90%/95% thresholds in this B1 session, but do not prove general accuracy superiority or establish native repeatability.", "speed_or_privacy_claim_certified_by_this_audit": false }, "limitations": [ "This review is unblinded and partly authored by the Kingy implementation agent. Preserve it as supplemental sensitivity evidence; do not relabel it as blind adjudication.", "The corrected semantic interpretation was not preregistered. It cannot overwrite the frozen primary results or turn the primary lexical score into a validated semantic accuracy metric.", "Only A1 and B1 received this case-level semantic audit. A2/A3 retain the same two primary nonpass IDs, but no claim of three completed native repetitions is possible.", "Forty unique questions on one synthetic corpus and one Mac are the units; three Kingy runs do not produce120 independent tasks or users.", "Factual labels can contain qualitative context such as final-inspection stage. Both candidates received the same labels; the results do not establish performance when ordinary users supply less structured questions.", "Incomplete native sessions and differing runtime/preprocessing/timing boundaries prevent a general comparative speed or reliability claim. Native egress enforcement remains unverified; no comparative privacy claim.", "A future fair accuracy comparison needs a preregistered semantic-equivalence policy and independent blinded assessment of disputed meanings before claiming an advantage." ], "source_outputs_unchanged_during_audit": true, "explicit_native_answerable_semantic_promotions": { "primary_definite_fail_to_contextual_pass": [ "v14-heldout-doc-01", "v14-heldout-doc-04", "v14-heldout-doc-08", "v14-heldout-doc-11", "v14-heldout-doc-16", "v14-heldout-doc-17", "v14-heldout-doc-20" ], "primary_unresolved_to_contextual_pass": [ "v14-heldout-cross-09", "v14-heldout-doc-06" ], "separate_abstention_and_citation_promotion": [ "v14-heldout-unknown-09" ], "remaining_answerable_support_ambiguity": [ "v14-heldout-cross-06" ], "remaining_native_definite_wrong_or_missing_answer_cases": [ "v14-heldout-cross-02", "v14-heldout-cross-04", "v14-heldout-cross-05", "v14-heldout-cross-10" ] } }