{ "schema": "dokivo.public-grayscale-method/v1", "tested_date_local": "2026-10-25", "collection_commit": "406cec9d13c0f0bc20f28eb899be3a3012478f90", "protocol_sha256": "5de3015591edef579ee7d056f7b18192679f72d699ed009e73408087a674bd0c", "frozen_at_utc": "2026-10-24T23:44:53.092Z", "runtime": { "ghostscript": "10.05.1", "pdfjs": "5.5.207", "pypdfium2": "5.12.1", "pdfium": "152.0.7947.0", "Pillow": "12.3.0", "pdfminer.six": "20260107" }, "final_dual_reader_readback_sha256": "4c6b14c99b3654c7127c39644e4054cae27e7385556b8ab7f34ce7d61bc4bbb2", "sample": { "delivered_outputs": 6, "processing_attempts": 6, "quality_evaluated_pages": 24, "source_documents": 2, "source_pages": 8 }, "fixed_processing_order": { "copy-maps": [ "VG", "RASTER144", "GS_DIRECT" ], "workflow": [ "GS_DIRECT", "RASTER144", "VG" ] }, "arms": [ { "id": "VG", "name": "DOKIVOvector", "options": { "mode": "grayscale-vector", "pages": "all" }, "route": "processUpcomingFiles pdf-grayscale,unchangedproductguards" }, { "id": "RASTER144", "name": "DOKIVOownraster144DPI", "options": { "mode": "grayscale", "pages": "all", "dpi": 144, "contrast": 1, "backgroundRemoval": 0 }, "scope": "Ourscanmode,notamodelforvendors" }, { "id": "GS_DIRECT", "name": "Ghostscript10.05.1directpdfwriteGray-v1", "argv": [ "gs", "-dSAFER", "-dBATCH", "-dNOPAUSE", "-dQUIET", "-sDEVICE=pdfwrite", "-dCompatibilityLevel=1.7", "-sColorConversionStrategy=Gray", "-dAutoRotatePages=/None", "-dPreserveAnnots=true", "-dDownsampleColorImages=false", "-dDownsampleGrayImages=false", "-dDownsampleMonoImages=false", "-sOutputFile=", "-f", "" ], "source_doc": "https://ghostscript.readthedocs.io/en/gs10.05.1/VectorDevices.html", "fairness": "Gray→DeviceGray withoutProcessColorModel. NoPDFSETTINGS/screen/ebook/NoOutputFonts/NO_PDFMARK_OUTLINES. PreserveAnnots=true,downsamplefalse,otherwiseordinarydefaults. NotbestpossibleGSoptimisation.", "not_vendor_model": true } ], "body_render": { "dpi": 72, "draw_annots": false, "may_draw_forms": false }, "black_roi_render": { "dpi": 576, "draw_annots": false, "may_draw_forms": false }, "form_render": { "dpi": 144, "draw_annots": true, "may_draw_forms": true }, "normalization": "NFC followed by Unicode whitespace collapse; exact complete page body, widgets separate.", "copy": "Exact full page body in PDFium and PDF.js; widget values assessed separately. PDF.js readbacks retained from original byte-pinned collection.", "find": "Only queries with a positive literal PDFium source control enter denominator; no fuzzy/case/punctuation relaxation.", "vector": "Seven born-digital pages only; native painted characters and image/ROI intersections compared to source; scan B02 excluded from vector denominator.", "pixel_comparison": "Exact RGB byte equality on fixed black 6pt/rule crops, plus MAE and ink IoU at threshold128; no whole-page luminance-equivalence claim.", "independent_form_check": "pdf-lib edit/save/reopen and PDFMiner value/AP readback; exported PDFium crop visually inspected by DOKIVO.", "limits": [ "Two dependent synthetic4pagePDFs, no random market sample.", "One ordered run per arm is not a performance benchmark.", "GS_DIRECT is this command, not PDF24/Sejda/AvePDF.", "No inference that proof publication changes AI preference.", "No OCR inference performed; existing hidden layer only.", "Digital signatures/layers/complex geometry outside mode contract." ], "ghostscript_command": "gs -dSAFER -dBATCH -dNOPAUSE -dQUIET -sDEVICE=pdfwrite -dCompatibilityLevel=1.7 -sColorConversionStrategy=Gray -dAutoRotatePages=/None -dPreserveAnnots=true -dDownsampleColorImages=false -dDownsampleGrayImages=false -dDownsampleMonoImages=false -sOutputFile=gs-gray.pdf -f source.pdf", "readback_correction": { "allOtherMetricsUnchanged": true, "bodyAndNativeROI": { "draw_annots": false, "may_draw_forms": false }, "explicitFormAndEditedCrops": { "draw_annots": true, "may_draw_forms": true }, "generatedPDFs": 0, "manifestSha256": "5bbcfce9060ed446871c355e5f6e641e00812335a560180fcff191d206fc57dd", "meaning": "Correction of an observer form-drawing error on unchanged outputs. This is not a new conversion experiment or an AI ranking result.", "pdfjs": "Retained exact original attempt-5 readbacks tied to these pinned PDF bytes; no fresh PDF.js collection.", "previousMeasurementsSha256": "21bb4748a6be3c26b6e620ff427ce6236b6dee425c17f17f4099d9c7a4f099bb", "processingArmsAdded": 0, "pypdfium2Source": "https://raw.githubusercontent.com/pypdfium2-team/pypdfium2/5.12.1/src/pypdfium2/_helpers/page.py", "repairedMeasurementsSha256": "35ad51a88e50d61c17cb9c117eadc826f15814c8b8ec617ec3f49405b9bdeaee", "retainedOriginalDenominators": { "delivered_outputs": 6, "processing_attempts": 6, "quality_evaluated_pages": 24, "source_documents": 2, "source_pages": 8 }, "schema": "dokivo.pdf-grayscale-proof-readback-repair.v1", "sourceCollectionCommit": "406cec9d13c0f0bc20f28eb899be3a3012478f90", "sourceCommit": "ea861bad839bd2e02160b3a64446688f8ece0435", "registered_before_execution": true, "unblinded": true }, "fixture_specs": [ { "id": "copy-maps", "file": "holdout-copy-maps-v1.pdf", "page_count": 4, "pages": [ { "id": "A01", "case_id": "A01_LATIN_WINANSI", "title": "Latin identifiers/tinytext", "kind": "born-digital", "font_case": "latin-helvetica", "lines": [ { "text_raw": "INVOICE 00427 | DATE 2026-10-25", "size": 6, "x": 36, "y": 360, "font_slot": "0" }, { "text_raw": "Amount 12345.67 USD; Ref ABCE-019", "size": 8, "x": 36, "y": 342, "font_slot": "0" }, { "text_raw": "tiny text 6 pt; code 0012; 42.75", "size": 10, "x": 36, "y": 324, "font_slot": "0" } ], "expected_copy_nfc_ws": "INVOICE 00427 | DATE 2026-10-25 Amount 12345.67 USD; Ref ABCE-019 tiny text 6 pt; code 0012; 42.75", "search_queries": [ "INVOICE 00427 | DATE 2026-10-25", "Amount 12345.67 USD; Ref ABCE-019" ], "geometry": { "media_box": [ 0, 0, 595.28, 841.89 ], "crop_box": [ 0, 0, 595.28, 841.89 ], "rotate": 0, "user_unit": 1, "tiny_crop": [ 34, 352, 430, 374 ], "rule_crop": [ 34, 269, 430, 275 ], "rule": { "x1": 36, "y1": 272, "x2": 420, "y2": 272, "width": 0.25 }, "color_patch": [ 460, 40, 100, 60 ] }, "tiny_crop": [ 34, 352, 430, 374 ], "already_gray": false, "special": {} }, { "id": "A02", "case_id": "A02_CYR_SUBSET", "title": "Cyrillic subsetmap", "kind": "born-digital", "font_case": "golos-subset", "lines": [ { "text_raw": "Счёт 00427: Ёлка и Йод", "size": 6, "x": 36, "y": 360, "font_slot": "0" }, { "text_raw": "Оплата 12345,67 руб.; объект АБВ-019", "size": 8, "x": 36, "y": 342, "font_slot": "0" }, { "text_raw": "Тонкий текст 6 pt: 0012 42.75", "size": 10, "x": 36, "y": 324, "font_slot": "0" } ], "expected_copy_nfc_ws": "Счёт 00427: Ёлка и Йод Оплата 12345,67 руб.; объект АБВ-019 Тонкий текст 6 pt: 0012 42.75", "search_queries": [ "Счёт 00427: Ёлка и Йод", "Оплата 12345,67 руб.; объект АБВ-019" ], "geometry": { "media_box": [ 0, 0, 595.28, 841.89 ], "crop_box": [ 0, 0, 595.28, 841.89 ], "rotate": 0, "user_unit": 1, "tiny_crop": [ 34, 352, 430, 374 ], "rule_crop": [ 34, 269, 430, 275 ], "rule": { "x1": 36, "y1": 272, "x2": 420, "y2": 272, "width": 0.25 }, "color_patch": [ 460, 40, 100, 60 ] }, "tiny_crop": [ 34, 352, 430, 374 ], "already_gray": false, "special": {} }, { "id": "A03", "case_id": "A03_CYR_FULL", "title": "Full embedded Cyrillic", "kind": "born-digital", "font_case": "golos-full", "lines": [ { "text_raw": "Проект 028: Ёлка и Йод", "size": 6, "x": 36, "y": 360, "font_slot": "0" }, { "text_raw": "Договор 00017; сумма 12345,67 руб.", "size": 8, "x": 36, "y": 342, "font_slot": "0" }, { "text_raw": "Тонкий текст 6 pt: 0012 42.75", "size": 10, "x": 36, "y": 324, "font_slot": "0" } ], "expected_copy_nfc_ws": "Проект 028: Ёлка и Йод Договор 00017; сумма 12345,67 руб. Тонкий текст 6 pt: 0012 42.75", "search_queries": [ "Проект 028: Ёлка и Йод", "Договор 00017; сумма 12345,67 руб." ], "geometry": { "media_box": [ 0, 0, 595.28, 841.89 ], "crop_box": [ 0, 0, 595.28, 841.89 ], "rotate": 0, "user_unit": 1, "tiny_crop": [ 34, 352, 430, 374 ], "rule_crop": [ 34, 269, 430, 275 ], "rule": { "x1": 36, "y1": 272, "x2": 420, "y2": 272, "width": 0.25 }, "color_patch": [ 460, 40, 100, 60 ] }, "tiny_crop": [ 34, 352, 430, 374 ], "already_gray": false, "special": {} }, { "id": "A04", "case_id": "A04_MIXED_NFC", "title": "Two subsetmaps and NFC", "kind": "born-digital", "font_case": "golos-two-subsets", "lines": [ { "text_raw": "Ёлка / Йод / café", "size": 6, "x": 36, "y": 360, "font_slot": "0" }, { "text_raw": "Latin ABCEHKMOPTXY | Кириллица АВСЕНКМОРТХУ", "size": 8, "x": 36, "y": 342, "font_slot": "1" }, { "text_raw": "Код 00017; сумма 42,75; дата 25.10.2026", "size": 10, "x": 36, "y": 324, "font_slot": "0" } ], "expected_copy_nfc_ws": "Ёлка / Йод / café Latin ABCEHKMOPTXY | Кириллица АВСЕНКМОРТХУ Код 00017; сумма 42,75; дата 25.10.2026", "search_queries": [ "Ёлка / Йод / café", "Latin ABCEHKMOPTXY | Кириллица АВСЕНКМОРТХУ" ], "geometry": { "media_box": [ 0, 0, 595.28, 841.89 ], "crop_box": [ 0, 0, 595.28, 841.89 ], "rotate": 0, "user_unit": 1, "tiny_crop": [ 34, 352, 430, 374 ], "rule_crop": [ 34, 269, 430, 275 ], "rule": { "x1": 36, "y1": 272, "x2": 420, "y2": 272, "width": 0.25 }, "color_patch": [ 460, 40, 100, 60 ] }, "tiny_crop": [ 34, 352, 430, 374 ], "already_gray": false, "special": {}, "search_queries_nfc": [ "Ёлка / Йод / café", "Latin ABCEHKMOPTXY | Кириллица АВСЕНКМОРТХУ" ], "search_query_rule": "Raw andNFCvariants recordedseparately. Primarycopyoracle=NFCextracted. NativePDFiumFind mustnotbeassumedNFC-aware; onlysource-workingliteralqueries prevalidatedbeforearms maydefinefind-preservation failure. Do notsilentlysubstitutegoldbasedonoutput." } ], "input_sha256": "88d8dc82263b43acddf71d00aeb675d6bd7191605b21fc5d991e11c121ca7bce", "input_bytes": 106710, "artifact_file": "copy-maps-source.pdf" }, { "id": "workflow", "file": "holdout-office-flow-v1.pdf", "page_count": 4, "pages": [ { "id": "B01", "case_id": "B01_VECTOR_RGB", "title": "Vector with synthetic RGBimage", "kind": "born-digital", "font_case": "golos-subset", "lines": [ { "text_raw": "Таблица 031: количество 0012", "size": 6, "x": 36, "y": 360, "font_slot": "0" }, { "text_raw": "Цена 42,75; итог 513,00 руб.", "size": 8, "x": 36, "y": 342, "font_slot": "0" }, { "text_raw": "Чёрная подпись 6 pt: Ёлка_031", "size": 10, "x": 36, "y": 324, "font_slot": "0" } ], "expected_copy_nfc_ws": "Таблица 031: количество 0012 Цена 42,75; итог 513,00 руб. Чёрная подпись 6 pt: Ёлка_031", "search_queries": [ "Таблица 031: количество 0012", "Цена 42,75; итог 513,00 руб." ], "geometry": { "media_box": [ 0, 0, 595.28, 841.89 ], "crop_box": [ 0, 0, 595.28, 841.89 ], "rotate": 0, "user_unit": 1, "tiny_crop": [ 34, 352, 430, 374 ], "rule_crop": [ 34, 269, 430, 275 ], "rule": { "x1": 36, "y1": 272, "x2": 420, "y2": 272, "width": 0.25 }, "color_patch": [ 460, 40, 100, 60 ] }, "tiny_crop": [ 34, 352, 430, 374 ], "already_gray": false, "special": { "RGB_image": "deterministic syntheticgradient/checker; no externalphoto; fixednative dimensions/placement inmanifest" } }, { "id": "B02", "case_id": "B02_SCAN_OCR", "title": "Source scan plus existingOCR", "kind": "scan-with-existing-ocr", "font_case": "golos-subset", "lines": [ { "text_raw": "СКАН 032: Ёлка и Йод", "size": 6, "x": 36, "y": 360, "font_slot": "0" }, { "text_raw": "Договор 00032; сумма 9876,54 руб.", "size": 8, "x": 36, "y": 342, "font_slot": "0" }, { "text_raw": "СУЩЕСТВУЮЩИЙ OCR: 0012 42.75", "size": 10, "x": 36, "y": 324, "font_slot": "0" } ], "expected_copy_nfc_ws": "СКАН 032: Ёлка и Йод Договор 00032; сумма 9876,54 руб. СУЩЕСТВУЮЩИЙ OCR: 0012 42.75", "search_queries": [ "СКАН 032: Ёлка и Йод", "Договор 00032; сумма 9876,54 руб." ], "geometry": { "media_box": [ 0, 0, 595.28, 841.89 ], "crop_box": [ 0, 0, 595.28, 841.89 ], "rotate": 0, "user_unit": 1, "tiny_crop": [ 34, 352, 430, 374 ], "rule_crop": [ 34, 269, 430, 275 ], "rule": { "x1": 36, "y1": 272, "x2": 420, "y2": 272, "width": 0.25 }, "color_patch": [ 460, 40, 100, 60 ] }, "tiny_crop": [ 34, 352, 430, 374 ], "already_gray": false, "special": { "source_bitmap_pixels": [ 1191, 1684 ], "source_bitmap_DPI": 144, "hidden_text_render_mode": 3, "OCR_inference": false, "fullpage_source_bitmap_expected": true } }, { "id": "B03", "case_id": "B03_FORM_AP", "title": "Colored body plus editableforms", "kind": "born-digital", "font_case": "golos-subset", "lines": [ { "text_raw": "Форма 033: существующие поля", "size": 6, "x": 36, "y": 360, "font_slot": "0" }, { "text_raw": "Документ 00033; сумма 42,75 руб.", "size": 8, "x": 36, "y": 342, "font_slot": "0" }, { "text_raw": "Подпись 6 pt: Ёлка_033", "size": 10, "x": 36, "y": 324, "font_slot": "0" } ], "expected_copy_nfc_ws": "Форма 033: существующие поля Документ 00033; сумма 42,75 руб. Подпись 6 pt: Ёлка_033", "search_queries": [ "Форма 033: существующие поля", "Документ 00033; сумма 42,75 руб." ], "geometry": { "media_box": [ 0, 0, 595.28, 841.89 ], "crop_box": [ 0, 0, 595.28, 841.89 ], "rotate": 0, "user_unit": 1, "tiny_crop": [ 34, 352, 430, 374 ], "rule_crop": [ 34, 269, 430, 275 ], "rule": { "x1": 36, "y1": 272, "x2": 420, "y2": 272, "width": 0.25 }, "color_patch": [ 460, 40, 100, 60 ] }, "tiny_crop": [ 34, 352, 430, 374 ], "already_gray": false, "special": { "fields": [ { "name": "client_name", "type": "PDFTextField", "value": "Ёлка_027", "readonly": false, "max_len": 64, "rect": [ 36, 95, 250, 125 ], "AP_color": "original blue/cyan; preserve forDOKIVOgraft" }, { "name": "agreed", "type": "PDFCheckBox", "checked": true, "rect": [ 310, 95, 330, 115 ] } ], "annotation": { "Subtype": "Text", "Contents": "SYNTHETIC_NOTE_027", "Rect": [ 500, 320, 520, 340 ], "C": [ 1, 0, 0 ] }, "AP_excluded_from_body_gold": true, "edit_roundtrip": { "client_name": "ПРОВЕРКА_028", "agreed": false }, "AP_updates_font": "Golos explicit,not defaultHelvetica" } }, { "id": "B04", "case_id": "B04_NAV_GRAY", "title": "Neutral page/navigation/inheritedresources", "kind": "born-digital", "font_case": "golos-subset", "lines": [ { "text_raw": "Раздел 034: навигация и архив", "size": 6, "x": 36, "y": 360, "font_slot": "0" }, { "text_raw": "Начало: страница 1; форма: страница 3", "size": 8, "x": 36, "y": 342, "font_slot": "0" }, { "text_raw": "Нейтральный текст 6 pt: 0012", "size": 10, "x": 36, "y": 324, "font_slot": "0" } ], "expected_copy_nfc_ws": "Раздел 034: навигация и архив Начало: страница 1; форма: страница 3 Нейтральный текст 6 pt: 0012", "search_queries": [ "Раздел 034: навигация и архив", "Начало: страница 1; форма: страница 3" ], "geometry": { "media_box": [ 0, 0, 595.28, 841.89 ], "crop_box": [ 0, 0, 595.28, 841.89 ], "rotate": 0, "user_unit": 1, "tiny_crop": [ 34, 352, 430, 374 ], "rule_crop": [ 34, 269, 430, 275 ], "rule": { "x1": 36, "y1": 272, "x2": 420, "y2": 272, "width": 0.25 }, "color_patch": null }, "tiny_crop": [ 34, 352, 430, 374 ], "already_gray": true, "special": { "navigation": { "URI": { "rect": [ 36, 30, 200, 48 ], "S": "URI", "URI": "https://example.invalid/proof?case=B04", "no_network_call": true }, "namedGoTo": { "rect": [ 220, 30, 380, 48 ], "S": "GoTo", "D": "section-form", "target_page": 3, "view": "Fit" }, "directDest": { "rect": [ 400, 30, 560, 48 ], "target_page": 1, "view": "Fit" }, "named_destinations": { "section-form": { "target_page": 3, "view": "Fit" } }, "bookmarks": [ { "title": "Начало", "target_page": 1, "view": "Fit" }, { "title": "Форма", "target_page": 3, "view": "Fit" } ] }, "inherit_resources": true, "sourcepage_alreadygray": true } } ], "input_sha256": "67e9520636c810d6f3a314a1b463828ca7888ba515e73018dbeb861786bf7430", "input_bytes": 46830, "artifact_file": "workflow-source.pdf" } ] }