{"observation":{"id":"4171c2f8-81c8-472e-abf5-a809333282b6","tool":"reducto","tool_name":"Reducto","criterion":"text-ocr-completeness","criterion_name":"Text & OCR Completeness","criterion_definition":"Extracts all readable content, including scanned pages, with accurate OCR and minimal omissions.","criterion_evidence_type":"transformation","criterion_rank_role":"decisive","criterion_rank_role_reason":"If the tool misses readable text or fails on scanned pages, it has not actually converted the PDF faithfully into Markdown. (3 of 3 judges)","scenario":"hybrid-earnings-report","scenario_name":"Hybrid Earnings Report","group_tag":"hybrid-earnings-report","scenario_description":"A long, real-world hybrid annual report used to benchmark end-to-end PDF-to-markdown conversion. It combines native text, financial tables, charts/graphics, and scanned signature/stamp regions, stressing preservation of document structure, reading order, and embedded visual content across a multi-section report.","modality":"pdf","input_text":null,"input_artifact_refs":[{"alt":null,"url":"https://cdn.futuresmart.ai/public/aidemos/f63a44b441704753b65d1b89dc9b4bbb.pdf?v=1","role":"input","filename":"Target-2015-Annual-Report.pdf"}],"stresses":["Native digital text extraction","Complex financial table preservation","Chart and graphic retention","Image-based signatures and stamps","Reading order across a long multi-section report","Markdown quality and consistency"],"verdict":"worked","score":null,"score_total":null,"note":"Converts the full 84-page native-digital annual report with no skipped pages; the output contains all 84 page-marker pairs, and checked high-risk numbers survive exactly, including the $51,550,988,273 market-value figure and 599,982,121 shares outstanding.","evidence_state":"verified","source":null,"artifacts":[{"url":"https://d3epheqghktydj.cloudfront.net/research-media-reducto-input1-hybridearnings-output-2e5bed5d1e5a.md","role":"output","alt":null},{"url":"https://cdn.futuresmart.ai/public/aidemos/658866653fdd4fe4aa62daa8d2b20f61.png?v=1","role":"output","alt":null},{"url":"https://cdn.futuresmart.ai/public/aidemos/ec1bb100ba61436fb886a28ef022e2a8.png?v=1","role":"input","alt":null},{"url":"https://cdn.futuresmart.ai/public/aidemos/88ed246b865e4fdeb4d198687d2bc10e.png?v=1","role":"output","alt":null},{"url":"https://cdn.futuresmart.ai/public/aidemos/51d9e7dedea64b97b04f95510a9f819a.png?v=1","role":"input","alt":null},{"url":"https://cdn.futuresmart.ai/public/aidemos/4e2fd80334ae454b93048a0a4f51fb43.png?v=1","role":"output","alt":null},{"url":"https://cdn.futuresmart.ai/public/aidemos/c2dfc4c7c3414357967fc4c3a5323583.png?v=1","role":"input","alt":null}],"run_id":"6e3160de-fe46-4b45-b071-72560b5c5d0e","study_title":"Convert a Complex PDF into Clean Markdown with an API","study_kind":"generation","research_task":"86b9h7t37","tested_at":null,"completeness":"input-and-output","input":{"state":"files","text":null,"files":[{"url":"https://cdn.futuresmart.ai/public/aidemos/f63a44b441704753b65d1b89dc9b4bbb.pdf?v=1","filename":"Target-2015-Annual-Report.pdf","alt":"Hybrid Earnings Report","role":"input"}],"modality":"pdf","stresses":["Native digital text extraction","Complex financial table preservation","Chart and graphic retention","Image-based signatures and stamps","Reading order across a long multi-section report","Markdown quality and consistency"]},"tool_page_slug":"reducto","tool_url":"https://aidemos.com/tools/reducto","permalink":"https://aidemos.com/evidence/4171c2f8-81c8-472e-abf5-a809333282b6","api_url":"https://ai.aidemos.com/v1/observations/4171c2f8-81c8-472e-abf5-a809333282b6"},"peers":[{"id":"dd1db974-fbae-43d1-8a89-bbbd8cf1255d","tool":"extend-ai","tool_name":"Extend AI","verdict":"mixed","score":null,"score_total":null,"note":"Reads low-clarity signer and auditor markings, but the stamp OCR is not perfect: 'LLP' is misread as '1LP'.","artifact_count":5,"thumbnail":"https://cdn.futuresmart.ai/public/aidemos/ce874b3599394bcdb0e813ae4006d734.png?v=1","evidence_url":"https://aidemos.com/evidence/dd1db974-fbae-43d1-8a89-bbbd8cf1255d"},{"id":"f77e8735-7e24-462a-bd4c-6e29808f7156","tool":"llamaparse","tool_name":"LlamaParse","verdict":"worked","score":null,"score_total":null,"note":"Recovers a blurred signature stamp as readable text, showing OCR can salvage low-quality scanned text rather than dropping it entirely.","artifact_count":2,"thumbnail":"https://cdn.futuresmart.ai/public/aidemos/6cbf1e9cd25a47e483a0144f1dcc4b61.png?v=1","evidence_url":"https://aidemos.com/evidence/f77e8735-7e24-462a-bd4c-6e29808f7156"},{"id":"b3b06de4-1c1f-451a-a26e-d1bb11daed26","tool":"tensorlake","tool_name":"Tensorlake","verdict":"mixed","score":null,"score_total":null,"note":"On a blurry Ernst & Young signoff, the OCR preserves the firm reference but makes a symbol-level mistake by rendering the ampersand as '+', so the text is close but not exact.","artifact_count":2,"thumbnail":"https://cdn.futuresmart.ai/public/aidemos/6cbf1e9cd25a47e483a0144f1dcc4b61.png?v=1","evidence_url":"https://aidemos.com/evidence/b3b06de4-1c1f-451a-a26e-d1bb11daed26"}],"other_criteria":[{"id":"55895767-e0f8-4be6-9d07-f0998a55a02d","criterion":"advanced-features","criterion_name":"Advanced Features","rank_role":"context","verdict":"failed","score":null,"score_total":null,"note":"On the bar-chart schema, extract.run gets only 3 of 5 Sales values right: 2011 and 2012 are silently pulled from a different table ($69.9B and $73.3B), with no confidence or warning signal.","artifact_count":3,"evidence_url":"https://aidemos.com/evidence/55895767-e0f8-4be6-9d07-f0998a55a02d"},{"id":"5b164bff-8ffe-46e2-b493-883b4ed9fed4","criterion":"advanced-features","criterion_name":"Advanced Features","rank_role":"context","verdict":"worked","score":null,"score_total":null,"note":"The separate schema-driven extract.run endpoint cleanly recovers both a financial table and a donut chart: all requested table rows and all five donut percentages come back exactly.","artifact_count":4,"evidence_url":"https://aidemos.com/evidence/5b164bff-8ffe-46e2-b493-883b4ed9fed4"},{"id":"7fb155c1-7886-4bd4-9e2d-1fec0f47b677","criterion":"advanced-features","criterion_name":"Advanced Features","rank_role":"context","verdict":"mixed","score":null,"score_total":null,"note":"The parse pipeline’s confidence field catches all 10 footer-number corruptions and also produces one false alarm on a correct securities table, so the signal is real but not perfectly calibrated.","artifact_count":6,"evidence_url":"https://aidemos.com/evidence/7fb155c1-7886-4bd4-9e2d-1fec0f47b677"},{"id":"1961fa7a-2d5a-4ece-83fe-23b4c0399568","criterion":"complex-document-handling","criterion_name":"Complex Document Handling","rank_role":"decisive","verdict":"mixed","score":null,"score_total":null,"note":"Processes the full 84-page mixed-content report in 25.5 seconds with no truncation and keeps the hardest nested table’s 27 values exact, but 10 footer numbers are corrupted in the back two-thirds of the file.","artifact_count":9,"evidence_url":"https://aidemos.com/evidence/1961fa7a-2d5a-4ece-83fe-23b4c0399568"},{"id":"277d3d32-8789-468a-b996-53077b678a24","criterion":"markdown-quality","criterion_name":"Markdown Quality","rank_role":"decisive","verdict":"struggled","score":null,"score_total":null,"note":"Produces usable but not clean markdown: the output contains 12 <signature> tags, 2 <empty> placeholders, and literal HTML tags such as <b>, <i>, and <u> instead of pure CommonMark.","artifact_count":3,"evidence_url":"https://aidemos.com/evidence/277d3d32-8789-468a-b996-53077b678a24"},{"id":"77c69713-eb97-47e0-8b45-b4d8c8902bc2","criterion":"reading-order-structure","criterion_name":"Reading Order & Structure","rank_role":"decisive","verdict":"struggled","score":null,"score_total":null,"note":"Drops heading structure on most of the filing: the FORM 10-K opening area receives no # or ## markup, and the report says 80 of 84 pages stay as plain paragraphs rather than navigable section headings.","artifact_count":4,"evidence_url":"https://aidemos.com/evidence/77c69713-eb97-47e0-8b45-b4d8c8902bc2"},{"id":"88babab8-a43b-4c7e-9f67-8dd32e3984bd","criterion":"reading-order-structure","criterion_name":"Reading Order & Structure","rank_role":"decisive","verdict":"worked","score":null,"score_total":null,"note":"Linearizes a two-column page correctly, reading the full left column, including all four bullets, before the right-column continuation appears, with no interleaving or sequence inversion.","artifact_count":2,"evidence_url":"https://aidemos.com/evidence/88babab8-a43b-4c7e-9f67-8dd32e3984bd"},{"id":"95351b6f-5f39-4442-81c6-7e668e79d68f","criterion":"table-preservation","criterion_name":"Table Preservation","rank_role":"decisive","verdict":"worked","score":null,"score_total":null,"note":"Preserves a genuinely row-paired 52-row state table as a clean grid, keeping the 1,792 stores total and 239,539 retail-square-footage total intact.","artifact_count":4,"evidence_url":"https://aidemos.com/evidence/95351b6f-5f39-4442-81c6-7e668e79d68f"},{"id":"0521f27e-23a0-457d-a653-8d443d84ee9b","criterion":"table-preservation","criterion_name":"Table Preservation","rank_role":"decisive","verdict":"failed","score":null,"score_total":null,"note":"Mis-merges two independent side-by-side lists into one table, forcing unrelated cost-of-sales and SG&A bullets into fake row pairs where no row correspondence exists in the source.","artifact_count":2,"evidence_url":"https://aidemos.com/evidence/0521f27e-23a0-457d-a653-8d443d84ee9b"},{"id":"d44d1388-5a3e-4021-94ac-d7281d773533","criterion":"visual-content-retention","criterion_name":"Visual Content Retention","rank_role":"decisive","verdict":"worked","score":null,"score_total":null,"note":"Retains visual content as real image outputs when requested: the portrait, donut chart, and bar chart are all returned as preserved figure/image crops rather than being dropped.","artifact_count":10,"evidence_url":"https://aidemos.com/evidence/d44d1388-5a3e-4021-94ac-d7281d773533"},{"id":"d6a44781-5816-4bba-9774-fad1fe0cbd76","criterion":"visual-content-retention","criterion_name":"Visual Content Retention","rank_role":"decisive","verdict":"failed","score":null,"score_total":null,"note":"Does not retain handwritten signature graphics as images: a signature crop returns only the <signature> placeholder plus the printed name/title, omitting the cursive mark itself.","artifact_count":2,"evidence_url":"https://aidemos.com/evidence/d6a44781-5816-4bba-9774-fad1fe0cbd76"}],"appears_in":[{"page_type":"ranking","slug":"pdf-to-markdown-apis","title":"Best AI Tools to Convert Complex PDFs into Clean Markdown with an API","url":"https://aidemos.com/best/pdf-to-markdown-apis","binding":"run"}],"same_scenario":[{"id":"dd1db974-fbae-43d1-8a89-bbbd8cf1255d","tool":"extend-ai","tool_name":"Extend AI","verdict":"mixed","score":null,"score_total":null,"note":"Reads low-clarity signer and auditor markings, but the stamp OCR is not perfect: 'LLP' is misread as '1LP'."},{"id":"f77e8735-7e24-462a-bd4c-6e29808f7156","tool":"llamaparse","tool_name":"LlamaParse","verdict":"worked","score":null,"score_total":null,"note":"Recovers a blurred signature stamp as readable text, showing OCR can salvage low-quality scanned text rather than dropping it entirely."},{"id":"b3b06de4-1c1f-451a-a26e-d1bb11daed26","tool":"tensorlake","tool_name":"Tensorlake","verdict":"mixed","score":null,"score_total":null,"note":"On a blurry Ernst & Young signoff, the OCR preserves the firm reference but makes a symbol-level mistake by rendering the ampersand as '+', so the text is close but not exact."}]}