[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85030-en":3,"doc-seo-85030-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85030,1374391975076,"Riley","https://ap-avatar.wpscdn.com/avatar/14000253ca4ec9f6853?x-image-process=image/resize,m_fixed,w_180,h_180&k=1783305029341752051",8,"Research & Report","Infinity Parser2 Technical Report","Infinity-Parser2 introduces a large multimodal model for end-to-end document parsing that combines a controllable data-synthesis pipeline with multi-task reinforcement learning to overcome limited annotated corpora. It delivers a scalable synthesis engine and an open-source Infinity-Doc2-5M dataset with 5M bilingual samples and rich annotations. A verifiable multi-task reward system unifies eight objectives, and two variants—Flash and Pro—reach state-of-the-art results on key benchmarks while generalizing to charts, chemical formulas, and document VQA.","INFINITY-PARSER2 TECHNICAL REPORT  \narXiv :2607 .07836v 1 [ cs .AI] 8 Jul 2026  \nINF Team  \n Models  Dataset  Code  Demo  \nABSTRACT  \nWe present Infinity-Parser2, a large multimodal model that couples a controllable data-synthesis pipeline with multi-task reinforcement learning for end-to-end document parsing, addressing the persistent scarcity of faithfully annotated parsing corpora. Our contributions are threefold. First, we build a scalable synthesis engine, pairing a controllable rendering framework with an iterative refinement loop, and use it to construct and open-source Infinity-Doc2-5M: a 5-million-sample bilingual (Chinese/English) corpus spanning diverse document types, annotated with element bounding boxes, canonical content forms (Markdown, HTML, LaTeX, SMILES, structured charts), and full-page reading order. Second, we introduce a verifiable, multi-task reward system that enables Joint Reinforcement Learning across eight co-trained objectives (document parsing, layout analysis, table parsing, math formula parsing, chart parsing, chemical formula parsing, document VQA, and general multimodal understanding), unifying perception, structure, and reasoning in a single optimization signal. Third, we release two variants under a shared architecture: Infinity-Parser2-Flash, optimized for low-latency inference with a 3.68× throughput gain over Infinity-Parser-7B, and Infinity-Parser2-Pro, engineered for precision-critical settings. Infinity-Parser2-Pro reaches state-of-the-art 87.6% onolmOCR-Bench and 74.3% on ParseBench, surpassing DeepSeek-OCR-2, PaddleOCR-VL-1.5, and MinerU2.5, with strong generalization to charts, chemical formulas, and document VQA.  \nolmOCR-bench  \n90  \n85  \n80  \n75  \n70  \n65  \n\n| 87.6 | 86.0 |\n| --- | --- |\n|  |  |\n|  | 78.5\u003Cbr>76.3\u003Cbr>75.2 |\n|  | 73.6 |\n|  |  |\n\nInfinity-Parser2-Pro Infinity-Parser2-Flash  \nPaddleOCR-VL-1.5 DeepSeek-OCR-2  \nMinerU2.5 Gemini-3.1-Pro  \nParseBench  \n85  \n75  \n65  \n55  \n45  \n35  \n25  \n\n| 74.3 |  |  |  |  |\n| --- | --- | --- | --- | --- |\n| 72.2 | 69.1 | 66.0 |  |  |\n|  |  |  |  |  |\n|  |  |  | 45.9 |  |\n|  |  |  |  | 41.2 |\n|  |  |  |  |  |\n\nInfinity-Parser2-Pro Infinity-Parser2-Flash  \nGemini-3.1-Pro PaddleOCR-VL-1.5  \nMinerU2.5 DeepSeek-OCR-2  \nOmniDocBench-v1.6  \nPaddleOCR-VL-1.5 Infinity-Parser2-Pro  \nMinerU2.5 Gemini-3-Pro  \nInfinity-Parser2-Flash DeepSeek-OCR-2  \n100  \n80  \n60  \n40  \n Infinity-Parser2-Pro  Infinity-Parser2-Flash  PaddleOCR-VL-1.5  DeepSeek-OCR-2  MinerU2.5  Gemini-3-Pro  \n\n|  |  |  |  |  | 94.76\u003Cbr>92.41\u003Cbr>84.60 | 89.53 | 91.40\u003Cbr>89.07 | 86.52 | 80.49 | 86.16 |  |  |  | 96.43\u003Cbr>93.16 |  | 93.68 | 88.89\u003Cbr>79.53 |  | 91.87 |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n|  |  |  |  |  |  |  |  |  |  |  | 73.19\u003Cbr>63.34 |  | 70.55 |  |  |  |  |  |  |\n| 52.41 | 46.05 | 50.21 |  | 51.62 |  |  |  |  |  | 49.73 |  | 47.02 |  |  | 43.42 |  |  |  |  |\n|  |  |  | 33.03\u003Cbr> | N/A |  |  |  |  |  | N/A\u003Cbr>N/A | N/A |  | /A |  | N/A\u003Cbr>N/A |  |  | 37.66\u003Cbr>/A\u003Cbr>\u003Cbr>N/A |  |\n\nN  \nN  \n20  \nLayout Analysis (D4LA) Table Parsing (PubTabNet) Chart Parsing (Chart2Table) Chemical Formula Parsing (CoSyn_Chemical) Document VQA (DocVQA) General MMU (AI2D)  \nFigure 1: Performance evaluation. Top: document parsing benchmarks. Bottom: cross-domain multi-task capabilities.  \nContents  \n1 Introduction 3  \n2 Related Work 4  \n2.1 Pipeline-based Methods .......................................... 4  \n2.2 End-to-end Methods ........................................... 4  \n2.3 RL-based Methods ............................................ 4  \n3 Data Curation 4  \n3.1 Data Iteration Flywheel .......................................... 5  \n3.2 DOM-Based Document Synthesis Engine ................................ 7  \n3.3 Dataset Preparation ............................................ 9  \n4 Model Training 11  \n4.1 Overall Training Strategy ......................................... 12  \n4.2 Supervised Fine-Tuni","cbCaid5EyeIPVwLo","https://ap.wps.com/l/cbCaid5EyeIPVwLo","pdf",27962272,1,44,"English","en",105,"# Introduction\n# Related Work\n## Pipeline-based Methods\n## End-to-end Methods\n## RL-based Methods\n# Data Curation\n## Data Iteration Flywheel\n## DOM-Based Document Synthesis Engine\n## Dataset Preparation\n# Model Training\n## Overall Training Strategy\n## Supervised Fine-Tuning\n## Joint Reinforcement Learning with Verifiable Rewards\n# Experiments\n## Implementation Details\n## Evaluation on Document Structure Tasks\n## Evaluation on Element-level Parsing Tasks\n## Evaluation on Reasoning and Generalization Tasks\n## Ablation Studies\n## Inference Speed\n## Evaluation on Real-World Document Information Extraction\n# Limitations\n# Conclusion\n# Contributions","[{\"question\":\"What problem does Infinity-Parser2 address?\",\"answer\":\"Infinity-Parser2 targets the scarcity of faithfully annotated document parsing corpora by enabling end-to-end parsing through controllable synthetic data generation and multi-task reinforcement learning.\"},{\"question\":\"What is Infinity-Doc2-5M and what does it include?\",\"answer\":\"Infinity-Doc2-5M is a 5-million-sample bilingual dataset spanning diverse document types, annotated with element bounding boxes, canonical content forms (including Markdown, HTML, LaTeX, SMILES, and structured charts), and full-page reading order.\"},{\"question\":\"How do Infinity-Parser2-Flash and Infinity-Parser2-Pro differ?\",\"answer\":\"Infinity-Parser2-Flash is optimized for low-latency inference with a stated throughput gain over Infinity-Parser-7B, while Infinity-Parser2-Pro is engineered for precision-critical settings and achieves strong benchmark results.\"}]",1784200501,111,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"infinity-parser2-technical-report","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/infinity-parser2-technical-report/85030/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does Infinity-Parser2 address?","Question",{"text":75,"@type":76},"Infinity-Parser2 targets the scarcity of faithfully annotated document parsing corpora by enabling end-to-end parsing through controllable synthetic data generation and multi-task reinforcement learning.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is Infinity-Doc2-5M and what does it include?",{"text":80,"@type":76},"Infinity-Doc2-5M is a 5-million-sample bilingual dataset spanning diverse document types, annotated with element bounding boxes, canonical content forms (including Markdown, HTML, LaTeX, SMILES, and structured charts), and full-page reading order.",{"name":82,"@type":73,"acceptedAnswer":83},"How do Infinity-Parser2-Flash and Infinity-Parser2-Pro differ?",{"text":84,"@type":76},"Infinity-Parser2-Flash is optimized for low-latency inference with a stated throughput gain over Infinity-Parser-7B, while Infinity-Parser2-Pro is engineered for precision-critical settings and achieves strong benchmark results.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]