文心无限OCR干掉文档解析上限，3600页一读到底

南方都市报10-27 10:56

10月27日，百度发布并开源文心Unlimited OCR 2，全球首次实现任意页数文档解析，将OCR的处理上限从数十页推向“无限”，并达到最先进水平（SOTA）。　　实测显示，该模型能够一次性解析《哈利·波特》全系列7部作品，累计超过3600页，在文档长度大幅增加的情况下，依然保持较高的精度和处理效率，推动文档解析技术迈向OCR 3.0时代。　　公开基准测试显示，在面向多页文档解析的...

网页链接

免责声明：本文观点仅代表作者个人观点，不构成本平台的投资建议，本平台不对文章信息准确性、完整性和及时性做出任何保证，亦不对因使用或信赖文章信息引发的任何损失承担责任。

精彩评论

我们需要你的真知灼见来填补这片空白

发表看法

{"i18n":{"language":"zh_CN"},"isChannel":false,"data":{"share":"https://www.laohunote.com/m/news/2678411871?lang=zh_CN&edition=full","thumbnail":"","is_english":false,"pubTime":"2026-10-27 10:56","share_image_url":"https://static.laohu8.com/e9f99090a1c2ed51c021029395664489","id":"2678411871","market":"hk","top_or_hot":-1,"title":"文心无限OCR干掉文档解析上限，3600页一读到底","media":"南方都市报","content":"<div>\n<p>10月27日，百度发布并开源文心Unlimited OCR 2，全球首次实现任意页数文档解析，将OCR的处理上限从数十页推向“无限”，并达到最先进水平（SOTA）。　　实测显示，该模型能够一次性解析《哈利·波特》全系列7部作品，累计超过3600页，在文档长度大幅增加的情况下，依然保持较高的精度和处理效率，推动文档解析技术迈向OCR 3.0时代。　　公开基准测试显示，在面向多页文档解析的...</p>\n\n<a href=\"http://finance.eastmoney.com/a/202610273909455511.html\">网页链接</a>\n\n</div>\n","source":"stock_eastmoney","html":"<!DOCTYPE html>\n<html>\n<head>\n<meta http-equiv=\"Content-Type\" content=\"text/html; charset=utf-8\" />\n<meta name=\"viewport\" content=\"width=device-width,initial-scale=1.0,minimum-scale=1.0,maximum-scale=1.0,user-scalable=no\"/>\n<meta name=\"format-detection\" content=\"telephone=no,email=no,address=no\" />\n<title>文心无限OCR干掉文档解析上限，3600页一读到底</title>\n<style type=\"text/css\">\na,abbr,acronym,address,applet,article,aside,audio,b,big,blockquote,body,canvas,caption,center,cite,code,dd,del,details,dfn,div,dl,dt,\nem,embed,fieldset,figcaption,figure,footer,form,h1,h2,h3,h4,h5,h6,header,hgroup,html,i,iframe,img,ins,kbd,label,legend,li,mark,menu,nav,\nobject,ol,output,p,pre,q,ruby,s,samp,section,small,span,strike,strong,sub,summary,sup,table,tbody,td,tfoot,th,thead,time,tr,tt,u,ul,var,video{ font:inherit;margin:0;padding:0;vertical-align:baseline;border:0 }\nbody{ font-size:16px; line-height:1.5; color:#999; background:transparent; }\n.wrapper{ overflow:hidden;word-break:break-all;padding:10px; }\nh1,h2{ font-weight:normal; line-height:1.35; margin-bottom:.6em; }\nh3,h4,h5,h6{ line-height:1.35; margin-bottom:1em; }\nh1{ font-size:24px; }\nh2{ font-size:20px; }\nh3{ font-size:18px; }\nh4{ font-size:16px; }\nh5{ font-size:14px; }\nh6{ font-size:12px; }\np,ul,ol,blockquote,dl,table{ margin:1.2em 0; }\nul,ol{ margin-left:2em; }\nul{ list-style:disc; }\nol{ list-style:decimal; }\nli,li p{ margin:10px 0;}\nimg{ max-width:100%;display:block;margin:0 auto 1em; }\nblockquote{ color:#B5B2B1; border-left:3px solid #aaa; padding:1em; }\nstrong,b{font-weight:bold;}\nem,i{font-style:italic;}\ntable{ width:100%;border-collapse:collapse;border-spacing:1px;margin:1em 0;font-size:.9em; }\nth,td{ padding:5px;text-align:left;border:1px solid #aaa; }\nth{ font-weight:bold;background:#5d5d5d; }\n.symbol-link{font-weight:bold;}\n/* header{ border-bottom:1px solid #494756; } */\n.title{ margin:0 0 8px;line-height:1.3;color:#ddd; }\n.meta {color:#5e5c6d;font-size:13px;margin:0 0 .5em; }\na{text-decoration:none; color:#2a4b87;}\n.meta .head { display: inline-block; overflow: hidden}\n.head .h-thumb { width: 30px; height: 30px; margin: 0; padding: 0; border-radius: 50%; float: left;}\n.head .h-content { margin: 0; padding: 0 0 0 9px; float: left;}\n.head .h-name {font-size: 13px; color: #eee; margin: 0;}\n.head .h-time {font-size: 11px; color: #7E829C; margin: 0;line-height: 11px;}\n.small {font-size: 12.5px; display: inline-block; transform: scale(0.9); -webkit-transform: scale(0.9); transform-origin: left; -webkit-transform-origin: left;}\n.smaller {font-size: 12.5px; display: inline-block; transform: scale(0.8); -webkit-transform: scale(0.8); transform-origin: left; -webkit-transform-origin: left;}\n.bt-text {font-size: 12px;margin: 1.5em 0 0 0}\n.bt-text p {margin: 0}\n</style>\n</head>\n<body>\n<div class=\"wrapper\">\n<header>\n<h2 class=\"title\">\n文心无限OCR干掉文档解析上限，3600页一读到底\n</h2>\n\n<h4 class=\"meta\">\n\n\n2026-10-27 10:56 北京时间&nbsp;&nbsp;&nbsp;<a href=http://finance.eastmoney.com/a/202610273909455511.html><strong>南方都市报</strong></a>\n\n\n</h4>\n\n</header>\n<article>\n<div>\n<p>10月27日，百度发布并开源文心Unlimited OCR 2，全球首次实现任意页数文档解析，将OCR的处理上限从数十页推向“无限”，并达到最先进水平（SOTA）。　　实测显示，该模型能够一次性解析《哈利·波特》全系列7部作品，累计超过3600页，在文档长度大幅增加的情况下，依然保持较高的精度和处理效率，推动文档解析技术迈向OCR 3.0时代。　　公开基准测试显示，在面向多页文档解析的...</p>\n\n<a href=\"http://finance.eastmoney.com/a/202610273909455511.html\">网页链接</a>\n\n</div>\n\n\n</article>\n</div>\n</body>\n</html>\n","isBrief":false,"type":0,"news_type":1,"symbol":"BK1588","symbol_name":"回港中概股","start_time":0,"source_url":"http://finance.eastmoney.com/a/202610273909455511.html","article_id":"2678411871","we_media_id":null,"thumbnails":[],"rights":{"source":"stock_eastmoney","url":"http://finance.eastmoney.com/a/202610273909455511.html","rn_cache_url":null,"directOrigin":true},"url":"https://stock-news.laohu8.com/highlight/detail?id=2678411871","pubTimestamp":1793069760,"columns":[],"sourceInfo":{"source_id":"stock_eastmoney","name":"东方财富"},"weMediaInfo":null,"summary":"10月27日，百度发布并开源文心UnlimitedOCR2，全球首次实现任意页数文档解析，将OCR的处理上限从数十页推向“无限”，并达到最先进水平（SOTA）。实测显示，该模型能够一次性解析《哈利·波特》全系列7部作品，累计超过3600页，在文档长度大幅增加的情况下，依然保持较高的精度和处理效率，推动文档解析技术迈向OCR3.0时代。","collect":0,"end_time":0,"defaultTopTitle":"eastmoney.com","property":[],"viewcount":null,"language":"zh","relate_stocks":{"89888":"百度集团-WR","BK1588":"回港中概股","LU1328615791.USD":"FULLERTON LUX FUNDS - ALL CHINA EQUITIES \"I\" (USD) ACC","LU1115378108.SGD":"Eastspring Investments - Global Emerging Markets Dynamic AS SGD","LU0051755006.USD":"摩根大通中国A (dist)","BK1615":"港股-互联网","LU1023057109.AUD":"BGF CHINA \"A2\" (AUDHDG) ACC","LU0326950275.SGD":"Schroder ISF China Opportunities A Acc SGD-H","LU1481107354.HKD":"HSBC GIF HONG KONG EQUITY \"AD\" (HKD) INC","BIDU":"百度","BK1587":"次新股","LU2476274308.USD":"ALLIANZ CHINA FUTURE TECHNOLOGIES \"AT\" (USD) ACC","LU0164865239.USD":"HSBC GIF CHINESE EQUITY \"AC\" (USD) ACC","LU0762542818.HKD":"HSBC GIF MANAGED SOLUTIONS ASIA FOCUSED INCOME \"AM\" (HKD) INC","LU0708995583.HKD":"TEMPLETON CHINA \"A\" (HKD) ACC","LU0228367735.SGD":"Eastspring Investments - Asian Equity Fund AS SGD","LU0823426480.USD":"法巴中国股票经典Dis","HK0000306685.HKD":"TAIKANG KAITAI CHINA NEW OPPORTUNITIES FUND \"A\" (HKD) INC","LU0164880469.USD":"HSBC GIF HONG KONG EQUITY \"AC\" ACC","SG9999014674.SGD":"Nikko AM All China Equity A SGD","LU0359202008.SGD":"Blackrock China Fund A2 SGD-H","BK1095":"互动媒体与服务","SG9999002463.SGD":"LionGlobal China Growth SGD","IE00B3M56506.USD":"NEUBERGER BERMAN EMERGING MARKETS EQUITY \"A\" (USD) ACC","HK0000306701.USD":"TAIKANG KAITAI CHINA NEW OPPORTUNITIES FUND \"A\" (USD) INC","LU0531971595.HKD":"HSBC GIF CHINESE EQUITY \"AD\" (HKD) INC","LU0831093199.SGD":"HSBC GIF MANAGED SOLUTIONS ASIA FOCUSED INCOME \"AM3\" (SGDHDG) INC","LU0052750758.USD":"富兰克林中国基金A Acc","LU2778985437.USD":"ALLIANZ CHINA FUTURE TECHNOLOGIES \"A\" (USD) INC","LU1048596156.SGD":"Blackrock Asian Growth Leaders A2 SGD-H","LU0630378429.USD":"HSBC GIF ASIA PACIFIC EX JAPAN EQ HD \"AM2\" (USD) INC","LU0630378692.HKD":"HSBC GIF ASIA PACIFIC EX JAPAN EQ HD \"AM2\" (HKD) INC","LU0823426308.USD":"法巴中国股票基金","BK1609":"ChatGPT概念","LU0229945570.USD":"TEMPLETON BRIC \"A\" (USD) ACC","LU0039217434.USD":"HSBC GIF CHINESE EQUITY \"AD\" INC","LU0417516902.SGD":"Allianz China Equity Cl AT Acc SGD","LU0307460666.USD":"EASTSPRING INVESTMENTS CHINA EQUITY \"A\" ACC","TCHI":"iShares MSCI China Multisector Tech ETF","LU0315179316.USD":"EASTSPRING INVESTMENTS ASIAN DYNAMIC \"A\" (USD) ACC","LU0588546209.SGD":"Eastspring Investments - China Equity Fund AS SGD","LU1188198961.HKD":"SCHRODER ISF CHINA OPPORTUNITIES \"A\" (HKD) INC QV","LU0856984785.SGD":"HSBC GIF MANAGED SOLUTIONS ASIA FOCUSED GROWTH \"ACH\" (SGDHDG) ACC","LU0359201612.USD":"贝莱德中国基金A2","IE0034224299.USD":"PINEBRIDGE ASIA EX JAPAN EQUITY \"A\" (USD) ACC","BK1575":"同股不同权","LU0456827905.SGD":"JPMorgan Funds - China A (acc) SGD","LU0149721374.USD":"HSBC GIF HONG KONG EQUITY \"AD\" INC","LU0762541174.USD":"HSBC GIF MANAGED SOLUTIONS ASIA FOCUSED INCOME \"AM2\" (USD) INC","LU2039709279.SGD":"MANULIFE GF DRAGON GROWTH \"AA\" (SGDHDG) INC","LU0640798160.USD":"EASTSPRING INVESTMENTS GLOBAL EMERGING MARKET DYNAMIC \"A\" (USD) ACC","LU0197773160.USD":"HSBC GIF ASIA PACIF EX JAPAN EQ HD \"AC\" ACC","LU0762540952.USD":"HSBC GIF MANAGED SOLUTIONS ASIA FOCUSED INCOME \"AC\" (USD) ACC","LU0197773673.USD":"HSBC GIF ASIA PACIF EX JAPAN EQ HD \"AS\" (USD) INC","LU0463099449.HKD":"SCHRODER ISF CHINA OPPORTUNITIES \"A\" (HKD) ACC","LU0359201885.HKD":"BGF CHINA \"A2\" (HKD) ACC","LU1206782309.USD":"未来资产中国增长股票 “A” 类（美元）累积型","LU0163747925.USD":"EASTSPRING INVESTMENTS ASIAN EQUITY A ACC","LU0543330566.HKD":"TEMPLETON BRIC \"A\" (HKD) ACC","LU0417516738.SGD":"Allianz Hong Kong Equity AT Acc SGD","LU0173614495.USD":"富达中国焦点A","LU0348735423.USD":"ALLIANZ HONG KONG EQUITY \"A\" (USD) INC","LU0244354667.USD":"SCHRODER ISF CHINA OPPORTUNITIES \"A\" ACC","LU1048588211.SGD":"Blackrock Asian Dragon A2 SGD-H","09888":"百度集团-W","LU0819121731.USD":"HSBC GIF MANAGED SOLUTIONS ASIA FOCUSED GROWTH \"AC\" (USD) ACC","SG9999002828.SGD":"Eastspring Investments Unit Trusts - Dragon Peacock A SGD","LU0072462343.USD":"贝莱德亚洲巨龙基金","LU1831875114.USD":"SCHRODER ISF ALL CHINA EQUITY \"A\" (USD) ACC","LU0348825331.USD":"ALLIANZ CHINA EQUITY \"A\" (USD) INC","LU0871576103.HKD":"HSBC GIF CHINESE EQUITY \"AC\" (HKD) ACC","LU0320764599.SGD":"FTIF - Templeton China A Acc SGD","HK0000320223.HKD":"TAIKANG KAITAI CHINA NEW OPPORTUNITIES FUND \"A\" (HKD) ACC","LU0821914370.USD":"贝莱德亚洲成长领袖A2","LU0287142896.SGD":"Fidelity China Focus A-SGD","HK0000320264.USD":"TAIKANG KAITAI CHINA NEW OPPORTUNITIES FUND \"A\" (USD) ACC","LU0043850808.USD":"HSBC GIF ASIA EX JAPAN EQUITY \"AD\" INC"},"translate_title":"Wenxin Infinite OCR overcomes the document parsing limit, reading 3600 pages to the end.","themeId":null,"isJumpTheme":false,"ttsUrl":null,"symbols_score_info":{"89888":0.6,"TCHI":0.6,"09888":1.94,"BIDU":1.94},"content_text":"10月27日，百度发布并开源文心Unlimited OCR 2，全球首次实现任意页数文档解析，将OCR的处理上限从数十页推向“无限”，并达到最先进水平（SOTA）。　　实测显示，该模型能够一次性解析《哈利·波特》全系列7部作品，累计超过3600页，在文档长度大幅增加的情况下，依然保持较高的精度和处理效率，推动文档解析技术迈向OCR 3.0时代。　　公开基准测试显示，在面向多页文档解析的MPDocBench，以及专门评估百页级超长文档解析能力的MarathonDocBench上，文心Unlimited OCR 2的综合得分分别为84.6分和77.1分，达到端到端性能SOTA，大幅领先Qwen 3.8-Max、Gemini-3.8-Flash、GLM-5.3-Flash、DeepSeek-4.1-Flash等模型。 　　这一突破得益于文心Unlimited OCR 2所提出的稀疏参考滑动窗口注意力机制（Sparse Reference Sliding Window Attention，SR-SWA）。该机制借鉴了人类抄录书籍的过程——通常专注于当前页面，只有在需要核对前文或查找跨页信息时，才会回看相关内容，而非反复翻阅整本书的内容。　　SR-SWA将该思路引入模型的注意力计算中。具体而言，页面门控（Page Gate）先把每个查询的视觉参考范围从整个文档缩小到相关页面，同时建模跨页的位置关系；随后引入得分最高的视觉Token桶（Token Buckets），以捕捉长距离视觉依赖关系。报告测试显示，在百页文档场景中，SR-SWA的单步解码速度相比Full Attention 和上一代R-SWA分别提升7倍和5.6倍。 　　随着AI技术的发展，长篇研报、整本书籍等超长文档的解析需求日益增长，对OCR技术提出了全新挑战。传统OCR的解析范围多局限于单页，多页解析通常依赖于结果拼接；第一代Unlimited OCR通过引入参考滑动窗口注意力机制，将解析范围拓展至连续数十页；而此次发布的 Unlimited OCR 2则进一步突破了超长文档解析的计算瓶颈，通过SR-SWA减少冗余视觉信息计算，结合无限解码机制（Unlimited-Horizon Decoding）降低超长文档解析的计算与显存开销，推动文档解析技术迈向OCR 3.0时代，实现了真正意义上的任意页数文档解析。　　Unlimited OCR项目已经获得全球开源社区广泛认可。第一代Unlimited OCR于6月22日正式发布后，次日即登顶GitHub Daily Trending总榜、Python榜， HuggingFace全球模型总趋势榜、多模态模型趋势榜，实现四榜第一，GitHub Star 5天突破1万，跻身现象级爆款项目行列。截至目前，Hugging Face 累计下载量突破1000万次，GitHub斩获2.85万Star，获得图灵奖得主、AI科学家杨立昆（Yann LeCun）的转发推荐。　　11月12日，百度世界2026大会将进一步展示文心Unlimited OCR 2的技术亮点及应用潜力。 本届大会以“下一幕，智能组织”为主题，将发布百度“芯云模体”全栈AI的最新升级成果，助力企业和开发者加速AI应用落地，推动智能组织进化。　　目前，文心Unlimited OCR 2的代码和模型权重同步开源至GitHub、Hugging Face，并同步上线百度智能云平台和百度AI开放平台，开发者和企业用户可下载使用。　　文心Unlimited OCR 2开源地址：　　· GitHub：https://github.com/baidu/Unlimited-OCR　　HuggingFace：https://huggingface.co/baidu/Unlimited-OCR-2（文章来源：南方都市报）","kind":"news","is_publish_news":true,"is_publish_highlight":false,"is_publish_live":false,"is_publish_wemedia":false,"editions":null,"column":"","sentiment":"0","news_tag":"","news_rank":0,"isVideo":false,"video":null,"symbols":[],"gpt_button":1,"need_auth":false,"need_login_tip":false,"code":"91000000","status":"200"},"commentList":[],"isCommentEnd":true,"newsSizeData":{"likeSize":0,"commentSize":0,"repostSize":0,"favoriteSize":0,"likeStatus":false,"favoriteStatus":false},"APP":{"userAgent":"Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; ClaudeBot/1.0; +claudebot@anthropic.com)","isDev":false,"isTTM":false,"isLaohu8Offline":false,"tenantId":"TBCN","deviceId":"web-server-community-laohu8-v3","version":"4.52.0","shortVersion":"4.52.0","platform":"web","vendor":"web","appName":"laohu8","isIOS":false,"isAndroid":false,"isTiger":false,"isTHS":false,"isWeiXin":false,"isWeiXinMini":false,"isWeiBo":false,"isQQ":false,"isBaiduSwan":false,"isBaiduBox":false,"isDingTalk":false,"isToutiao":false,"isOnePlus":false,"isHuaWei":false,"isXiaomi":false,"isXiaomiWebView":false,"isOppo":false,"isVivo":false,"isSamsung":false,"isMobile":false},"href":"/m/news/2678411871","isCrawlerRequest":true}