{"id":108264,"date":"2026-09-21T12:32:57","date_gmt":"2026-09-21T04:32:57","guid":{"rendered":"https:\/\/www.wsisp.com\/helps\/108264.html"},"modified":"2026-09-21T12:32:57","modified_gmt":"2026-09-21T04:32:57","slug":"%e6%96%87%e6%a1%a3%e6%99%ba%e8%83%bd%e4%b8%93%e9%a2%98%ef%bc%9a%e5%b0%8f%e6%a8%a1%e5%9e%8b%e5%90%8e%e8%ae%ad%e7%bb%83%e6%80%8e%e4%b9%88%e9%80%89%e6%a0%b7%e6%9c%ac%ef%bc%9fpaddleocr-vl-1-6%e6%8c%89","status":"publish","type":"post","link":"https:\/\/www.wsisp.com\/helps\/108264.html","title":{"rendered":"\u6587\u6863\u667a\u80fd\u4e13\u9898\uff1a\u5c0f\u6a21\u578b\u540e\u8bad\u7ec3\u600e\u4e48\u9009\u6837\u672c\uff1fPaddleOCR-VL-1.6\u6309\u4e09\u7c7b\u5f31\u70b9\u6316\u6398"},"content":{"rendered":"<p>&#xff08;1&#xff09;\u300a\u4e09\u5e74\u9762\u8bd5\u4e94\u5e74\u6a21\u62df\u300bAIGC \/ LLM \/ AI Agent \u7b97\u6cd5\u5de5\u7a0b\u5e08\u4e0e\u5f00\u53d1\u5de5\u7a0b\u5e08\u6c42\u804c\u9762\u8bd5\u79d8\u7c4d&#xff0c;\u72ec\u5bb6\u8d44\u6e90\u89c1 WeThinkIn\/AIGC-Interview-Book&#xff0c;\u6b22\u8fce Star&#xff01; &#xff08;2&#xff09;AIGC \/ LLM \/ AI Agent \u7b97\u6cd5\u5c97\u4e0e\u5f00\u53d1\u5c97\u6c42\u804c\u9762\u8bd5\u5185\u63a8\u5b66\u4e60\u793e\u7fa4&#xff0c;\u6db5\u76d6 AIGC\u3001LLM \u5927\u6a21\u578b\u3001AI Agent\u3001\u4f20\u7edf\u6df1\u5ea6\u5b66\u4e60\u3001\u81ea\u52a8\u9a7e\u9a76\u3001\u673a\u5668\u5b66\u4e60\u3001\u8ba1\u7b97\u673a\u89c6\u89c9\u3001\u81ea\u7136\u8bed\u8a00\u5904\u7406\u3001\u5f3a\u5316\u5b66\u4e60\u3001\u5927\u6570\u636e\u6316\u6398\u3001\u5177\u8eab\u667a\u80fd\u3001\u5143\u5b87\u5b99\u3001AGI \u7b49\u65b9\u5411\u7684\u6700\u65b0\u9762\u8bd5\u5e72\u8d27\u4e0e\u6838\u5fc3\u77e5\u8bc6&#xff0c;\u6b22\u8fce\u52a0\u5165(https:\/\/t.zsxq.com\/33pJ0)&#xff01;<\/p>\n<hr \/>\n<p>\u8bba\u6587\u6807\u9898&#xff1a;PaddleOCR-VL-1.6: Expanding the Frontier of Document Parsing with Under-Optimized Region Refinement and Progressive Post-Training \u8bba\u6587\u4f5c\u8005&#xff1a;\u767e\u5ea6\u98de\u6868 \u53d1\u8868\u65f6\u95f4&#xff1a;2026\u5e746\u67082\u65e5<\/p>\n<hr \/>\n<h4>\u76ee\u5f55<\/h4>\n<ul>\n<li>\n<ul>\n<li>\u4e00\u3001\u4e24\u9636\u6bb5\u6d41\u6c34\u7ebf\u6cbf\u75281.5\u76840.9B\u67b6\u6784<\/li>\n<li>\u4e8c\u3001\u8fb9\u754c\u8106\u5f31\u3001\u8986\u76d6\u7a00\u758f\u3001\u76d1\u7763\u4e0d\u53ef\u9760\u4e09\u8def\u6316\u6398<\/li>\n<li>\u4e09\u3001\u591a\u4e13\u5bb6\u5171\u8bc6\u4e0e\u6e32\u67d3\u5f15\u5bfc\u6807\u6ce8<\/li>\n<li>\u56db\u3001CPT\u5230SFT\u518d\u5230GRPO\u7684\u9012\u8fdb\u540e\u8bad\u7ec3<\/li>\n<li>\u4e94\u3001OmniDocBench v1.6\u603b\u520696.33\u4e0e\u5206\u9879\u77ed\u677f<\/li>\n<li>\u516d\u3001\u672c\u5730\u90e8\u7f72\u4e0e\u63a8\u7406\u5b9e\u8df5<\/li>\n<li>\u4e03\u3001\u603b\u7ed3\u4e0e\u601d\u8003<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<p>\u672c\u6587\u8981\u70b9 &#xff08;1&#xff09;PaddleOCR-VL-1.6\u662f\u767e\u5ea6\u98de\u68682026\u5e746\u6708\u53d1\u5e03\u7684\u6587\u6863\u89e3\u6790\u6a21\u578b&#xff0c;0.9B\u67b6\u6784\u4e0e1.5\u76f8\u540c\u3002 &#xff08;2&#xff09;OmniDocBench v1.6\u603b\u520696.33&#xff0c;\u6bd41.5\u768494.93\u9ad81.40\u5206&#xff0c;\u8868\u683cTEDS\u4ece91.67\u5347\u523094.76\u3002 &#xff08;3&#xff09;\u505a\u6cd5\u4e0a\u6700\u503c\u5f97\u770b\u7684\u662f\u6309\u8fb9\u754c\u8106\u5f31\u3001\u8986\u76d6\u7a00\u758f\u3001\u76d1\u7763\u4e0d\u53ef\u9760\u4e09\u8def\u6316\u6837\u672c&#xff0c;\u518d\u8d70CPT\u3001SFT\u3001GRPO\u3002 &#xff08;4&#xff09;\u77ed\u677f\u662f\u9605\u8bfb\u987a\u5e8f\u6ca1\u62ff\u7b2c\u4e00&#xff0c;\u4e2d\u6587\u624b\u5199\u5b9a\u4f4d\u8bc6\u522b\u4ece89.52\u6389\u523085.90\u3002 &#xff08;5&#xff09;\u6587\u672b\u9644\u6a21\u578b\u5361\u7684\u5b98\u65b9\u63a8\u7406\u793a\u4f8b\u3002<\/p>\n<hr \/>\n<p>\u767e\u5ea6\u98de\u6868\u57282026\u5e746\u67082\u65e5\u628aPaddleOCR-VL-1.6\u7684\u6280\u672f\u62a5\u544a\u6302\u4e0aarXiv&#xff0c;\u7f16\u53f72606.03264\u3002\u6a21\u578b\u5361\u628a\u6743\u91cd\u53d1\u5e03\u5199\u62105\u670828\u65e5\u3002\u6211\u628a\u62a5\u544a\u3001\u6a21\u578b\u5361\u548cHTML\u7248\u5bf9\u4e86\u4e00\u904d\u3002<\/p>\n<p>0.9B\u67b6\u6784\u4e0e1.5\u76f8\u540c&#xff0c;\u7248\u9762\u6a21\u578bPP-DocLayoutV3\u6ca1\u52a8&#xff0c;\u540e\u8bad\u7ec3\u53ea\u52a8\u8bc6\u522b\u90a3\u4e00\u652f\u3002OmniDocBench v1.6\u603b\u520696.33\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260921043250-6ab0b37281cdb.png\" alt=\"PaddleOCR-VL-1.6\u5728OmniDocBench v1.6\u548cReal5\u4e0a\u7684\u67f1\u72b6\u5bf9\u6bd4\" \/><\/p>\n<h3>\u4e00\u3001\u4e24\u9636\u6bb5\u6d41\u6c34\u7ebf\u6cbf\u75281.5\u76840.9B\u67b6\u6784<\/h3>\n<p>\u6587\u6863\u89e3\u6790\u5206\u4e24\u6bb5\u3002PP-DocLayoutV3\u5148\u51fa\u591a\u8fb9\u5f62\u6846\u3001\u7c7b\u522b\u548c\u9605\u8bfb\u987a\u5e8f&#xff0c;\u88c1\u5757\u518d\u6309\u6307\u4ee4\u8bc6\u522b\u6b63\u6587\u3001\u8868\u683c\u3001\u516c\u5f0f\u3001\u56fe\u8868\u548c\u5370\u7ae0&#xff0c;\u540e\u5904\u7406\u62fc\u6210Markdown\u6216JSON\u3002<\/p>\n<p>\u8bc6\u522b\u6a21\u578b0.9B&#xff0c;\u539f\u751f\u5206\u8fa8\u7387\u89c6\u89c9\u7f16\u7801\u5668\u3001\u81ea\u9002\u5e94MLP\u8fde\u63a5\u5668&#xff0c;\u518d\u52a0\u4e0aERNIE-4.5-0.3B\u3002\u6587\u5b57\u5b9a\u4f4d\u4e0d\u8d70\u7248\u9762\u6a21\u578b&#xff0c;0.9B\u81ea\u5df1\u505a\u7aef\u5230\u7aef\u68c0\u6d4b\u548c\u8bc6\u522b\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260921043253-6ab0b375e483a.png\" alt=\"\u7cfb\u7edf\u603b\u89c8\uff0c\u5de6\u4fa7\u67b6\u6784\u5757\u4ecd\u6807\u62101.5\uff0c\u62a5\u544a\u5199\u660e1.6\u8bc6\u522b\u6a21\u578b\u6cbf\u7528\u8fd9\u5957\u7ed3\u6784\" \/><\/p>\n<p>1.5\u5df2\u7ecf\u591f\u5f3a\u3002\u5269\u4e0b\u7684\u9519\u4e0d\u518d\u50cf\u5747\u5300\u566a\u58f0\u3002\u6709\u7684\u6837\u672c\u8f7b\u8f7b\u4e00\u6270\u52a8\u8f93\u51fa\u5c31\u8df3&#xff0c;\u957f\u5c3e\u7248\u5f0f\u51e0\u4e4e\u6ca1\u90bb\u5c45&#xff0c;\u6807\u7b7e\u4e5f\u6709\u4ece\u4e00\u5f00\u59cb\u5c31\u7ad9\u4e0d\u4f4f\u7684\u30020.9B\u5bf9\u6570\u636e\u6548\u7387\u66f4\u654f\u611f&#xff0c;\u540e\u9762\u90fd\u56f4\u7740\u600e\u4e48\u628a\u6837\u672c\u9009\u5230\u8fd8\u6ca1\u5b66\u7a33\u7684\u5730\u65b9\u3002<\/p>\n<h3>\u4e8c\u3001\u8fb9\u754c\u8106\u5f31\u3001\u8986\u76d6\u7a00\u758f\u3001\u76d1\u7763\u4e0d\u53ef\u9760\u4e09\u8def\u6316\u6398<\/h3>\n<p>\u62a5\u544a\u628a\u8fd9\u4e9b\u5730\u65b9\u53eb\u6b20\u4f18\u5316\u533a\u57df\u3002\u6316\u6cd5\u5bf9\u77401.5\u672c\u8eab&#xff0c;\u4e0d\u9760\u628a\u8bed\u6599\u6574\u4f53\u518d\u94fa\u4e00\u5c42\u3002\u6709\u7684\u8def\u95ee\u8f93\u51fa\u7a33\u4e0d\u7a33&#xff0c;\u6709\u7684\u8def\u95ee\u8fd9\u7c7b\u7248\u5f0f\u89c1\u6ca1\u89c1\u591f\u3002\u6807\u7b7e\u810f\u4e0d\u810f\u88ab\u5355\u72ec\u5f00\u51fa\u6765&#xff0c;\u67e5\u7684\u662f\u771f\u503c\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260921043254-6ab0b37658114.png\" alt=\"\u4e09\u8def\u6b20\u4f18\u5316\u533a\u57df\u6316\u6398\u4e0e\u540e\u7eed\u6807\u6ce8\u6d41\u7a0b\" \/><\/p>\n<p>\u8fb9\u754c\u8106\u5f31\u9488\u5bf9\u51b3\u7b56\u9762\u53d1\u98d8\u7684\u6837\u672c\u3002\u4ed6\u4eec\u7559\u4e86\u8bad\u7ec3\u65e5\u7a0b\u6700\u540e8%\u91cc\u7684\u516b\u4e2acheckpoint&#xff0c;\u518d\u7ed9\u540c\u4e00\u5f20\u56fe\u52a0\u5341\u516d\u79cd\u4e0d\u6539\u8bed\u4e49\u7684\u6270\u52a8&#xff0c;\u50cf\u7d20\u5e73\u79fb\u3001JPEG\u538b\u7f29\u3001\u566a\u58f0\u3001\u6a21\u7cca\u3001\u975e\u5747\u5300\u7f29\u653e\u90fd\u5728\u91cc\u9762\u3002\u6bcf\u4e2a\u6837\u672c\u5f97\u5230128\u6b21\u9884\u6d4b&#xff0c;\u4e24\u4e24\u7b97\u5f52\u4e00\u5316\u7f16\u8f91\u8ddd\u79bb&#xff0c;\u4e00\u51718128\u5bf9\u3002<\/p>\n<p>\u4e3a\u4e86\u5c11\u88ab\u6807\u70b9\u683c\u5f0f\u8fd9\u7c7b\u5c0f\u5dee\u5f02\u5e26\u7740\u8d70&#xff0c;\u53ea\u53d6\u6700\u5927\u7684128\u5bf9\u505a\u5e73\u5747\u5f53\u4f5c\u8106\u5f31\u5206\u6570&#xff0c;\u6309\u5206\u6570\u53d6\u524d1%&#xff0c;\u518d\u52a0\u8fdb\u51fa\u73b0\u8fc7\u6a21\u578b\u9000\u5316\u7684\u6837\u672c\u3002\u5341\u591a\u79cd\u5e38\u89c4\u589e\u5f3a\u53e0\u5728\u4e00\u8d77&#xff0c;\u4ecd\u538b\u4e0d\u4f4f\u5176\u4e2d\u4e00\u90e8\u5206\u3002<\/p>\n<p>\u8986\u76d6\u7a00\u758f\u662f\u53e6\u4e00\u7c7b\u9519\u3002\u76f8\u4f3c\u6a21\u5f0f\u8bad\u7ec3\u91cc\u51fa\u73b0\u8fc7&#xff0c;\u6a21\u578b\u8fd8\u662f\u4f1a\u5199\u9519&#xff0c;\u5468\u56f4\u5206\u5e03\u6491\u5f97\u592a\u8584\u3002\u4ed6\u4eec\u7528\u6587\u6863\u4e13\u7528\u7f16\u7801\u5668\u5efa\u76f8\u4f3c\u5ea6\u56fe&#xff0c;\u628a\u9608\u503c\u4e00\u6b65\u6b65\u62ac\u9ad8&#xff0c;\u8ba9\u8fde\u901a\u5757\u88c2\u5f00&#xff0c;\u7559\u4e0b\u5c40\u90e8\u5bc6\u5ea6\u4f4e\u3001\u5bb9\u6613\u88ab\u5927\u7c07\u541e\u6389\u7684\u5c0f\u5757\u3002K-Means\u9884\u5148\u5b9aK\u3001\u6bcf\u4e2a\u70b9\u90fd\u5fc5\u987b\u8fdb\u7c07&#xff0c;\u7a00\u6709\u7248\u5f0f\u5bb9\u6613\u88ab\u5438\u8fdb\u90bb\u8fd1\u7684\u5bc6\u96c6\u56e2\u3002\u79cd\u5b50\u62ff\u53bb\u5185\u90e8\u5927\u6c60\u5b50\u5b9a\u5411\u68c0\u7d22&#xff0c;\u62a5\u544a\u70b9\u540d\u8865\u4e86\u53e4\u7c4d\u3001\u751f\u50fb\u5b57\u548c\u5de5\u4e1a\u8868\u683c\u3002<\/p>\n<p>\u76d1\u7763\u4e0d\u53ef\u9760\u7ba1\u5df2\u7ecf\u6709\u6807\u7b7e\u3001\u4f46\u6807\u7b7e\u53ef\u80fd\u5e26\u504f\u7684\u6837\u672c\u3002\u6a21\u578b\u6709\u65f6\u4f1a\u975e\u5e38\u81ea\u4fe1\u5730\u91cd\u590d\u540c\u4e00\u79cd\u9519&#xff0c;\u62a5\u544a\u628a\u8fd9\u7c7b\u5931\u8d25\u8bb0\u5728\u6807\u7b7e\u4e0a\u3002\u5916\u90e8\u4e13\u5bb6\u662fQianfan-OCR\u3001GLM-OCR\u548cMinerU2.5-Pro\u3002\u539f\u6807\u7b7e\u88ab\u4efb\u610f\u4e00\u5bb6\u652f\u6301\u5c31\u4fdd\u7559\u3002\u4e09\u5bb6\u90fd\u4e0d\u8ba4\u3001\u4f46\u81f3\u5c11\u4e24\u5bb6\u4e13\u5bb6\u5f7c\u6b64\u4e00\u81f4&#xff0c;\u5c31\u7528\u5171\u8bc6\u8986\u76d6\u3002\u4ecd\u8bf4\u4e0d\u5230\u4e00\u5757&#xff0c;\u6807\u6210\u672a\u89e3\u51b3&#xff0c;\u4ea4\u7ed9\u7cbe\u4fee\u6807\u6ce8\u3002<\/p>\n<h3>\u4e09\u3001\u591a\u4e13\u5bb6\u5171\u8bc6\u4e0e\u6e32\u67d3\u5f15\u5bfc\u6807\u6ce8<\/h3>\n<p>\u68c0\u7d22\u56de\u6765\u7684\u56fe\u6ca1\u6709\u6807\u7b7e&#xff0c;\u672a\u89e3\u51b3\u6837\u672c\u7f3a\u5916\u90e8\u652f\u6301&#xff0c;\u4e24\u6761\u90fd\u8981\u91cd\u65b0\u5199\u6210\u80fd\u8bad\u7684\u76d1\u7763\u3002\u81ea\u52a8\u6807\u6ce8\u6362\u53e6\u4e00\u7ec4\u4e13\u5bb6&#xff0c;PaddleOCR-VL-1.5\u3001GLM-OCR\u548cMinerU2.5-Pro\u3002\u81f3\u5c11\u4e24\u5bb6\u4e00\u81f4&#xff0c;\u5171\u8bc6\u76f4\u63a5\u5f53\u6807\u7b7e\u3002\u8fbe\u4e0d\u6210\u5171\u8bc6\u7684&#xff0c;\u8fdb\u5165ERNIE 5.0\u7684Judge-and-Refine\u5faa\u73af\u3002<\/p>\n<p>\u8868\u683c\u548c\u516c\u5f0f\u5f88\u96be\u62ffLaTeX\u3001HTML\u76f4\u63a5\u8ddf\u539f\u56fe\u5bf9&#xff0c;\u4ed6\u4eec\u5148\u628a\u5019\u9009\u8f93\u51fa\u6e32\u67d3\u6210\u56fe&#xff0c;\u8ba9\u5224\u5b9a\u505a\u56fe\u548c\u56fe\u7684\u6bd4\u5bf9&#xff0c;\u53bb\u627e\u9519\u884c\u3001\u9519\u5217\u3001\u8de8\u5355\u5143\u683c\u548c\u5185\u5bb9\u653e\u9519\u4f4d\u3002\u4e13\u5bb6\u8f93\u51fa\u53ea\u5728\u7b2c\u4e00\u8f6e\u6ce8\u5165&#xff0c;\u540e\u9762\u51e0\u8f6e\u53ea\u7528\u5f53\u524d\u9884\u6d4b\u548c\u4e0a\u4e00\u8f6e\u5dee\u5f02&#xff0c;\u907f\u514d\u4e09\u5bb6\u8bf4\u6cd5\u53cd\u590d\u628a\u8f68\u8ff9\u62fd\u56de\u53bb\u3002\u4ecd\u5bf9\u4e0d\u4e0a\u7684\u9001\u5230\u4eba\u5de5&#xff0c;\u6700\u540e\u4e00\u8f6e\u8f93\u51fa\u5f53\u4f5c\u9884\u6807\u6ce8\u3002\u901a\u8fc7\u7387\u548c\u4eba\u5de5\u91cf\u62a5\u544a\u6ca1\u5199\u3002<\/p>\n<h3>\u56db\u3001CPT\u5230SFT\u518d\u5230GRPO\u7684\u9012\u8fdb\u540e\u8bad\u7ec3<\/h3>\n<p>\u8bad\u7ec3\u4ece1.5\u7684checkpoint\u63a5\u7740\u505a\u3002\u65b0\u6570\u636e\u6309\u53ef\u9760\u7a0b\u5ea6\u62c6\u8fdb\u4e09\u622a\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260921043254-6ab0b376ceb78.png\" alt=\"\u4ece1.5\u6743\u91cd\u51fa\u53d1\uff0cCPT\u3001SFT\u3001GRPO\u4e09\u9636\u6bb5\u628a\u6570\u636e\u4ece1680\u4e07\u6536\u52304.9\u4e07\" \/><\/p>\n<p>CPT\u5148\u704c\u65b0\u5206\u5e03\u3002\u8bed\u6599\u662f1.5\u7684\u5168\u90e8SFT\u3001\u4e00\u90e8\u5206\u9884\u8bad\u7ec3&#xff0c;\u518d\u52a0\u4e0a\u65b0\u68c0\u7d22\u5230\u7684\u5168\u90e8\u6837\u672c&#xff0c;\u4e00\u51711680\u4e07\u6761&#xff0c;\u6807\u7b7e\u6362\u6210\u6700\u65b0\u7248\u3002\u5168\u53c2\u6570\u30011\u4e2aepoch\u3001batch 1024\u3001\u5b66\u4e60\u73873e-5\u3002\u53e4\u7c4d\u3001\u751f\u50fb\u5b57\u8fd9\u7c7b\u504f\u79fb&#xff0c;\u9760\u77ed\u77ed\u4e00\u6bb5SFT\u5403\u4e0d\u7a33&#xff0c;\u6240\u4ee5\u653e\u5728\u6700\u524d\u9762\u3002<\/p>\n<p>SFT\u6539\u6536\u7a84\u3002\u6cbf\u75281.5\u7684\u4e0d\u786e\u5b9a\u611f\u77e5\u805a\u7c7b\u91c7\u6837\u6316\u96be\u4f8b&#xff0c;\u7559\u4e0b\u4e13\u5bb6\u672a\u5171\u8bc6\u540e\u7cbe\u4fee\u7684\u6837\u672c&#xff0c;\u518d\u6536\u8fdb\u7ea0\u6b63\u8fc7\u7684\u65e7\u6807\u7b7e&#xff0c;\u5408\u8ba1730\u4e07\u6761&#xff0c;\u5b66\u4e60\u73871e-5&#xff0c;\u4ecd\u662f1\u4e2aepoch\u3001\u5168\u53c2\u6570\u3002<\/p>\n<p>\u5f3a\u5316\u5b66\u4e60\u66f4\u6311\u98df\u3002\u8bed\u8a00\u6a21\u578b\u53ea\u67090.3B&#xff0c;\u6837\u672c\u9009\u98d8\u4e86&#xff0c;\u53ef\u80fd\u4e00\u5c0f\u64ae\u96be\u4f8b\u6da8\u4e86\u3001\u6574\u4f53\u6389\u4e0b\u53bb\u3002\u4ed6\u4eec\u7528SFT\u6a21\u578b\u5f53\u521d\u59cb\u7b56\u7565&#xff0c;\u6bcf\u6761\u91c716\u6b21&#xff0c;\u6e29\u5ea60.85&#xff0c;top-p 0.9&#xff0c;top-k 32\u3002\u6700\u597d\u4e00\u6b21\u4ecd\u5f88\u4f4e\u3001\u5e73\u5747\u5df2\u7ecf\u5f88\u9ad8\u3001\u6700\u597d\u548c\u5e73\u5747\u5dee\u592a\u5c0f\u3001\u5956\u52b1\u65b9\u5dee\u63a5\u8fd1\u96f6&#xff0c;\u8fd9\u56db\u7c7b\u5148\u6254\u6389\u3002<\/p>\n<p>\u7559\u4e0b\u7684\u6253\u9ad8\u6f5c\u529b\u5206\u6570\u3002\u4e3b\u9879\u662f\u6700\u597d\u4e00\u6b21\u51cf\u53bb\u5e73\u5747&#xff0c;\u518d\u6309\u751f\u6210\u4e0d\u786e\u5b9a\u5ea6\u548c\u5956\u52b1\u65b9\u5dee\u52a0\u6743&#xff0c;\u03b1\u53d61\u3001\u03b2\u53d62\u3002\u4e0d\u786e\u5b9a\u5ea6\u7528\u957f\u5ea6\u5f52\u4e00\u5316\u7684\u5e8f\u5217\u7f6e\u4fe1\u5ea6\u6765\u7b97&#xff0c;\u907f\u514d\u957f\u8f93\u51fa\u5929\u7136\u628a\u4f3c\u7136\u6253\u4f4e\u3002OCR\u3001\u56fe\u8868\u3001\u8868\u683c\u3001\u516c\u5f0f\u3001\u5370\u7ae0\u3001\u6587\u5b57\u5b9a\u4f4d\u5404\u53d6\u524d8\u5343\u6761&#xff0c;\u4e00\u51714.9\u4e07\u6761&#xff0c;\u518d\u8bad2\u4e2aepoch&#xff0c;\u5b66\u4e60\u73872e-6\u3002\u7ec4\u5927\u5c0f16&#xff0c;\u8ddf\u4e86DAPO\u7684clip-higher&#xff0c;\u03b5_high\u53d60.28&#xff0c;\u7ec4\u5185\u5956\u52b1\u65b9\u5dee\u4e3a\u96f6\u7684\u66f4\u65b0\u76f4\u63a5\u4e22\u6389\u3002<\/p>\n<p>\u5956\u52b1\u5148\u8fc7\u6709\u6548\u6027\u95e8&#xff0c;\u683c\u5f0f\u574f\u3001\u622a\u65ad\u3001\u9000\u5316\u8bb0\u96f6&#xff0c;\u518d\u4e58\u7ed3\u6784\u9879\u548c\u76f8\u4f3c\u5ea6\u3002\u8868\u683c\u7528TEDS&#xff0c;\u516c\u5f0f\u7528CDM&#xff0c;\u6b63\u6587\u548c\u5370\u7ae0\u75281\u51cf\u5f52\u4e00\u5316\u7f16\u8f91\u8ddd\u79bb&#xff0c;\u56fe\u8868\u7528RMS-F1&#xff0c;\u5b9a\u4f4d\u7528\u7f16\u8f91\u76f8\u4f3c\u5ea6\u52a0\u6743F1\u3002<\/p>\n<table>\n<tr>\u4efb\u52a1\u6709\u6548\u6027\u95e8\u7ed3\u6784\u9879\u76f8\u4f3c\u5ea6<\/tr>\n<tbody>\n<tr>\n<td>\u8868\u683c<\/td>\n<td>\u9000\u5316\u3001\u622a\u65ad\u3001OTSL\u89e3\u6790\u5931\u8d25<\/td>\n<td>\u5355\u5143\u683cLaTeX\u5408\u6cd5\u6027\u3001\u77e9\u5f62\u5316\u7f16\u8f91\u4ee3\u4ef7<\/td>\n<td>TEDS<\/td>\n<\/tr>\n<tr>\n<td>OCR<\/td>\n<td>\u9000\u5316\u3001\u622a\u65ad\u3001\u884c\u5185LaTeX\u975e\u6cd5<\/td>\n<td>\u53d61<\/td>\n<td>1-NED<\/td>\n<\/tr>\n<tr>\n<td>\u516c\u5f0f<\/td>\n<td>\u9000\u5316\u3001\u622a\u65ad\u3001LaTeX\u975e\u6cd5<\/td>\n<td>\u53d61<\/td>\n<td>CDM<\/td>\n<\/tr>\n<tr>\n<td>\u5370\u7ae0<\/td>\n<td>\u9000\u5316\u3001\u622a\u65ad<\/td>\n<td>\u53d61<\/td>\n<td>1-NED<\/td>\n<\/tr>\n<tr>\n<td>\u56fe\u8868<\/td>\n<td>\u9000\u5316\u3001\u622a\u65ad\u3001Markdown\u8868\u975e\u6cd5<\/td>\n<td>\u77e9\u5f62\u5316\u7f16\u8f91\u4ee3\u4ef7<\/td>\n<td>RMS-F1<\/td>\n<\/tr>\n<tr>\n<td>\u5b9a\u4f4d<\/td>\n<td>\u9000\u5316\u3001\u622a\u65ad\u3001\u8f93\u51fa\u683c\u5f0f\u975e\u6cd5<\/td>\n<td>\u53d61<\/td>\n<td>\u7f16\u8f91\u76f8\u4f3c\u5ea6\u52a0\u6743F1<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h3>\u4e94\u3001OmniDocBench v1.6\u603b\u520696.33\u4e0e\u5206\u9879\u77ed\u677f<\/h3>\n<p>v1.6\u52a0\u4e86\u591a\u7c92\u5ea6\u81ea\u9002\u5e94\u5339\u914d&#xff0c;\u5e76\u65b0\u589e296\u9875Hard\u5b50\u96c6\u3002\u6587\u672c\u548c\u9605\u8bfb\u987a\u5e8f\u7528\u7f16\u8f91\u8ddd\u79bb&#xff0c;\u8868\u683c\u7528TEDS&#xff0c;\u516c\u5f0f\u7528CDM\u3002<\/p>\n<table>\n<tr>\u6a21\u578b\u53c2\u6570\u91cfOmniDocBench v1.6\u603b\u5206\u6587\u672cEdit\u2193\u516c\u5f0fCDM\u8868\u683cTEDS\u9605\u8bfb\u987a\u5e8fEdit\u2193<\/tr>\n<tbody>\n<tr>\n<td>PaddleOCR-VL-1.6<\/td>\n<td>0.9B<\/td>\n<td>96.33<\/td>\n<td>0.033<\/td>\n<td>97.49<\/td>\n<td>94.76<\/td>\n<td>0.127<\/td>\n<\/tr>\n<tr>\n<td>MinerU2.5-Pro<\/td>\n<td>1.2B<\/td>\n<td>95.75<\/td>\n<td>0.036<\/td>\n<td>97.45<\/td>\n<td>93.42<\/td>\n<td>0.120<\/td>\n<\/tr>\n<tr>\n<td>GLM-OCR<\/td>\n<td>0.9B<\/td>\n<td>95.22<\/td>\n<td>0.044<\/td>\n<td>97.18<\/td>\n<td>92.83<\/td>\n<td>0.133<\/td>\n<\/tr>\n<tr>\n<td>PaddleOCR-VL-1.5<\/td>\n<td>0.9B<\/td>\n<td>94.93<\/td>\n<td>0.038<\/td>\n<td>96.89<\/td>\n<td>91.67<\/td>\n<td>0.130<\/td>\n<\/tr>\n<tr>\n<td>Youtu-Parsing<\/td>\n<td>2.5B<\/td>\n<td>93.74<\/td>\n<td>0.044<\/td>\n<td>93.63<\/td>\n<td>92.02<\/td>\n<td>0.116<\/td>\n<\/tr>\n<tr>\n<td>Gemini 3 Pro<\/td>\n<td>\u672a\u516c\u5f00<\/td>\n<td>92.91<\/td>\n<td>0.064<\/td>\n<td>95.99<\/td>\n<td>89.15<\/td>\n<td>0.165<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u5206\u9879\u91cc\u8df3\u5f97\u6700\u5927\u7684\u662f\u8868\u683c&#xff0c;\u7ed3\u6784TEDS\u4ece94.37\u5347\u523097.11\u3002\u9605\u8bfb\u987a\u5e8f0.127&#xff0c;\u843d\u540eYoutu-Parsing\u76840.116\u548cMinerU2.5-Pro\u76840.120\u30026\u6708\u4ea4\u7a3f\u65f696.33\u662f\u699c\u9996&#xff0c;7\u6708OvisOCR2\u62a5\u544a\u7ed9\u51fa96.58&#xff0c;\u4e24\u4efd\u62a5\u544a\u7684\u8bc4\u6d4b\u90fd\u6307\u5411\u5b98\u65b9\u699c\u3002<\/p>\n<p>Real5\u6309\u626b\u63cf\u3001\u5f2f\u66f2\u3001\u5c4f\u6444\u3001\u5149\u7167\u3001\u503e\u659c\u4e94\u7ec4\u771f\u5b9e\u62cd\u6444\u6761\u4ef6\u8003\u30021.6\u603b\u520693.19&#xff0c;1.5\u662f92.05&#xff0c;Gemini 3 Pro\u662f89.24&#xff0c;Qwen3-VL-235B-A22B-Instruct\u662f88.90\u3002<\/p>\n<p>\u5185\u90e81258\u5f20\u96be\u8868Overall TEDS 91.71&#xff0c;1801\u5f20\u56fe\u8868RMS-F1 91.74&#xff0c;300\u5f20\u5370\u7ae0NED 0.119&#xff0c;\u90fd\u662f\u81ea\u5bb6\u8003\u5377\u3002\u6587\u5b57\u5b9a\u4f4d\u4e5d\u7ef4\u5e73\u5747\u4ece86.21\u5347\u523087.47&#xff0c;\u6a21\u7cca\u4ece84.22\u5347\u523090.59\u3002\u4e2d\u6587\u624b\u5199\u5374\u4ece89.52\u6389\u523085.90&#xff0c;\u82f1\u6587\u5370\u5237\u4ece86.89\u5fae\u964d\u523086.51&#xff0c;\u53e4\u7c4d\u53ea\u4ece85.23\u632a\u523085.98\u3002<\/p>\n<p>\u8fb9\u754c\u8106\u5f31\u90a3\u8def\u7528\u4e86\u5927\u91cf\u6a21\u7cca\u6270\u52a8&#xff0c;\u8986\u76d6\u7a00\u758f\u70b9\u540d\u8865\u4e86\u53e4\u7c4d&#xff0c;\u4e2d\u6587\u624b\u5199\u843d\u5728\u4e3b\u653b\u5bf9\u8c61\u4e4b\u5916\u3002<\/p>\n<p>\u6d88\u878d\u4ece1.5\u51fa\u53d1&#xff0c;CPT\u523095.62&#xff0c;SFT\u523096.25&#xff0c;RL\u523096.33\u3002\u8868\u683cTEDS\u5728SFT\u5df2\u7ecf\u523094.74&#xff0c;RL\u53ea\u518d\u52a00.02\u3002\u5927\u5934\u5728\u524d\u9762\u4e24\u6bb5\u76d1\u7763\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260921043255-6ab0b377ab16f.png\" alt=\"OmniDocBench v1.6\u603b\u5206\u548c\u8868\u683cTEDS\u968fCPT\u3001SFT\u3001RL\u4e09\u9636\u6bb5\u7684\u53d8\u5316\" \/><\/p>\n<h3>\u516d\u3001\u672c\u5730\u90e8\u7f72\u4e0e\u63a8\u7406\u5b9e\u8df5<\/h3>\n<p>\u6a21\u578b\u5361\u5f3a\u8c03\u67b6\u6784\u548c1.5\u517c\u5bb9\u3002\u5b98\u65b9\u63a8\u8350\u8d70PaddleOCR\u6d41\u6c34\u7ebf\u505a\u6574\u9875\u89e3\u6790&#xff0c;transformers\u793a\u4f8b\u53ea\u8986\u76d6\u5143\u7d20\u7ea7\u8bc6\u522b\u548c\u6587\u5b57\u5b9a\u4f4d\u3002\u5b89\u88c5\u547d\u4ee4\u6765\u81ea\u6a21\u578b\u5361\u3002<\/p>\n<p>python <span class=\"token parameter variable\">-m<\/span> pip <span class=\"token function\">install<\/span> paddlepaddle-gpu<span class=\"token operator\">&#061;&#061;<\/span><span class=\"token number\">3.2<\/span>.1 <span class=\"token parameter variable\">-i<\/span> https:\/\/www.paddlepaddle.org.cn\/packages\/stable\/cu126\/<br \/>\npython <span class=\"token parameter variable\">-m<\/span> pip <span class=\"token function\">install<\/span> <span class=\"token parameter variable\">-U<\/span> <span class=\"token string\">&#034;paddleocr[doc-parser]&gt;&#061;3.6.0&#034;<\/span><br \/>\npaddleocr doc_parser <span class=\"token parameter variable\">-i<\/span> https:\/\/paddle-model-ecology.bj.bcebos.com\/paddlex\/imgs\/demo_image\/paddleocr_vl_demo.png <span class=\"token parameter variable\">&#8211;pipeline_version<\/span> v1.6<\/p>\n<p>pipeline_version\u5199\u6210v1.6\u624d\u4f1a\u8d70\u5230\u65b0\u6743\u91cd\u3002Python\u63a5\u53e3\u662fPaddleOCRVL(pipeline_version&#061;&#034;v1.6&#034;)&#xff0c;\u7ed3\u679c\u53ef\u4ee5save_to_json\u6216save_to_markdown\u3002\u6a21\u578b\u5361\u8981\u6c42PaddlePaddle 3.2.1\u53ca\u4ee5\u4e0a&#xff0c;macOS\u7528\u6237\u6309\u8bf4\u660e\u7528Docker\u3002\u52a0\u901f\u53ef\u7528\u5b98\u65b9vLLM\u955c\u50cf&#xff0c;\u6a21\u578b\u540dPaddleOCR-VL-1.6-0.9B\u3002\u5143\u7d20\u7ea7\u8c03\u7528\u8d70transformers&gt;&#061;5.0.0&#xff0c;\u4efb\u52a1\u9760\u77ed\u6307\u4ee4\u5207\u6362\u3002<\/p>\n<p><span class=\"token keyword\">from<\/span> PIL <span class=\"token keyword\">import<\/span> Image<br \/>\n<span class=\"token keyword\">import<\/span> torch<br \/>\n<span class=\"token keyword\">from<\/span> transformers <span class=\"token keyword\">import<\/span> AutoProcessor<span class=\"token punctuation\">,<\/span> AutoModelForImageTextToText<\/p>\n<p>model_path <span class=\"token operator\">&#061;<\/span> <span class=\"token string\">&#034;PaddlePaddle\/PaddleOCR-VL-1.6&#034;<\/span><br \/>\nimage_path <span class=\"token operator\">&#061;<\/span> <span class=\"token string\">&#034;test.png&#034;<\/span><br \/>\ntask <span class=\"token operator\">&#061;<\/span> <span class=\"token string\">&#034;ocr&#034;<\/span>  <span class=\"token comment\"># ocr \/ table \/ chart \/ formula \/ spotting \/ seal<\/span><\/p>\n<p>image <span class=\"token operator\">&#061;<\/span> Image<span class=\"token punctuation\">.<\/span><span class=\"token builtin\">open<\/span><span class=\"token punctuation\">(<\/span>image_path<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span>convert<span class=\"token punctuation\">(<\/span><span class=\"token string\">&#034;RGB&#034;<\/span><span class=\"token punctuation\">)<\/span><br \/>\norig_w<span class=\"token punctuation\">,<\/span> orig_h <span class=\"token operator\">&#061;<\/span> image<span class=\"token punctuation\">.<\/span>size<br \/>\n<span class=\"token keyword\">if<\/span> task <span class=\"token operator\">&#061;&#061;<\/span> <span class=\"token string\">&#034;spotting&#034;<\/span> <span class=\"token keyword\">and<\/span> orig_w <span class=\"token operator\">&lt;<\/span> <span class=\"token number\">1500<\/span> <span class=\"token keyword\">and<\/span> orig_h <span class=\"token operator\">&lt;<\/span> <span class=\"token number\">1500<\/span><span class=\"token punctuation\">:<\/span><br \/>\n    image <span class=\"token operator\">&#061;<\/span> image<span class=\"token punctuation\">.<\/span>resize<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">(<\/span>orig_w <span class=\"token operator\">*<\/span> <span class=\"token number\">2<\/span><span class=\"token punctuation\">,<\/span> orig_h <span class=\"token operator\">*<\/span> <span class=\"token number\">2<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">,<\/span> Image<span class=\"token punctuation\">.<\/span>Resampling<span class=\"token punctuation\">.<\/span>LANCZOS<span class=\"token punctuation\">)<\/span><\/p>\n<p>max_pixels <span class=\"token operator\">&#061;<\/span> <span class=\"token number\">2048<\/span> <span class=\"token operator\">*<\/span> <span class=\"token number\">28<\/span> <span class=\"token operator\">*<\/span> <span class=\"token number\">28<\/span> <span class=\"token keyword\">if<\/span> task <span class=\"token operator\">&#061;&#061;<\/span> <span class=\"token string\">&#034;spotting&#034;<\/span> <span class=\"token keyword\">else<\/span> <span class=\"token number\">1280<\/span> <span class=\"token operator\">*<\/span> <span class=\"token number\">28<\/span> <span class=\"token operator\">*<\/span> <span class=\"token number\">28<\/span><br \/>\nPROMPTS <span class=\"token operator\">&#061;<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token string\">&#034;ocr&#034;<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token string\">&#034;OCR:&#034;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token string\">&#034;table&#034;<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token string\">&#034;Table Recognition:&#034;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token string\">&#034;formula&#034;<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token string\">&#034;Formula Recognition:&#034;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token string\">&#034;chart&#034;<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token string\">&#034;Chart Recognition:&#034;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token string\">&#034;spotting&#034;<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token string\">&#034;Spotting:&#034;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token string\">&#034;seal&#034;<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token string\">&#034;Seal Recognition:&#034;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><\/p>\n<p>DEVICE <span class=\"token operator\">&#061;<\/span> <span class=\"token string\">&#034;cuda&#034;<\/span> <span class=\"token keyword\">if<\/span> torch<span class=\"token punctuation\">.<\/span>cuda<span class=\"token punctuation\">.<\/span>is_available<span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span> <span class=\"token keyword\">else<\/span> <span class=\"token string\">&#034;cpu&#034;<\/span><br \/>\nmodel <span class=\"token operator\">&#061;<\/span> AutoModelForImageTextToText<span class=\"token punctuation\">.<\/span>from_pretrained<span class=\"token punctuation\">(<\/span><br \/>\n    model_path<span class=\"token punctuation\">,<\/span> torch_dtype<span class=\"token operator\">&#061;<\/span>torch<span class=\"token punctuation\">.<\/span>bfloat16<br \/>\n<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span>to<span class=\"token punctuation\">(<\/span>DEVICE<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span><span class=\"token builtin\">eval<\/span><span class=\"token punctuation\">(<\/span><span class=\"token punctuation\">)<\/span><br \/>\nprocessor <span class=\"token operator\">&#061;<\/span> AutoProcessor<span class=\"token punctuation\">.<\/span>from_pretrained<span class=\"token punctuation\">(<\/span>model_path<span class=\"token punctuation\">)<\/span><br \/>\nmessages <span class=\"token operator\">&#061;<\/span> <span class=\"token punctuation\">[<\/span><span class=\"token punctuation\">{<\/span><br \/>\n    <span class=\"token string\">&#034;role&#034;<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token string\">&#034;user&#034;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token string\">&#034;content&#034;<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token punctuation\">[<\/span><br \/>\n        <span class=\"token punctuation\">{<\/span><span class=\"token string\">&#034;type&#034;<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token string\">&#034;image&#034;<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token string\">&#034;image&#034;<\/span><span class=\"token punctuation\">:<\/span> image<span class=\"token punctuation\">}<\/span><span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token punctuation\">{<\/span><span class=\"token string\">&#034;type&#034;<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token string\">&#034;text&#034;<\/span><span class=\"token punctuation\">,<\/span> <span class=\"token string\">&#034;text&#034;<\/span><span class=\"token punctuation\">:<\/span> PROMPTS<span class=\"token punctuation\">[<\/span>task<span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">}<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">,<\/span><br \/>\n<span class=\"token punctuation\">}<\/span><span class=\"token punctuation\">]<\/span><br \/>\ninputs <span class=\"token operator\">&#061;<\/span> processor<span class=\"token punctuation\">.<\/span>apply_chat_template<span class=\"token punctuation\">(<\/span><br \/>\n    messages<span class=\"token punctuation\">,<\/span><br \/>\n    add_generation_prompt<span class=\"token operator\">&#061;<\/span><span class=\"token boolean\">True<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    tokenize<span class=\"token operator\">&#061;<\/span><span class=\"token boolean\">True<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    return_dict<span class=\"token operator\">&#061;<\/span><span class=\"token boolean\">True<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    return_tensors<span class=\"token operator\">&#061;<\/span><span class=\"token string\">&#034;pt&#034;<\/span><span class=\"token punctuation\">,<\/span><br \/>\n    images_kwargs<span class=\"token operator\">&#061;<\/span><span class=\"token punctuation\">{<\/span><span class=\"token string\">&#034;size&#034;<\/span><span class=\"token punctuation\">:<\/span> <span class=\"token punctuation\">{<\/span><br \/>\n        <span class=\"token string\">&#034;shortest_edge&#034;<\/span><span class=\"token punctuation\">:<\/span> processor<span class=\"token punctuation\">.<\/span>image_processor<span class=\"token punctuation\">.<\/span>min_pixels<span class=\"token punctuation\">,<\/span><br \/>\n        <span class=\"token string\">&#034;longest_edge&#034;<\/span><span class=\"token punctuation\">:<\/span> max_pixels<span class=\"token punctuation\">,<\/span><br \/>\n    <span class=\"token punctuation\">}<\/span><span class=\"token punctuation\">}<\/span><span class=\"token punctuation\">,<\/span><br \/>\n<span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">.<\/span>to<span class=\"token punctuation\">(<\/span>model<span class=\"token punctuation\">.<\/span>device<span class=\"token punctuation\">)<\/span><br \/>\noutputs <span class=\"token operator\">&#061;<\/span> model<span class=\"token punctuation\">.<\/span>generate<span class=\"token punctuation\">(<\/span><span class=\"token operator\">**<\/span>inputs<span class=\"token punctuation\">,<\/span> max_new_tokens<span class=\"token operator\">&#061;<\/span><span class=\"token number\">512<\/span><span class=\"token punctuation\">)<\/span><br \/>\n<span class=\"token keyword\">print<\/span><span class=\"token punctuation\">(<\/span>processor<span class=\"token punctuation\">.<\/span>decode<span class=\"token punctuation\">(<\/span>outputs<span class=\"token punctuation\">[<\/span><span class=\"token number\">0<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">[<\/span>inputs<span class=\"token punctuation\">[<\/span><span class=\"token string\">&#034;input_ids&#034;<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">.<\/span>shape<span class=\"token punctuation\">[<\/span><span class=\"token operator\">&#8211;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">:<\/span><span class=\"token operator\">&#8211;<\/span><span class=\"token number\">1<\/span><span class=\"token punctuation\">]<\/span><span class=\"token punctuation\">)<\/span><span class=\"token punctuation\">)<\/span><\/p>\n<p>max_new_tokens&#061;512\u662f\u5143\u7d20\u7ea7\u957f\u5ea6&#xff0c;\u6574\u9875Markdown\u4e0d\u8981\u6307\u671b\u8fd9\u6bb5\u4ee3\u7801\u3002\u6587\u5b57\u5b9a\u4f4d\u77ed\u8fb9\u4f4e\u4e8e1500\u4f1a\u5148\u653e\u5927\u4e00\u500d&#xff0c;max_pixels\u63d0\u52302048\u00d728\u00d728\u3002<\/p>\n<h3>\u4e03\u3001\u603b\u7ed3\u4e0e\u601d\u8003<\/h3>\n<p>\u6700\u503c\u5f97\u6284\u7684\u662f\u5c0f\u6a21\u578b\u540e\u8bad\u7ec3\u600e\u4e48\u628a\u6837\u672c\u9009\u8584\u3002\u4e09\u8def\u6316\u6398\u628a\u8fd8\u4e0d\u7a33\u3001\u89c1\u5f97\u5c11\u3001\u6807\u7b7e\u810f\u62c6\u5f00&#xff0c;CPT\u3001SFT\u3001RL\u518d\u6309\u53ef\u9760\u7a0b\u5ea6\u5f80\u4e0b\u6536\u3002RL\u8981\u7ec4\u5185\u6bd4\u51fa\u5dee\u5f02&#xff0c;\u7b56\u7565\u8fd8\u5f97\u5076\u5c14\u591f\u5f97\u7740\u66f4\u597d\u7684\u8f93\u51fa\u30024.9\u4e07\u6761\u53ea\u628a\u603b\u5206\u4ece96.25\u63a8\u523096.33&#xff0c;\u5927\u5934\u5728\u524d\u9762\u4e24\u6bb5\u76d1\u7763\u3002<\/p>\n<p>\u9605\u8bfb\u987a\u5e8f\u6ca1\u62ff\u5230\u7b2c\u4e00\u3002\u4e2d\u6587\u624b\u5199\u6389\u4e863.62\u4e2a\u767e\u5206\u70b9\u3002\u7248\u9762\u6a21\u578b\u6574\u4ee3\u6ca1\u52a8&#xff0c;\u5207\u9519\u7684\u533a\u57df\u540e\u9762\u770b\u4e0d\u89c1\u3002\u5185\u90e8\u8868\u3001\u56fe\u3001\u7ae0\u662f\u81ea\u5bb6\u8003\u5377\u3002\u81ea\u52a8\u6807\u6ce8\u7684\u901a\u8fc7\u7387\u62a5\u544a\u6ca1\u5199\u3002\u8fd9\u4e9b\u7f3a\u53e3\u4e0d\u5f71\u54cd\u628a\u6316\u5f31\u70b9\u3001\u518d\u6309\u9636\u6bb5\u704c\u6570\u636e\u8fd9\u4e00\u5957\u62ff\u8d70\u7528\u3002<\/p>\n<hr \/>\n<p>\u53c2\u8003\u94fe\u63a5<\/p>\n<ul>\n<li>\u6280\u672f\u62a5\u544a\u539f\u6587 arXiv 2606.03264 v1&#xff0c;2026\u5e746\u6708&#xff0c;HTML\u7248<\/li>\n<li>\u6a21\u578b\u6743\u91cd\u4e0e\u63a8\u7406\u793a\u4f8b PaddlePaddle\/PaddleOCR-VL-1.6<\/li>\n<li>\u6e90\u7801 PaddlePaddle\/PaddleOCR<\/li>\n<li>OmniDocBench\u699c\u5355<\/li>\n<li>\u524d\u7f6e\u7248\u672c PaddleOCR-VL-1.5&#xff0c;arXiv 2601.21957<\/li>\n<li>\u5bf9\u6bd4\u65b9\u6848 MinerU2.5-Pro\u3001GLM-OCR\u3001OvisOCR2<\/li>\n<li>\u771f\u5b9e\u62cd\u6444\u57fa\u51c6 Real5-OmniDocBench<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>&#xff08;1&#xff09;\u300a\u4e09\u5e74\u9762\u8bd5\u4e94\u5e74\u6a21\u62df\u300bAIGC \/ LLM \/ AI Agent \u7b97\u6cd5\u5de5\u7a0b\u5e08\u4e0e\u5f00\u53d1\u5de5\u7a0b\u5e08\u6c42\u804c\u9762\u8bd5\u79d8\u7c4d&#xff0c;\u72ec\u5bb6\u8d44\u6e90\u89c1 WeThinkIn\/AIGC-Interview-Book&#xff0c;\u6b22\u8fce Star&#xff01; &#xff08;2&#xff09;AIGC \/ LLM \/ AI Agent \u7b97\u6cd5\u5c97\u4e0e\u5f00\u53d1\u5c97\u6c42\u804c\u9762\u8bd5\u5185\u63a8\u5b66\u4e60\u793e\u7fa4&#xff0c;\u6db5\u76d6 AIGC\u3001LLM \u5927\u6a21\u578b\u3001AI Agent\u3001\u4f20\u7edf\u6df1\u5ea6\u5b66\u4e60\u3001\u81ea\u52a8\u9a7e\u9a76\u3001\u673a\u5668\u5b66\u4e60\u3001\u8ba1<\/p>\n","protected":false},"author":2,"featured_media":108259,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[11653,12349,52,50],"topic":[],"class_list":["post-108264","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-server","tag-11653","tag-12349","tag-aigc","tag-50"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v20.3 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>\u6587\u6863\u667a\u80fd\u4e13\u9898\uff1a\u5c0f\u6a21\u578b\u540e\u8bad\u7ec3\u600e\u4e48\u9009\u6837\u672c\uff1fPaddleOCR-VL-1.6\u6309\u4e09\u7c7b\u5f31\u70b9\u6316\u6398 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.wsisp.com\/helps\/108264.html\" \/>\n<meta property=\"og:locale\" content=\"zh_CN\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"\u6587\u6863\u667a\u80fd\u4e13\u9898\uff1a\u5c0f\u6a21\u578b\u540e\u8bad\u7ec3\u600e\u4e48\u9009\u6837\u672c\uff1fPaddleOCR-VL-1.6\u6309\u4e09\u7c7b\u5f31\u70b9\u6316\u6398 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"og:description\" content=\"&#xff08;1&#xff09;\u300a\u4e09\u5e74\u9762\u8bd5\u4e94\u5e74\u6a21\u62df\u300bAIGC \/ LLM \/ AI Agent \u7b97\u6cd5\u5de5\u7a0b\u5e08\u4e0e\u5f00\u53d1\u5de5\u7a0b\u5e08\u6c42\u804c\u9762\u8bd5\u79d8\u7c4d&#xff0c;\u72ec\u5bb6\u8d44\u6e90\u89c1 WeThinkIn\/AIGC-Interview-Book&#xff0c;\u6b22\u8fce Star&#xff01; &#xff08;2&#xff09;AIGC \/ LLM \/ AI Agent \u7b97\u6cd5\u5c97\u4e0e\u5f00\u53d1\u5c97\u6c42\u804c\u9762\u8bd5\u5185\u63a8\u5b66\u4e60\u793e\u7fa4&#xff0c;\u6db5\u76d6 AIGC\u3001LLM \u5927\u6a21\u578b\u3001AI Agent\u3001\u4f20\u7edf\u6df1\u5ea6\u5b66\u4e60\u3001\u81ea\u52a8\u9a7e\u9a76\u3001\u673a\u5668\u5b66\u4e60\u3001\u8ba1\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.wsisp.com\/helps\/108264.html\" \/>\n<meta property=\"og:site_name\" content=\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\" \/>\n<meta property=\"article:published_time\" content=\"2026-09-21T04:32:57+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260921043250-6ab0b37281cdb.png\" \/>\n<meta name=\"author\" content=\"admin\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u4f5c\u8005\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin\" \/>\n\t<meta name=\"twitter:label2\" content=\"\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 \u5206\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/108264.html\",\"url\":\"https:\/\/www.wsisp.com\/helps\/108264.html\",\"name\":\"\u6587\u6863\u667a\u80fd\u4e13\u9898\uff1a\u5c0f\u6a21\u578b\u540e\u8bad\u7ec3\u600e\u4e48\u9009\u6837\u672c\uff1fPaddleOCR-VL-1.6\u6309\u4e09\u7c7b\u5f31\u70b9\u6316\u6398 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"isPartOf\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\"},\"datePublished\":\"2026-09-21T04:32:57+00:00\",\"dateModified\":\"2026-09-21T04:32:57+00:00\",\"author\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.wsisp.com\/helps\/108264.html#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.wsisp.com\/helps\/108264.html\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/108264.html#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/www.wsisp.com\/helps\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u6587\u6863\u667a\u80fd\u4e13\u9898\uff1a\u5c0f\u6a21\u578b\u540e\u8bad\u7ec3\u600e\u4e48\u9009\u6837\u672c\uff1fPaddleOCR-VL-1.6\u6309\u4e09\u7c7b\u5f31\u70b9\u6316\u6398\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#website\",\"url\":\"https:\/\/www.wsisp.com\/helps\/\",\"name\":\"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3\",\"description\":\"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41\",\"name\":\"admin\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"contentUrl\":\"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery\",\"caption\":\"admin\"},\"sameAs\":[\"http:\/\/wp.wsisp.com\"],\"url\":\"https:\/\/www.wsisp.com\/helps\/author\/admin\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u6587\u6863\u667a\u80fd\u4e13\u9898\uff1a\u5c0f\u6a21\u578b\u540e\u8bad\u7ec3\u600e\u4e48\u9009\u6837\u672c\uff1fPaddleOCR-VL-1.6\u6309\u4e09\u7c7b\u5f31\u70b9\u6316\u6398 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.wsisp.com\/helps\/108264.html","og_locale":"zh_CN","og_type":"article","og_title":"\u6587\u6863\u667a\u80fd\u4e13\u9898\uff1a\u5c0f\u6a21\u578b\u540e\u8bad\u7ec3\u600e\u4e48\u9009\u6837\u672c\uff1fPaddleOCR-VL-1.6\u6309\u4e09\u7c7b\u5f31\u70b9\u6316\u6398 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","og_description":"&#xff08;1&#xff09;\u300a\u4e09\u5e74\u9762\u8bd5\u4e94\u5e74\u6a21\u62df\u300bAIGC \/ LLM \/ AI Agent \u7b97\u6cd5\u5de5\u7a0b\u5e08\u4e0e\u5f00\u53d1\u5de5\u7a0b\u5e08\u6c42\u804c\u9762\u8bd5\u79d8\u7c4d&#xff0c;\u72ec\u5bb6\u8d44\u6e90\u89c1 WeThinkIn\/AIGC-Interview-Book&#xff0c;\u6b22\u8fce Star&#xff01; &#xff08;2&#xff09;AIGC \/ LLM \/ AI Agent \u7b97\u6cd5\u5c97\u4e0e\u5f00\u53d1\u5c97\u6c42\u804c\u9762\u8bd5\u5185\u63a8\u5b66\u4e60\u793e\u7fa4&#xff0c;\u6db5\u76d6 AIGC\u3001LLM \u5927\u6a21\u578b\u3001AI Agent\u3001\u4f20\u7edf\u6df1\u5ea6\u5b66\u4e60\u3001\u81ea\u52a8\u9a7e\u9a76\u3001\u673a\u5668\u5b66\u4e60\u3001\u8ba1","og_url":"https:\/\/www.wsisp.com\/helps\/108264.html","og_site_name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","article_published_time":"2026-09-21T04:32:57+00:00","og_image":[{"url":"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2026\/09\/20260921043250-6ab0b37281cdb.png"}],"author":"admin","twitter_card":"summary_large_image","twitter_misc":{"\u4f5c\u8005":"admin","\u9884\u8ba1\u9605\u8bfb\u65f6\u95f4":"4 \u5206"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.wsisp.com\/helps\/108264.html","url":"https:\/\/www.wsisp.com\/helps\/108264.html","name":"\u6587\u6863\u667a\u80fd\u4e13\u9898\uff1a\u5c0f\u6a21\u578b\u540e\u8bad\u7ec3\u600e\u4e48\u9009\u6837\u672c\uff1fPaddleOCR-VL-1.6\u6309\u4e09\u7c7b\u5f31\u70b9\u6316\u6398 - \u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","isPartOf":{"@id":"https:\/\/www.wsisp.com\/helps\/#website"},"datePublished":"2026-09-21T04:32:57+00:00","dateModified":"2026-09-21T04:32:57+00:00","author":{"@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41"},"breadcrumb":{"@id":"https:\/\/www.wsisp.com\/helps\/108264.html#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.wsisp.com\/helps\/108264.html"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.wsisp.com\/helps\/108264.html#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/www.wsisp.com\/helps"},{"@type":"ListItem","position":2,"name":"\u6587\u6863\u667a\u80fd\u4e13\u9898\uff1a\u5c0f\u6a21\u578b\u540e\u8bad\u7ec3\u600e\u4e48\u9009\u6837\u672c\uff1fPaddleOCR-VL-1.6\u6309\u4e09\u7c7b\u5f31\u70b9\u6316\u6398"}]},{"@type":"WebSite","@id":"https:\/\/www.wsisp.com\/helps\/#website","url":"https:\/\/www.wsisp.com\/helps\/","name":"\u7f51\u7855\u4e92\u8054\u5e2e\u52a9\u4e2d\u5fc3","description":"\u9999\u6e2f\u670d\u52a1\u5668_\u9999\u6e2f\u4e91\u670d\u52a1\u5668\u8d44\u8baf_\u670d\u52a1\u5668\u5e2e\u52a9\u6587\u6863_\u670d\u52a1\u5668\u6559\u7a0b","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.wsisp.com\/helps\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/358e386c577a3ab51c4493330a20ad41","name":"admin","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/www.wsisp.com\/helps\/#\/schema\/person\/image\/","url":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","contentUrl":"https:\/\/gravatar.wp-china-yes.net\/avatar\/?s=96&d=mystery","caption":"admin"},"sameAs":["http:\/\/wp.wsisp.com"],"url":"https:\/\/www.wsisp.com\/helps\/author\/admin"}]}},"_links":{"self":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/108264","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/comments?post=108264"}],"version-history":[{"count":0,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/posts\/108264\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media\/108259"}],"wp:attachment":[{"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/media?parent=108264"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/categories?post=108264"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/tags?post=108264"},{"taxonomy":"topic","embeddable":true,"href":"https:\/\/www.wsisp.com\/helps\/wp-json\/wp\/v2\/topic?post=108264"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}