{"id":8318,"date":"2026-03-10T05:05:00","date_gmt":"2026-03-09T20:05:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=8318"},"modified":"2026-03-07T17:07:41","modified_gmt":"2026-03-07T08:07:41","slug":"firered-ocr-technical-report","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=8318","title":{"rendered":"FireRed-OCR Technical Report\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>FireRed-OCR Technical Report\u00a0<\/strong>[30.0]<br>\u672c\u7a3f\u3067\u306f,\u6c4e\u7528VLM\u3092\u5c02\u9580\u5bb6\u3092\u89e3\u6790\u3059\u308b\u30d4\u30af\u30bb\u30eb\u7cbe\u5ea6\u69cb\u9020\u6587\u66f8\u306b\u5909\u63db\u3059\u308b\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30afFireRed-OCR\u3092\u7d39\u4ecb\u3059\u308b\u3002 \u9ad8\u54c1\u8cea\u306a\u69cb\u9020\u5316\u30c7\u30fc\u30bf\u306e\u4e0d\u8db3\u306b\u5bfe\u51e6\u3059\u308b\u305f\u3081,Geometry + Semantics&#8217;s Data Factory\u3092\u69cb\u7bc9\u3057\u305f\u3002 \u672c\u7a3f\u3067\u306f,\u753b\u7d20\u30ec\u30d9\u30eb\u306e\u8a8d\u8b58\u304b\u3089\u8ad6\u7406\u69cb\u9020\u751f\u6210\u3078\u30e2\u30c7\u30eb\u3092\u5c0e\u304f\u4e09\u6bb5\u968e\u30d7\u30ed\u30b0\u30ec\u30c3\u30b7\u30d6\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6226\u7565\u3092\u63d0\u6848\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2603.01840v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2603.01840v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Mon, 02 Mar 2026 13:19:23 GMT)<\/li>\n\n\n\n<li>OCR\u306e\u6539\u5584\u306e\u767a\u8868\u304c\u7d9a\u304f\u3002\u672c\u8ad6\u6587\u3067\u306f\u300cThis curriculum includes: (1) Multi-task Pre-alignment to ground the model\u2019s understanding of document structure; (2) Specialized SFT for standardizing full- image Markdown output; and (3) Format-Constrained Group Relative Policy Optimization (GRPO), which utilizes reinforcement learning to enforce strict syntactic validity and structural integrity (e g , table closure, formula syntax).\u00a0\u300d\u3068\u3044\u3046\u30a2\u30d7\u30ed\u30fc\u30c1\u3067MLLM\u3092\u5f37\u5316\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/FireRedTeam\/FireRed-OCR\">GitHub &#8211; FireRedTeam\/FireRed-OCR \u00b7 GitHub<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[285],"class_list":["post-8318","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-ocr"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8318","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=8318"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8318\/revisions"}],"predecessor-version":[{"id":8319,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8318\/revisions\/8319"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=8318"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=8318"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=8318"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}