{"id":7594,"date":"2025-10-13T03:46:00","date_gmt":"2025-10-12T18:46:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7594"},"modified":"2025-10-12T12:50:33","modified_gmt":"2025-10-12T03:50:33","slug":"think-natively-unlocking-multilingual-reasoning-with-consistency-enhanced-reinforcement-learning","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7594","title":{"rendered":"Think Natively: Unlocking Multilingual Reasoning with Consistency-Enhanced Reinforcement Learning"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Think Natively: Unlocking Multilingual Reasoning with Consistency-Enhanced Reinforcement Learning\u00a0<\/strong>[85.7]<br>\u672c\u7a3f\u3067\u306f,\u8a00\u8a9e\u4e00\u8cab\u6027\u5831\u916c\u3068\u8a00\u8a9e\u9593\u601d\u8003\u30a2\u30e9\u30a4\u30e1\u30f3\u30c8\u5831\u916c\u306b\u3088\u3063\u3066\u8a13\u7df4\u3055\u308c\u305fM-Thinker\u3092\u63d0\u6848\u3059\u308b\u3002 M-Thinker\u306f2\u3064\u306e\u30de\u30eb\u30c1\u8a00\u8a9e\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067100%\u8fd1\u3044\u8a00\u8a9e\u4e00\u8cab\u6027\u3068\u512a\u308c\u305f\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u9054\u6210\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2510.07300v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2510.07300v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 08 Oct 2025 17:55:02 GMT)<\/li>\n\n\n\n<li>\u300cWe propose M-Thinker, which both achieves the input-output language consistency with a Language Consistency reward and enhances the multilingual reasoning performance with a Cross-lingual Thinking Alignment reward.\u300d\u3068\u5165\u529b\u30fb\u601d\u8003\u30fb\u51fa\u529b\u3067\u8a00\u8a9e\u3092\u4e00\u81f4\u3055\u305b\u308b\u624b\u6cd5\u306e\u63d0\u6848\u3002\u6027\u80fd\u5411\u4e0a\u306b\u3064\u306a\u304c\u308b\u5834\u5408\u3082\u3042\u308a\u305d\u3046\u306a\u306e\u304c\u8208\u5473\u6df1\u3044\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/XZhang00\/M-Thinker\">GitHub &#8211; XZhang00\/M-Thinker: Code for &#8220;Think Natively: Unlocking Multilingual Reasoning with Consistency-Enhanced Reinforcement Learning&#8221;.<\/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":[232,267],"class_list":["post-7594","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-lrm","tag-multilingual"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7594","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=7594"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7594\/revisions"}],"predecessor-version":[{"id":7595,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7594\/revisions\/7595"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7594"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7594"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7594"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}