{"id":6374,"date":"2025-03-10T04:23:00","date_gmt":"2025-03-09T19:23:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=6374"},"modified":"2025-03-10T04:23:00","modified_gmt":"2025-03-09T19:23:00","slug":"start-self-taught-reasoner-with-tools","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=6374","title":{"rendered":"START: Self-taught Reasoner with Tools\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>START: Self-taught Reasoner with Tools\u00a0<\/strong>[51.4]<br>\u30c4\u30fc\u30eb\u7d71\u5408\u9577\u30c1\u30a7\u30fc\u30f3\u30fb\u30aa\u30d6\u30fb\u30b7\u30fc\u30af\u30ec\u30c3\u30c8(CoT)\u63a8\u8ad6LSM\u3067\u3042\u308bSTART(Self-Taught Reasoner with Tools)\u3092\u7d39\u4ecb\u3059\u308b\u3002 START\u306f\u8907\u96d1\u306a\u8a08\u7b97\u3001\u81ea\u5df1\u30c1\u30a7\u30c3\u30af\u3001\u591a\u69d8\u306a\u65b9\u6cd5\u306e\u63a2\u7d22\u3001\u305d\u3057\u3066\u81ea\u5df1\u8001\u5316\u3092\u884c\u3046\u3053\u3068\u304c\u3067\u304d\u308b\u3002 \u57fa\u790e\u3068\u306a\u308bQwQ-32B\u3092\u8457\u3057\u304f\u4e0a\u56de\u308a\u3001\u6700\u5148\u7aef\u306e\u30aa\u30fc\u30d7\u30f3\u30a6\u30a7\u30a4\u30c8\u30e2\u30c7\u30ebR1-Distill-Qwen-32B\u306b\u5339\u6575\u3059\u308b\u6027\u80fd\u3092\u9054\u6210\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2503.04625v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2503.04625v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 06 Mar 2025 17:11:51 GMT)<\/li>\n\n\n\n<li>\u30c4\u30fc\u30eb\u7d71\u5408\u578b\u306eCoT\u3092\u884c\u3046START (Self-Taught Reasoner with Tools)\u306e\u63d0\u6848\u3001\u300cHint-infer: code\/math data is processed by QwQ, with responses truncated at predefined terminators. Context-aware hints from a Hint-Library are injected at truncation points (including endpoints), and QwQ resumes inference using a code interpreter for Python execution feedback.\u300d\u3068\u300cb) Hint-RFT: Hint-infer outputs undergo rule-based scoring, filtering, and content modification to create Dseed .\u300d\u306e\uff12\u3064\u304c\u30ad\u30fc\u30dd\u30a4\u30f3\u30c8\u3002\u30eb\u30fc\u30eb\u30fb\u30c6\u30f3\u30d7\u30ec\u30fc\u30c8\u3092\u3046\u307e\u304f\u7d71\u5408\u3057\u3066\u3044\u3063\u3066\u3044\u308b\u5370\u8c61\u3067\u3001\u3053\u306e\u624b\u306e\u5de5\u592b\u306f\u8272\u3005\u3042\u308a\u5f97\u305d\u3046\u3002<\/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":[59,356],"class_list":["post-6374","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-chain-of-thought","tag-self-x"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6374","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=6374"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6374\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=6374"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=6374"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=6374"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}