{"id":5999,"date":"2025-01-06T05:53:00","date_gmt":"2025-01-05T20:53:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=5999"},"modified":"2025-01-06T05:53:00","modified_gmt":"2025-01-05T20:53:00","slug":"confidence-v-s-critique-a-decomposition-of-self-correction-capability-for-llms","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=5999","title":{"rendered":"Confidence v.s. Critique: A Decomposition of Self-Correction Capability for LLMs"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Confidence v.s. Critique: A Decomposition of Self-Correction Capability for LLMs\u00a0<\/strong>[34.2]<br>\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb(LLM)\u306f\u81ea\u5df1\u751f\u6210\u5fdc\u7b54\u3092\u88dc\u6b63\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u308b\u304c\u3001\u81ea\u5df1\u88dc\u6b63\u5f8c\u306e\u7cbe\u5ea6\u306e\u4f4e\u4e0b\u3082\u89b3\u5bdf\u3055\u308c\u3066\u3044\u308b\u3002 \u81ea\u5df1\u8a02\u6b63\u80fd\u529b\u306f\u3001\u81ea\u4fe1(\u56de\u7b54\u3092\u6b63\u3059\u81ea\u4fe1)\u3068\u6279\u5224(\u9593\u9055\u3063\u305f\u56de\u7b54\u3092\u6b63\u3057\u3044\u3082\u306e\u306b\u3059\u308b)\u306b\u5206\u89e3\u3057\u307e\u3059\u3002 \u6211\u3005\u306e\u6226\u7565\u306f\u4e21\u65b9\u306e\u80fd\u529b\u306b\u304a\u3044\u3066\u30d0\u30cb\u30e9SFT\u3088\u308a\u512a\u308c\u3066\u304a\u308a\u3001\u81ea\u5df1\u88dc\u6b63\u5f8c\u306e\u7cbe\u5ea6\u306f\u306f\u308b\u304b\u306b\u9ad8\u3044\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2412.19513v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2412.19513v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Fri, 27 Dec 2024 08:09:11 GMT)<\/li>\n\n\n\n<li>Confidence score\u3068Critic\u306e\u5206\u6790\u3001\u304a\u3088\u3073\u3001\u81ea\u5df1\u4fee\u6b63\u80fd\u529b\u3092\u9ad8\u3081\u308b\u624b\u6cd5\u306e\u63d0\u6848<\/li>\n\n\n\n<li>\u300cConfidence prompt\/ICL example can lead higer CL and lower CS; critique prompt\/ICL example can cause lower CL and higher CS.\u300d\uff08Confidence Level (CL) and Critique Score (CS)\uff09\u3068\u30c8\u30ec\u30fc\u30c9\u30aa\u30d5\u306e\u95a2\u4fc2\u306b\u3042\u308b\u3068\u306e\u3053\u3068\u3002<\/li>\n\n\n\n<li>\u4e21\u8005\u3092\u6539\u5584\u3059\u308b\u305f\u3081\u306b\u300cCritique Improvement Tuning (CCT), which can be divided into Confidence Level Improvement Tuning (CLT) and Critique Score Improvement Tuning (CST).\u300d\u3092\u63d0\u6848<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/Zhe-Young\/SelfCorrectDecompose\">GitHub &#8211; Zhe-Young\/SelfCorrectDecompose: Code for paper &#8220;Confidence v.s. Critique: A Decomposition of Self-Correction Capability for LLMs&#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":[84,223,356],"class_list":["post-5999","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-critic","tag-llm","tag-self-x"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5999","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=5999"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5999\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5999"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5999"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5999"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}