{"id":5872,"date":"2024-12-09T03:21:00","date_gmt":"2024-12-08T18:21:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=5872"},"modified":"2024-12-09T03:21:00","modified_gmt":"2024-12-08T18:21:00","slug":"self-improvement-in-language-models-the-sharpening-mechanism","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=5872","title":{"rendered":"Self-Improvement in Language Models: The Sharpening Mechanism\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Self-Improvement in Language Models: The Sharpening Mechanism\u00a0<\/strong>[70.9]<br>\u8a00\u8a9e\u30e2\u30c7\u30ea\u30f3\u30b0\u306b\u304a\u3051\u308b\u6700\u8fd1\u306e\u7814\u7a76\u306f\u3001\u8a00\u8a9e\u30e2\u30c7\u30eb\u304c\u5916\u90e8\u304b\u3089\u306e\u30d5\u30a3\u30fc\u30c9\u30d0\u30c3\u30af\u306a\u3057\u306b\u3088\u308a\u9ad8\u3044\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u9054\u6210\u3059\u308b\u305f\u3081\u306b\u3001\u8a00\u8a9e\u4e16\u4ee3\u3092\u8a55\u4fa1\u3057\u3001\u6d17\u7df4\u3059\u308b\u81ea\u5df1\u6539\u5584\u306e\u53ef\u80fd\u6027\u3092\u9ad8\u3081\u3066\u3044\u308b\u3002 \u6211\u3005\u306f\u3001\u30ec\u30f3\u30ba\u3092\u901a\u3057\u3066\u81ea\u5df1\u6539\u5584\u306e\u80fd\u529b\u306b\u3064\u3044\u3066\u3001\u65b0\u305f\u306a\u8996\u70b9\u3092\u63d0\u4f9b\u3059\u308b\u3002 \u8a00\u8a9e\u30e2\u30c7\u30eb\u306f\u3001\u6b63\u3057\u3044\u5fdc\u7b54\u3092\u751f\u6210\u3059\u308b\u5834\u5408\u3088\u308a\u3082\u3001\u5fdc\u7b54\u54c1\u8cea\u306e\u691c\u8a3c\u304c\u512a\u308c\u3066\u3044\u308b\u3068\u3044\u3046\u89b3\u5bdf\u306b\u611f\u9298\u3092\u53d7\u3051\u3066\u3001\u5f8c\u5b66\u7fd2\u306b\u304a\u3044\u3066\u3001\u30e2\u30c7\u30eb\u81ea\u4f53\u3092\u691c\u8a3c\u5bfe\u8c61\u3068\u3057\u3066\u3001\u81ea\u5df1\u6539\u5584\u3092\u5f62\u5f0f\u5316\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2412.01951v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2412.01951v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Mon, 02 Dec 2024 20:24:17 GMT)<\/li>\n\n\n\n<li>\u300cMotivated by the observation that language models are often better at verifying response quality than they are at generating correct responses, we formalize self-improvement as using the model itself as a verifier during post-training in order to \u201csharpen\u201d the model to one placing large mass on high-quality sequences, thereby amortizing the expensive inference-time computation of generating good sequences.\u300d\u3068\u3044\u3046\u7814\u7a76<\/li>\n\n\n\n<li><a href=\"https:\/\/devneko.jp\/wordpress\/?tag=critic\">\u6700\u8fd1\u3088\u304f\u898b\u308b\u30ad\u30fc\u30ef\u30fc\u30c9critic \u2013 arXiv\u6700\u65b0\u8ad6\u6587\u306e\u7d39\u4ecb<\/a>\u306b\u3082\u95a2\u9023\u3059\u308b\u9762\u767d\u3044\u7814\u7a76<\/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":[84,356],"class_list":["post-5872","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-critic","tag-self-x"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5872","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=5872"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5872\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5872"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5872"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5872"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}