{"id":6197,"date":"2025-02-10T04:53:00","date_gmt":"2025-02-09T19:53:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=6197"},"modified":"2025-02-10T04:53:00","modified_gmt":"2025-02-09T19:53:00","slug":"s1-simple-test-time-scaling","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=6197","title":{"rendered":"s1: Simple test-time scaling"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>s1: Simple test-time scaling&nbsp;<\/strong>[148.4]<br>\u30c6\u30b9\u30c8\u6642\u9593\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u306f\u3001\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u6539\u5584\u3059\u308b\u305f\u3081\u306b\u4f59\u5206\u306a\u30c6\u30b9\u30c8\u6642\u9593\u8a08\u7b97\u3092\u4f7f\u7528\u3059\u308b\u8a00\u8a9e\u30e2\u30c7\u30ea\u30f3\u30b0\u306b\u5bfe\u3059\u308b\u3001\u6709\u671b\u306a\u65b0\u3057\u3044\u30a2\u30d7\u30ed\u30fc\u30c1\u3067\u3042\u308b\u3002 \u30c6\u30b9\u30c8\u30bf\u30a4\u30e0\u306e\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u3068\u5f37\u529b\u306a\u63a8\u8ad6\u6027\u80fd\u3092\u5b9f\u73fe\u3059\u308b\u305f\u3081\u306e\u6700\u3082\u30b7\u30f3\u30d7\u30eb\u306a\u30a2\u30d7\u30ed\u30fc\u30c1\u3092\u63a2\u3057\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2501.19393v2\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2501.19393v2\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Mon, 03 Feb 2025 16:31:30 GMT)<\/li>\n\n\n\n<li>\u300cWe show that SFT on only 1,000 examples suffices to build a competitive reasoning model matching o1-preview and produces a model that lies on the pareto frontier \u300d\u3068\u3044\u3046\u5831\u544a\u3002\u300cFirst, we curate a small dataset s1K of 1,000 questions paired with reasoning traces relying on three criteria we validate through ablations: difficulty, diversity, and quality. Second, we develop budget forcing to control test-time compute by forcefully terminating the model\u2019s thinking process or lengthening it by appending \u201cWait\u201d multiple times to the model\u2019s generation when it tries to end.\u300d\u3068Wait\u3092\u4f7f\u3046\u306e\u304c\u7279\u5fb4\u7684\uff08<a href=\"https:\/\/devneko.jp\/wordpress\/?p=3917\">Think before you speak: Training Language Models With Pause Tokens \u2013 arXiv\u6700\u65b0\u8ad6\u6587\u306e\u7d39\u4ecb<\/a>\u3092\u601d\u3044\u51fa\u3059\uff09<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/simplescaling\/s1\">GitHub &#8211; simplescaling\/s1: s1: Simple test-time scaling<\/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":[232,366],"class_list":["post-6197","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-lrm","tag-slow-thinking"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6197","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=6197"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6197\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=6197"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=6197"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=6197"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}