{"id":5169,"date":"2024-07-22T04:21:00","date_gmt":"2024-07-21T19:21:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=5169"},"modified":"2024-07-22T04:21:00","modified_gmt":"2024-07-21T19:21:00","slug":"arena-learning-build-data-flywheel-for-llms-post-training-via-simulated-chatbot-arena","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=5169","title":{"rendered":"Arena Learning: Build Data Flywheel for LLMs Post-training via Simulated Chatbot Arena"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Arena Learning: Build Data Flywheel for LLMs Post-training via Simulated Chatbot Arena&nbsp;<\/strong>[126.7]<br>AI\u99c6\u52d5\u306e\u30a2\u30ce\u30c6\u30fc\u30b7\u30e7\u30f3\u3092\u4f7f\u3063\u3066\u30a2\u30ea\u30fc\u30ca\u306e\u6226\u3044\u3092\u30b7\u30df\u30e5\u30ec\u30fc\u30c8\u3059\u308b\u305f\u3081\u306b\u8a2d\u8a08\u3055\u308c\u305f\u3001\u9769\u65b0\u7684\u306a\u30aa\u30d5\u30e9\u30a4\u30f3\u6226\u7565\u3067\u3042\u308bArena Learning\u3092\u7d39\u4ecb\u3057\u307e\u3059\u3002 Arena Learning\u306f\u3001\u30aa\u30d5\u30e9\u30a4\u30f3\u30b7\u30df\u30e5\u30ec\u30fc\u30b7\u30e7\u30f3\u3068\u30aa\u30f3\u30e9\u30a4\u30f3\u30b3\u30f3\u30da\u30c6\u30a3\u30b7\u30e7\u30f3\u306e\u6b63\u78ba\u306a\u8a55\u4fa1\u3068\u4e00\u8cab\u6027\u3092\u4fdd\u8a3c\u3059\u308b\u3002 \u30bf\u30fc\u30b2\u30c3\u30c8\u30e2\u30c7\u30eb\u3067\u3042\u308bWizardLM-$beta$\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3059\u308b\u305f\u3081\u306bArena Learning\u3092\u9069\u7528\u3057\u3001\u5927\u5e45\u306a\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u5411\u4e0a\u3092\u793a\u3057\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2407.10627v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2407.10627v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Mon, 15 Jul 2024 11:26:07 GMT)<\/li>\n\n\n\n<li>ChatBot Arena\u306e\u8a55\u4fa1\u3092\u518d\u73fe\u3059\u308b\u74b0\u5883\u3092AI\u3067\u5b9f\u73fe\u3001\u300cThis paper introduces Arena Learning, a simulated offline chatbot arena that utilizes AI LLMs to bypass the manual and time-intensive cost typically associated with preparing the arena battle data, while preserving the core advantages of the arena-based evaluation and training.\u300d\u3001\u300cFurthermore, the model trained iteratively on synthetic data generated by Arena Learning exhibits significant performance improvements using various training strategies.\u300d\u3068\u306e\u3053\u3068\u3002<\/li>\n\n\n\n<li>\u81ea\u5df1\u6539\u5584\u3001\u5408\u6210\u30c7\u30fc\u30bf\u6d3b\u7528\u306e\u6587\u8108\u3067\u3082\u975e\u5e38\u306b\u8208\u5473\u6df1\u3044\u3002<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>AgentInstruct: Toward Generative Teaching with Agentic Flows&nbsp;<\/strong>[12.2]<br>\u6211\u3005\u306f\u3001\u30dd\u30b9\u30c8\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u306b\u5408\u6210\u30c7\u30fc\u30bf\u3092\u4f7f\u3046\u3053\u3068\u3001\u7279\u306b\u3001\u4ed6\u306e\u30e2\u30c7\u30eb\u306b\u65b0\u3057\u3044\u30b9\u30ad\u30eb\u3084\u632f\u308b\u821e\u3044\u3092\u6559\u3048\u308b\u305f\u3081\u306b\u3001\u5f37\u529b\u306a\u30e2\u30c7\u30eb\u3067\u30c7\u30fc\u30bf\u3092\u4f5c\u6210\u3059\u308b\u3053\u3068\u306b\u91cd\u70b9\u3092\u7f6e\u3044\u3066\u3044\u308b\u3002 \u672c\u7a3f\u3067\u306f,\u591a\u7a2e\u591a\u69d8\u306a\u9ad8\u54c1\u8cea\u306a\u5408\u6210\u30c7\u30fc\u30bf\u3092\u81ea\u52d5\u751f\u6210\u3059\u308b\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3067\u3042\u308bAgentInstruct\u3092\u7d39\u4ecb\u3059\u308b\u3002 \u30c6\u30ad\u30b9\u30c8\u7de8\u96c6,\u5275\u9020\u7684\u57f7\u7b46,\u30c4\u30fc\u30eb\u4f7f\u7528,\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0,\u7406\u89e3\u306e\u7406\u89e3\u306a\u3069,\u3055\u307e\u3056\u307e\u306a\u30b9\u30ad\u30eb\u3092\u5b66\u7fd2\u3059\u308b\u305f\u3081\u306e,2500\u4e07\u5bfe\u306e\u30dd\u30b9\u30c8\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u4f5c\u6210\u3059\u308b\u3053\u3068\u3067,AgentInstruct\u306e\u6709\u7528\u6027\u3092\u5b9f\u8a3c\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2407.03502v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2407.03502v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Wed, 03 Jul 2024 21:01:12 GMT)<\/li>\n\n\n\n<li>\u4e0a\u8a18\u3068\u306f\u7570\u306a\u308aAgentic\u306a\u30c7\u30fc\u30bf\u5408\u6210\u30a2\u30d7\u30ed\u30fc\u30c1\u3082\u6709\u671b\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":[223,390,576],"class_list":["post-5169","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-llm","tag-synthetic-data","tag-576"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5169","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=5169"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5169\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5169"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5169"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5169"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}