{"id":3853,"date":"2023-09-27T04:34:00","date_gmt":"2023-09-26T19:34:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=3853"},"modified":"2023-09-27T04:34:00","modified_gmt":"2023-09-26T19:34:00","slug":"mint-multi-turn-interaction-%e3%83%99%e3%83%b3%e3%83%81%e3%83%9e%e3%83%bc%e3%82%af","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=3853","title":{"rendered":"MINT: Multi-turn INTeraction \u30d9\u30f3\u30c1\u30de\u30fc\u30af"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>MINT: Evaluating LLMs in Multi-turn Interaction with Tools and Language Feedback\u00a0<\/strong>[78.6]<br>\u6211\u3005\u306fMINT\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3092\u5c0e\u5165\u3057\u3001\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb\u306e\u30de\u30eb\u30c1\u30bf\u30fc\u30f3\u30a4\u30f3\u30bf\u30e9\u30af\u30b7\u30e7\u30f3\u306b\u3088\u308b\u8ab2\u984c\u89e3\u6c7a\u80fd\u529b\u3092\u8a55\u4fa1\u3059\u308b\u3002 LLM\u306f\u4e00\u822c\u7684\u306b\u3001\u30c4\u30fc\u30eb\u30a4\u30f3\u30bf\u30e9\u30af\u30b7\u30e7\u30f3\u3068\u8a00\u8a9e\u30d5\u30a3\u30fc\u30c9\u30d0\u30c3\u30af\u306e\u6069\u6075\u3092\u53d7\u3051\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2309.10691v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2309.10691v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Tue, 19 Sep 2023 15:25:42 GMT)<\/li>\n\n\n\n<li>\u30de\u30eb\u30c1\u30bf\u30fc\u30f3\u30a4\u30f3\u30bf\u30e9\u30af\u30b7\u30e7\u30f3\u3092\u524d\u63d0\u3068\u3057\u305f\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u306e\u63d0\u6848\u3002\u300cBetter single-turn performance does not guarantee better multi-turn performance.\u300d\u300cSurprisingly, on LLMs we evaluated, we found supervised instruction-finetuning (SIFT) and reinforcement learning from human feedback (RLHF) generally hurt multi-turn capabilities.\u300d\u3068\u3044\u3046\u7d50\u679c\u304c\u8208\u5473\u6df1\u3044\u3002SIFT\u3084RLHF\u304c\u60aa\u5f71\u97ff\u3092\u4e0e\u3048\u308b\u306e\u306f\u672c\u5f53\u306a\u3093\u3060\u308d\u3046\u304b\u3002\u3002\uff08\u30d5\u30a3\u30fc\u30c9\u30d0\u30c3\u30af\u306bGPT-4\u3092\u7528\u3044\u3066\u3044\u308b\u5f71\u97ff\u304c\u3042\u308b\u306e\u304b\u306f\u77e5\u308a\u305f\u3044\u3068\u3053\u308d\u3002text-bison\u3067\u30d5\u30a3\u30fc\u30c9\u30d0\u30c3\u30af\u3092\u3057\u305f\u5834\u5408\u306b\u540c\u50be\u5411\u306a\u306e\u304b\u306a\u3069\u304c\u6c17\u306b\u306a\u308b\uff09<\/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":[223,517],"class_list":["post-3853","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-llm","tag-517"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/3853","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=3853"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/3853\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=3853"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=3853"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=3853"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}