{"id":4778,"date":"2024-04-24T05:47:00","date_gmt":"2024-04-23T20:47:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=4778"},"modified":"2024-04-24T05:47:00","modified_gmt":"2024-04-23T20:47:00","slug":"many-shot-in-context-learning","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=4778","title":{"rendered":"Many-Shot In-Context Learning"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Many-Shot In-Context Learning\u00a0<\/strong>[57.6]<br>\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb (LLMs) \u306f\u3001\u6587\u8108\u5185\u5b66\u7fd2 (ICL) \u306b\u304a\u3044\u3066\u512a\u308c\u3066\u3044\u308b \u6211\u3005\u306f\u3001\u591a\u7a2e\u591a\u69d8\u306a\u751f\u6210\u7684\u304a\u3088\u3073\u8b58\u5225\u7684\u30bf\u30b9\u30af\u306b\u304a\u3051\u308b\u9855\u8457\u306a\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u5411\u4e0a\u3092\u89b3\u5bdf\u3059\u308b\u3002 Reinforced \u3068 Unsupervised ICL \u306f\u591a\u767a\u7684\u306a\u30b7\u30b9\u30c6\u30e0\u3067\u306f\u6975\u3081\u3066\u6709\u52b9\u3067\u3042\u308b\u3053\u3068\u304c\u308f\u304b\u3063\u305f\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2404.11018v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2404.11018v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 17 Apr 2024 02:49:26 GMT)<\/li>\n\n\n\n<li>Gemini 1.5\u306a\u3069\u3067\u53ef\u80fd\u306b\u306a\u3063\u305fMany shot\uff08500 shot\u306a\u3069\uff09\u306a\u3069\u306e\u52b9\u679c\u306e\u5206\u6790\u3002\u6027\u80fd\u304c\u4e0a\u304c\u308b\u4f8b\u304c\u591a\u3044\u304c\u300cOn some tasks (e g , code verifier, planning), we did observe slight performance deterioration beyond a certain number of shots.\u300d\u3068\u306e\u3053\u3068\u3002Reinforced ICL\u3001Unsupervised ICL\u00a0\u3068\u3044\u3046\u4eba\u9593\u3092\u4ecb\u3055\u306a\u3044ICL\u3082\u691c\u8a3c\u3057\u3066\u3044\u3066\u300cWe found that, for problem-solving domains where human-generated rationales are expensive to obtain, Reinforced and Unsupervised ICL can obtain strong performance when compared to ICL with human data.\u300d\u3068\u306e\u3053\u3068\u3002<\/li>\n\n\n\n<li>\u9577\u3044\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u306e\u5229\u70b9\u3092\u30a2\u30d4\u30fc\u30eb\u3059\u308b\u8ad6\u6587\u3002SSM\u3060\u3068\u3069\u3046\u306a\u3093\u308b\u3093\u3084\u308d\u3068\u3044\u3046\u8208\u5473\u304c\u3042\u308b\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":[200],"class_list":["post-4778","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-in-context-learning"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4778","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=4778"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4778\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4778"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4778"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4778"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}