{"id":5057,"date":"2024-07-05T05:21:00","date_gmt":"2024-07-04T20:21:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=5057"},"modified":"2024-07-05T05:21:00","modified_gmt":"2024-07-04T20:21:00","slug":"llms-instead-of-human-judges-a-large-scale-empirical-study-across-20-nlp-evaluation-tasks","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=5057","title":{"rendered":"LLMs instead of Human Judges? A Large Scale Empirical Study across 20 NLP Evaluation Tasks"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>LLMs instead of Human Judges? A Large Scale Empirical Study across 20 NLP Evaluation Tasks\u00a0<\/strong>[106.1]<br>\u4eba\u9593\u306e\u5224\u65ad\u306e\u4ee3\u308f\u308a\u306bLCM\u306b\u3088\u308b\u5224\u65ad\u3067NLP\u30e2\u30c7\u30eb\u3092\u8a55\u4fa1\u3059\u308b\u50be\u5411\u304c\u9ad8\u307e\u3063\u3066\u3044\u308b\u3002 \u4eba\u9593\u306e\u30c7\u30fc\u30bf\u3068\u306e\u6bd4\u8f03\u304c\u306a\u3044\u3068\u3001\u3053\u308c\u3089\u306e\u8a55\u4fa1\u306e\u6709\u52b9\u6027\u304c\u61f8\u5ff5\u3055\u308c\u308b\u3002 JUDGE-BENCH\u306f\u3001\u4eba\u9593\u306e\u30a2\u30ce\u30c6\u30fc\u30b7\u30e7\u30f3\u3092\u6301\u306420\u500b\u306eNLP\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306e\u96c6\u5408\u3067\u3042\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2406.18403v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2406.18403v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 26 Jun 2024 14:56:13 GMT)<\/li>\n\n\n\n<li>\u3088\u304f\u7528\u3044\u3089\u308c\u308b\u30c6\u30af\u30cb\u30c3\u30af\u3067\u3042\u308bLLM\u3092\u7528\u3044\u305f\u8a55\u4fa1\u306b\u95a2\u3059\u308b\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3002\u300cGPT-4o ranks first across several evaluation scenarios, but the Llama-3-70B and Mixtral-8x22B open models are relatively close, and outperform GPT-4o on some assessment types such as categorical sentence acceptability (CoLa) and graded summary quality (Summeval).\u300d\u3068\u306e\u7d50\u679c\u3002\u6709\u52b9\u6027\u306f\u30bf\u30b9\u30af\u306b\u3088\u3063\u3066\u304b\u306a\u308a\u9055\u3046\u5370\u8c61\u304c\u3042\u308b\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/dmg-illc\/JUDGE-BENCH\">GitHub &#8211; dmg-illc\/JUDGE-BENCH<\/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":[224,517,647,648],"class_list":["post-5057","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-llm-as-a-judge","tag-517","tag-647","tag-648"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5057","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=5057"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5057\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5057"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5057"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5057"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}