{"id":4683,"date":"2024-04-18T05:38:00","date_gmt":"2024-04-17T20:38:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=4683"},"modified":"2024-04-18T05:38:00","modified_gmt":"2024-04-17T20:38:00","slug":"metal-towards-multilingual-meta-evaluation","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=4683","title":{"rendered":"METAL: Towards Multilingual Meta-Evaluation\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>METAL: Towards Multilingual Meta-Evaluation\u00a0<\/strong>[12.9]<br>\u672c\u7814\u7a76\u3067\u306f,\u591a\u8a00\u8a9e\u30b7\u30ca\u30ea\u30aa\u306b\u304a\u3044\u3066,Large Language Models (LLMs) \u3092\u8a55\u4fa1\u5bfe\u8c61\u3068\u3057\u3066\u30a8\u30f3\u30c9\u30fb\u30c4\u30fc\u30fb\u30a8\u30f3\u30c9\u8a55\u4fa1\u3092\u884c\u3046\u305f\u3081\u306e\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3092\u63d0\u6848\u3059\u308b\u3002 \u8981\u7d04\u4f5c\u696d\u306e\u305f\u3081\u306e\u6bcd\u56fd\u8a9e\u8a71\u8005\u5224\u5b9a\u3092\u542b\u308010\u8a00\u8a9e\u3092\u5bfe\u8c61\u3068\u3057\u305f\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u4f5c\u6210\u3059\u308b\u3002 GPT-3.5-Turbo, GPT-4, PaLM2\u3092\u7528\u3044\u305fLCM\u8a55\u4fa1\u5668\u306e\u6027\u80fd\u306e\u6bd4\u8f03\u3092\u884c\u3063\u305f\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2404.01667v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2404.01667v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Tue, 02 Apr 2024 06:14:54 GMT)<\/li>\n\n\n\n<li>\u30de\u30eb\u30c1\u30ea\u30f3\u30ac\u30eb\u306aLLM\u8a55\u4fa1\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306e\u63d0\u6848\u3001GPT-4\u306f\u3084\u306f\u308a\u512a\u79c0\u3002\u3060\u304c\u300cFinally, we analyze human and LLM reasoning and observe that LLMs often provide incorrect justifications for their scores, thus showing that more research is needed to be able to use LLM-based evaluators with confidence in the multilingual setting.\u300d\u30fb\u30fb\u30fb\u3002\u308f\u308a\u3068\u3088\u304f\u8a00\u308f\u308c\u3066\u3044\u308b\u3053\u3068\u3067\u306f\u3042\u308b\u30fb\u30fb\u30fb\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/hadarishav\/METAL\">hadarishav\/METAL: Code and data repo for NAACL&#8217;24 findings paper &#8220;METAL: Towards Multilingual Meta Evaluation&#8221; (github.com)<\/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":[223,267,517],"class_list":["post-4683","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-llm","tag-multilingual","tag-517"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4683","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=4683"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4683\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4683"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4683"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4683"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}