{"id":4953,"date":"2024-06-07T04:34:00","date_gmt":"2024-06-06T19:34:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=4953"},"modified":"2024-06-07T04:34:00","modified_gmt":"2024-06-06T19:34:00","slug":"the-fine-tuning-paradox-boosting-translation-quality-without-sacrificing-llm-abilities","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=4953","title":{"rendered":"The Fine-Tuning Paradox: Boosting Translation Quality Without Sacrificing LLM Abilities\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>The Fine-Tuning Paradox: Boosting Translation Quality Without Sacrificing LLM Abilities\u00a0<\/strong>[18.2]<br>\u6a5f\u68b0\u7ffb\u8a33\u306e\u305f\u3081\u306e\u5fae\u8abf\u6574\u5927\u578b\u8a00\u8a9e\u30e2\u30c7\u30eb(LLM)\u306f\u3001\u5168\u4f53\u7684\u306a\u7ffb\u8a33\u54c1\u8cea\u304c\u6539\u5584\u3055\u308c\u3066\u3044\u308b\u3002 \u30e2\u30c7\u30eb\u30b5\u30a4\u30ba\u306f70\u5104\u304b\u308965\u5104\u307e\u3067\u306e\u7bc4\u56f2\u3067,LLaMA\u304a\u3088\u3073\u30d5\u30a1\u30eb\u30b3\u30f3\u7cfb\u306e\u30e2\u30c7\u30eb\u306b\u5bfe\u3057\u3066\u5e83\u7bc4\u306a\u7ffb\u8a33\u8a55\u4fa1\u3092\u884c\u3046\u3002 \u30d5\u30a9\u30fc\u30de\u30eb\u306a\u30b9\u30c6\u30a2\u30ea\u30f3\u30b0\u3092\u884c\u3046\u80fd\u529b\u306e\u4f4e\u4e0b\u3001\u6570\u30b7\u30e7\u30c3\u30c8\u306e\u4f8b\u306b\u3088\u308b\u6280\u8853\u7684\u7ffb\u8a33\u306e\u4f5c\u6210\u3001\u6587\u66f8\u30ec\u30d9\u30eb\u306e\u7ffb\u8a33\u3092\u884c\u3046\u80fd\u529b\u306e\u4f4e\u4e0b\u3092\u89b3\u5bdf\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2405.20089v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2405.20089v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 30 May 2024 14:25:56 GMT)<\/li>\n\n\n\n<li>\u300cOur results show that while fine-tuning improves the general translation quality of LLMs, several abilities degrade.\u300d\u306b\u5bfe\u3057\u3066\u300cWe show that incorporating a mix of monolingual and parallel data during fine-tuning can preserve abilities of LLMs.\u300d\u3068\u306e\u3053\u3068<\/li>\n\n\n\n<li>\u7ffb\u8a33\u7279\u5316\u306b\u3057\u305f\u3089\u4e0b\u304c\u308b\u80fd\u529b\u3082\u3042\u308b\u306e\u306f\u5f53\u7136\u3060\u3068\u601d\u3046\u304c\u3001\u5358\u8a00\u8a9e\u306a\u30c7\u30fc\u30bf\u3092\u5165\u308c\u308b\u306e\u307f\u3067\u80fd\u529b\u306e\u7dad\u6301\u304c\u53ef\u80fd\u3068\u3044\u3046\u306e\u306b\u9a5a\u304d\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":[223,602],"class_list":["post-4953","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-llm","tag-602"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4953","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=4953"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4953\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4953"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4953"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4953"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}