{"id":4326,"date":"2024-01-22T06:29:00","date_gmt":"2024-01-21T21:29:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=4326"},"modified":"2024-01-22T06:29:00","modified_gmt":"2024-01-21T21:29:00","slug":"tuning-language-models-by-proxy","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=4326","title":{"rendered":"Tuning Language Models by Proxy\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Tuning Language Models by Proxy\u00a0<\/strong>[117.1]<br>\u30d7\u30ed\u30ad\u30b7\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306f\u3001\u30d6\u30e9\u30c3\u30af\u30dc\u30c3\u30af\u30b9LM\u4e0a\u3067\u52d5\u4f5c\u3059\u308b\u8efd\u91cf\u306a\u5fa9\u53f7\u6642\u9593\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u3067\u3042\u308b\u3002 \u6211\u3005\u306e\u7814\u7a76\u306f\u3001\u5c0f\u3055\u304f\u8abf\u6574\u3055\u308c\u305fLM\u3092\u4f7f\u7528\u3057\u3066\u3001\u5927\u898f\u6a21\u3067\u6f5c\u5728\u7684\u306b\u30d7\u30ed\u30d7\u30e9\u30a4\u30a8\u30bf\u30ea\u306aLM\u3092\u52b9\u7387\u7684\u306b\u30ab\u30b9\u30bf\u30de\u30a4\u30ba\u3059\u308b\u53ef\u80fd\u6027\u3092\u5b9f\u8a3c\u3057\u3066\u3044\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2401.08565v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2401.08565v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Tue, 16 Jan 2024 18:49:55 GMT)<\/li>\n\n\n\n<li>\uff08\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3057\u305f\uff09\u5c0f\u898f\u6a21LM\u3092\u7528\u3044\u3066\u5927\u898f\u6a21LM\u306e\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u3092\u884c\u3048\u308b\u3068\u3044\u3046\u5831\u544a\u3002untuned\u306a\u30e2\u30c7\u30eb\u3068tuned\u306a\u30e2\u30c7\u30eb\u306e\u5dee\u3092\u898b\u308b\u30a2\u30d7\u30ed\u30fc\u30c1\u3002\u300cwhen we apply proxy-tuning to LLAMA2-70B using proxies of only 7B size, we can close 88% of the gap between LLAMA2-70B and its truly-tuned CHAT version\u300d\u3068\u306e\u3053\u3068\u3002<\/li>\n\n\n\n<li>\u300cproxy-tuning addresses an important issue about how to efficiently adapt proprietary models to diverse use cases.\u300d\u3068\u3042\u308b\u901a\u308a\u30d3\u30b8\u30cd\u30b9\u3067\u306e\u30e6\u30fc\u30b9\u30b1\u30fc\u30b9\u306f\u591a\u305d\u3046\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":[150,223],"class_list":["post-4326","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-fine-tuning","tag-llm"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4326","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=4326"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4326\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4326"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4326"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4326"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}