{"id":5467,"date":"2024-09-16T04:06:00","date_gmt":"2024-09-15T19:06:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=5467"},"modified":"2024-09-16T04:06:00","modified_gmt":"2024-09-15T19:06:00","slug":"synthetic-continued-pretraining","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=5467","title":{"rendered":"SYNTHETIC CONTINUED PRETRAINING"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Synthetic continued pretraining\u00a0<\/strong>[29.7]<br>\u4e0e\u3048\u3089\u308c\u305f\u4e8b\u5b9f\u3092\u5b66\u3076\u305f\u3081\u306b\u306f\u3001\u30e2\u30c7\u30eb\u306f\u6570\u767e\u304b\u3089\u6570\u5343\u306e\u591a\u69d8\u306a\u8868\u73fe\u3067\u8a13\u7df4\u3055\u308c\u306a\u3051\u308c\u3070\u306a\u3089\u306a\u3044\u3002 \u672c\u7814\u7a76\u3067\u306f,\u3088\u308a\u5b66\u7fd2\u3057\u3084\u3059\u3044\u5927\u898f\u6a21\u30b3\u30fc\u30d1\u30b9\u3092\u5408\u6210\u3059\u308b\u305f\u3081\u306e\u5408\u6210\u7d99\u7d9a\u4e8b\u524d\u5b66\u7fd2\u3092\u63d0\u6848\u3059\u308b\u3002 \u5408\u6210\u30c7\u30fc\u30bf\u62e1\u5f35\u30a2\u30eb\u30b4\u30ea\u30ba\u30e0\u3067\u3042\u308bEntiGraph\u3067\u3053\u306e\u63d0\u6848\u3092\u30a4\u30f3\u30b9\u30bf\u30f3\u30b9\u5316\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2409.07431v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2409.07431v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 11 Sep 2024 17:21:59 GMT)<\/li>\n\n\n\n<li>\u30ca\u30ec\u30c3\u30b8\u30b0\u30e9\u30d5\u3092\u4ecb\u3057\u3066\u5408\u6210\u30c7\u30fc\u30bf\u3092\u69cb\u7bc9\u3059\u308bEntiGraph\u306e\u63d0\u6848\u3002\u300cSynthetic continued pretraining with EntiGraph demonstrates consistent scaling in downstream closed-book QA performance up to a 600M token synthetic corpus, whereas baselines such as continued pretraining on the small corpus or synthetic paraphrases show no improvement or asymptote early.\u300d\u3068\u306e\u3053\u3068\u3067\u6709\u52b9\u6027\u3092\u78ba\u8a8d<\/li>\n\n\n\n<li>\u62bd\u8c61\u7684\u306a\u300c\u77e5\u8b58\u300d\u3092\u4ecb\u3057\u305f\u307b\u3046\u304c\u3001\u8868\u73fe\u306e\u5909\u63db\u3088\u308a\u3082\u826f\u3044\uff08\u5b66\u7fd2\u306b\u5229\u7528\u53ef\u80fd\u306a\uff09\u60c5\u5831\u3092\u63d0\u4f9b\u3067\u304d\u308b\u3068\u3044\u3046\u89e3\u91c8\u3067\u826f\u3044\u306e\u3060\u308d\u3046\u304b\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":[42,223,356],"class_list":["post-5467","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-autonomous-agent","tag-llm","tag-self-x"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5467","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=5467"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5467\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5467"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5467"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5467"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}