{"id":5025,"date":"2024-06-24T04:42:00","date_gmt":"2024-06-23T19:42:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=5025"},"modified":"2024-06-24T04:42:00","modified_gmt":"2024-06-23T19:42:00","slug":"datacomp-lm-in-search-of-the-next-generation-of-training-sets-for-language-models","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=5025","title":{"rendered":"DataComp-LM: In search of the next generation of training sets for language models"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>DataComp-LM: In search of the next generation of training sets for language models&nbsp;<\/strong>[193.3]<br>DataComp for Language Models (DCLM)\u306f\u3001\u5236\u5fa1\u3055\u308c\u305f\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u5b9f\u9a13\u306e\u305f\u3081\u306e\u30c6\u30b9\u30c8\u30d9\u30c3\u30c9\u3067\u3042\u308a\u3001\u8a00\u8a9e\u30e2\u30c7\u30eb\u3092\u6539\u5584\u3059\u308b\u3053\u3068\u3092\u76ee\u7684\u3068\u3057\u3066\u3044\u308b\u3002 \u6211\u3005\u306f\u3001Common Crawl\u304b\u3089\u62bd\u51fa\u3055\u308c\u305f240T\u30c8\u30fc\u30af\u30f3\u306e\u6a19\u6e96\u5316\u30b3\u30fc\u30d1\u30b9\u3001OpenLM\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306b\u57fa\u3065\u304f\u52b9\u679c\u7684\u306a\u4e8b\u524d\u5b66\u7fd2\u30ec\u30b7\u30d4\u300153\u306e\u4e0b\u6d41\u8a55\u4fa1\u30b9\u30a4\u30fc\u30c8\u3092\u63d0\u4f9b\u3059\u308b\u3002 DCLM\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u306e\u53c2\u52a0\u8005\u306f\u3001412M\u304b\u30897B\u30d1\u30e9\u30e1\u30fc\u30bf\u306e\u30e2\u30c7\u30eb\u30b9\u30b1\u30fc\u30eb\u3067\u306e\u91cd\u8907\u3001\u30d5\u30a3\u30eb\u30bf\u30ea\u30f3\u30b0\u3001\u30c7\u30fc\u30bf\u6df7\u5408\u306a\u3069\u306e\u30c7\u30fc\u30bf\u30ad\u30e5\u30ec\u30fc\u30b7\u30e7\u30f3\u6226\u7565\u3092\u5b9f\u9a13\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2406.11794v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2406.11794v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Mon, 17 Jun 2024 17:42:57 GMT)<\/li>\n\n\n\n<li>\u8a00\u8a9e\u30e2\u30c7\u30eb\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6642\u306e\u30c7\u30fc\u30bf\u30ad\u30e5\u30ec\u30fc\u30b7\u30e7\u30f3\u306e\u305f\u3081\u306e\u30d9\u30f3\u30c1\u30de\u30fc\u30afDataComp for Language Models (DCLM)\u306e\u63d0\u6848\u3002\u91cd\u8981\u306a\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067\u5c0f\u3055\u3081\u306e\u30c8\u30e9\u30c3\u30af\u3082\u7528\u610f\u3055\u308c\u3066\u3044\u308b\uff08\u6700\u5c0f\u30c8\u30e9\u30c3\u30af\u306f412M\u30d1\u30e9\u30e1\u30fc\u30bf\u30018.2B\u5b66\u7fd2\u7528\u30c8\u30fc\u30af\u30f3\uff08\u5143\u30c7\u30fc\u30bf469B\uff09\u3001\u5b66\u7fd2\u7528\u306e\u8a08\u7b97\u91cf\u306f2.0e19FLOPs\u3001H100\u63db\u7b97\u306726\u6642\u9593\uff09\u304c\u3001\u305d\u308c\u306b\u3057\u3066\u3082\u53c2\u52a0\u3059\u308b\u306b\u3082\u7d50\u69cb\u306a\u74b0\u5883\u304c\u5fc5\u8981\u305d\u3046\u3002\u3002\u3002<\/li>\n\n\n\n<li>\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u30b5\u30a4\u30c8\u306f<a href=\"https:\/\/www.datacomp.ai\/dclm\/\">DataComp<\/a><\/li>\n<\/ul>\n\n\n\n<p>\u95a2\u9023\u3059\u308b\u3082\u306e\u3068\u3057\u3066\u4e0b\u8a18\u8ad6\u6587\u3082\u53c2\u8003\u306b\u306a\u308b\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Data-Centric AI in the Age of Large Language Models\u00a0<\/strong>[51.2]<br>\u672c\u7a3f\u3067\u306f,\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb(LLM)\u306b\u7740\u76ee\u3057\u305f,AI\u7814\u7a76\u306b\u304a\u3051\u308b\u30c7\u30fc\u30bf\u4e2d\u5fc3\u306e\u8996\u70b9\u3092\u63d0\u6848\u3059\u308b\u3002 \u672c\u7814\u7a76\u3067\u306f,LLM\u306e\u767a\u9054\u6bb5\u968e(\u4e8b\u524d\u5b66\u7fd2\u3084\u5fae\u8abf\u6574\u306a\u3069)\u3084\u63a8\u8ad6\u6bb5\u968e(\u6587\u8108\u5185\u5b66\u7fd2\u306a\u3069)\u306b\u304a\u3044\u3066,\u30c7\u30fc\u30bf\u304c\u6709\u52b9\u3067\u3042\u308b\u3053\u3068\u3092\u793a\u3059\u3002 \u30c7\u30fc\u30bf\u3092\u4e2d\u5fc3\u3068\u3057\u305f4\u3064\u306e\u30b7\u30ca\u30ea\u30aa\u3092\u7279\u5b9a\u3057\u3001\u30c7\u30fc\u30bf\u4e2d\u5fc3\u306e\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3068\u30c7\u30fc\u30bf\u30ad\u30e5\u30ec\u30fc\u30b7\u30e7\u30f3\u3001\u30c7\u30fc\u30bf\u5c5e\u6027\u3001\u77e5\u8b58\u4f1d\u9054\u3001\u63a8\u8ad6\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u5316\u3092\u30ab\u30d0\u30fc\u3057\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2406.14473v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2406.14473v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 20 Jun 2024 16:34:07 GMT)<\/li>\n\n\n\n<li>LLM\u306e\u6642\u4ee3\u306b\u304a\u3044\u3066\u3082\u30c7\u30fc\u30bf\u306f\u91cd\u8981\u3001DataCOMP\u306b\u3064\u3044\u3066\u306f\u300cDataComp is a more suitable starting point due to its scale and the promising initial findings.\u300d\u3068\u8a18\u8f09\u3002<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>\u95a2\u9023\u3059\u308b\u3082\u306e\u3068\u3057\u3066\u4e0b\u8a18\u8ad6\u6587\u3082\u53c2\u8003\u306b\u306a\u308b\u3002<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[101,102,223,491],"class_list":["post-5025","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-dataset-condensation","tag-dataset-distillation","tag-llm","tag-491"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5025","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=5025"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5025\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5025"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5025"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5025"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}