{"id":5420,"date":"2024-09-09T06:58:00","date_gmt":"2024-09-08T21:58:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=5420"},"modified":"2024-09-09T06:58:00","modified_gmt":"2024-09-08T21:58:00","slug":"olmoe-reflection-70b-tanuki-8x8b-8b","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=5420","title":{"rendered":"OLMoE, Reflection-70B, Tanuki-8\u00d78B, 8B"},"content":{"rendered":"\n<p>OSS\u307e\u305f\u306f\u516c\u958b\u30e2\u30c7\u30eb\u306eLLM\u958b\u767a\u306f\u5f15\u304d\u7d9a\u304d\u76db\u3093\u3067\u3042\u308b\u3002<\/p>\n\n\n\n<p>OLMoE-1B-7B\u306fMoE\u69cb\u6210\u306e\u5c0f\u898f\u6a21\u30fb\u9ad8\u6027\u80fd\u30e2\u30c7\u30eb\uff08\u30ea\u30dd\u30b8\u30c8\u30ea\uff1a<a href=\"https:\/\/huggingface.co\/allenai\/OLMoE-1B-7B-0924\">allenai\/OLMoE-1B-7B-0924 \u00b7 Hugging Face<\/a>\u3001Apache-2\u30e9\u30a4\u30bb\u30f3\u30b9\uff09\u3067\u8ad6\u6587\u3082\u516c\u8868\u3055\u308c\u3066\u3044\u308b\u3002<\/p>\n\n\n\n<p>Reflection-70B\u306f\u73fe\u6642\u70b9\u3067\u6700\u9ad8\u30ec\u30d9\u30eb\u306e\u6027\u80fd\u3092\u3082\u3064\u516c\u958b\u30e2\u30c7\u30eb\uff08\u30ea\u30dd\u30b8\u30c8\u30ea\uff1a<a href=\"https:\/\/huggingface.co\/mattshumer\/Reflection-Llama-3.1-70B\">mattshumer\/Reflection-Llama-3.1-70B \u00b7 Hugging Face<\/a>\u3001Llama 3.1\u30e9\u30a4\u30bb\u30f3\u30b9\uff09\u3001\u73fe\u72b6\u8a73\u7d30\u306f\u516c\u958b\u3055\u308c\u3066\u3044\u306a\u3055\u305d\u3046\u3060\u304c\u3001\u300cReflection Llama-3.1 70B is (currently) the world&#8217;s top open-source LLM, trained with a new technique called Reflection-Tuning that teaches a LLM to detect mistakes in its reasoning and correct course. The model was trained on synthetic data generated by\u00a0<a href=\"https:\/\/glaive.ai\/\">Glaive<\/a>.\u300d\u3068\u306e\u8a18\u8f09\u304c\u3042\u308b\u3002<\/p>\n\n\n\n<p>\u65e5\u672c\u3067\u3082<a href=\"https:\/\/weblab.t.u-tokyo.ac.jp\/2024-08-30\/\">\u6771\u4eac\u5927\u5b66\u677e\u5c3e\u30fb\u5ca9\u6fa4\u7814\u7a76\u5ba4 GENIAC\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306b\u304a\u3044\u3066\u3001\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb\u300cTanuki-8\u00d78B\u300d\u3092\u958b\u767a\u30fb\u516c\u958b | \u6771\u4eac\u5927\u5b66\u677e\u5c3e\u30fb\u5ca9\u6fa4\u7814\u7a76\u5ba4\uff08\u677e\u5c3e\u7814\uff09- Matsuo Lab (u-tokyo.ac.jp)<\/a>\u304c\u767a\u8868\u3055\u308c\u3066\u3044\u308b\u306e\u304c\u7d20\u6674\u3089\u3057\u3044\u3002\u3053\u3061\u3089\u306f\u958b\u767a\u904e\u7a0b\u304c<a href=\"https:\/\/zenn.dev\/matsuolab\/articles\/377f7ae8b1169e\">\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30ebTanuki-8B, 8x8B\u306e\u4f4d\u7f6e\u3065\u3051\u3084\u958b\u767a\u6307\u91dd\u306a\u3069 (zenn.dev)<\/a>\u3067\u516c\u958b\u3055\u308c\u3066\u3044\u308b\u3002\u5408\u6210\u30c7\u30fc\u30bf\u306e\u6d3b\u7528\u306f\u6700\u8fd1\u767a\u8868\u3055\u308c\u305f\u6d77\u5916\u30e2\u30c7\u30eb\u3067\u3082\u591a\u7528\u3055\u308c\u3066\u3044\u308b\u30c6\u30af\u30cb\u30c3\u30af\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>OLMoE: Open Mixture-of-Experts Language Models\u00a0<\/strong>[180.2]<br>OLMoE\u306f\u3001Sparse Mixture-of-Experts (MoE)\u3092\u5229\u7528\u3057\u305f\u3001\u5b8c\u5168\u306b\u30aa\u30fc\u30d7\u30f3\u3067\u6700\u5148\u7aef\u306e\u8a00\u8a9e\u30e2\u30c7\u30eb\u3067\u3042\u308b\u3002 OLMoE-1B-7B\u306f70\u5104(B)\u306e\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u6301\u3064\u304c\u3001\u5165\u529b\u30c8\u30fc\u30af\u30f3\u3054\u3068\u306b1B\u3057\u304b\u4f7f\u7528\u3057\u306a\u3044\u3002 5\u5146\u306e\u30c8\u30fc\u30af\u30f3\u3067\u4e8b\u524d\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3057\u3001\u3055\u3089\u306bOLMoE-1B-7B-\u30a4\u30f3\u30b9\u30c8\u30e9\u30af\u30c8\u3092\u4f5c\u6210\u3059\u308b\u3088\u3046\u306b\u9069\u5fdc\u3057\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2409.02060v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2409.02060v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Tue, 03 Sep 2024 17:08:20 GMT)<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Reflection-Tuning: Data Recycling Improves LLM Instruction-Tuning\u00a0<\/strong>[79.3]<br>\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u30bb\u30c3\u30c8\u306e\u4f4e\u54c1\u8cea\u30c7\u30fc\u30bf\u306f\u3001\u901a\u5e38\u3001\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306e\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u306b\u6709\u5bb3\u3067\u3042\u308b\u3002 \u6211\u3005\u306f\u300c\u53cd\u5c04\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u300d\u3068\u547c\u3070\u308c\u308b\u65b0\u3057\u3044\u624b\u6cd5\u3092\u63d0\u6848\u3059\u308b\u3002 \u3053\u306e\u30a2\u30d7\u30ed\u30fc\u30c1\u3067\u306f\u3001\u30aa\u30e9\u30af\u30ebLSM\u3092\u4f7f\u7528\u3057\u3066\u3001\u30c7\u30fc\u30bf\u5185\u306e\u547d\u4ee4\u3084\u5fdc\u7b54\u306e\u8cea\u3092\u691c\u67fb\u3057\u3001\u5411\u4e0a\u3059\u308b\u3053\u3068\u3067\u3001\u5143\u306e\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u30c7\u30fc\u30bf\u3092\u30ea\u30b5\u30a4\u30af\u30eb\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2310.11716v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2310.11716v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 18 Oct 2023 05:13:47 GMT)<\/li>\n\n\n\n<li><strong>Selective Reflection-Tuning: Student-Selected Data Recycling for LLM Instruction-Tuning\u00a0<\/strong>[39.7]<br>\u6700\u8fd1\u306e\u591a\u304f\u306e\u624b\u6cd5\u306f\u30c7\u30fc\u30bf\u54c1\u8cea\u306e\u6539\u5584\u306b\u91cd\u70b9\u3092\u7f6e\u3044\u3066\u3044\u308b\u304c\u3001\u5b66\u751f\u30e2\u30c7\u30eb\u3068\u306e\u4e92\u63db\u6027\u3092\u898b\u843d\u3068\u3057\u3066\u3044\u308b\u3053\u3068\u304c\u591a\u3044\u3002 \u672c\u7a3f\u3067\u306f,\u6559\u5e2b\u306eLLM\u306e\u30ea\u30d5\u30ec\u30af\u30b7\u30e7\u30f3\u3068\u30a4\u30f3\u30c8\u30ed\u30b9\u30da\u30af\u30b7\u30e7\u30f3\u3092\u76f8\u4e57\u5316\u3057\u3066\u65e2\u5b58\u306e\u30c7\u30fc\u30bf\u54c1\u8cea\u3092\u6539\u5584\u3059\u308b\u65b0\u30d1\u30e9\u30c0\u30a4\u30e0\u3067\u3042\u308bSelective Reflection-Tuning\u3092\u7d39\u4ecb\u3059\u308b\u3002 \u3053\u306e\u6559\u5e2b\u3068\u5b66\u751f\u306e\u5354\u8abf\u4f5c\u696d\u306b\u3088\u308a\u3001\u9ad8\u54c1\u8cea\u3067\u751f\u5f92\u4e92\u63db\u306e\u6388\u696d\u5fdc\u7b54\u30da\u30a2\u304c\u4f5c\u6210\u3055\u308c\u3001\u7d50\u679c\u3068\u3057\u3066\u30b5\u30f3\u30d7\u30eb\u52b9\u7387\u306e\u3088\u3044\u6307\u5c0e\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u304c\u3082\u305f\u3089\u3055\u308c\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2402.10110v2\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2402.10110v2\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Fri, 07 Jun 2024 20:23:21 GMT)<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/tianyi-lab\/Reflection_Tuning\">GitHub &#8211; tianyi-lab\/Reflection_Tuning: [ACL&#8217;24] Selective Reflection-Tuning: Student-Selected Data Recycling for LLM Instruction-Tuning<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>OSS\u307e\u305f\u306f\u516c\u958b\u30e2\u30c7\u30eb\u306eLLM\u958b\u767a\u306f\u5f15\u304d\u7d9a\u304d\u76db\u3093\u3067\u3042\u308b\u3002 OLMoE-1B-7B\u306fMoE\u69cb\u6210\u306e\u5c0f\u898f\u6a21\u30fb\u9ad8\u6027\u80fd\u30e2\u30c7\u30eb\uff08\u30ea\u30dd\u30b8\u30c8\u30ea\uff1aallenai\/OLMoE-1B-7B-0924 \u00b7 Hugging Face\u3001Apache &hellip; <a href=\"https:\/\/devneko.jp\/wordpress\/?p=5420\" class=\"more-link\"><span class=\"screen-reader-text\">&#8220;OLMoE, Reflection-70B, Tanuki-8\u00d78B, 8B&#8221; \u306e<\/span>\u7d9a\u304d\u3092\u8aad\u3080<\/a><\/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":[223,293],"class_list":["post-5420","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-llm","tag-oss"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5420","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=5420"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5420\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5420"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5420"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5420"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}