{"id":5055,"date":"2024-07-01T06:44:00","date_gmt":"2024-06-30T21:44:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=5055"},"modified":"2024-07-01T06:44:00","modified_gmt":"2024-06-30T21:44:00","slug":"gemma2-criticgpt","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=5055","title":{"rendered":"Gemma2, CriticGPT"},"content":{"rendered":"\n<p>Google\u304b\u3089\u516c\u958b\u30e2\u30c7\u30eb\u3068\u3057\u3066\u306f\u898f\u6a21\u306e\u5927\u304d\u3044LLM Gemma2\u304c\u30ea\u30ea\u30fc\u30b9\u3055\u308c\u305f\u30029B\u306827B\u306e\u516c\u958b\u3002Llama3\u306a\u3069\u7af6\u5408\u3059\u308b\u516c\u958b\u30e2\u30c7\u30eb\u3092\u8d85\u3048\u308b\u6027\u80fd\u3068\u306e\u3053\u3068\u3002\u30c6\u30af\u30cb\u30ab\u30eb\u30ec\u30dd\u30fc\u30c8\uff08<a href=\"https:\/\/storage.googleapis.com\/deepmind-media\/gemma\/gemma-2-report.pdf\">gemma-2-report.pdf (storage.googleapis.com)<\/a>\uff09\u306b\u306f\u300cThe 9 billion and 27 billion parameter models are available today, with a 2 billion parameter model to be released shortly.\u300d\u3068\u3042\u308b\u3002\u300cWe also train the 2B and 9B models with knowledge distillation (Hinton et al , 2015) instead of next token prediction. The resulting models deliver the best performance for their size, and even offer competitive alternatives to models that are 2-3\u00d7 bigger.\u300d\u3068\u84b8\u7559\u3092\u52b9\u679c\u7684\u306b\u4f7f\u3063\u3066\u3044\u308b\u3082\u306e\u9762\u767d\u3044\u30025. Ablations\u3092\u307f\u308b\u306b\u52b9\u679c\u306f\u5927\u304d\u305d\u3046<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u30cb\u30e5\u30fc\u30b9\u30ea\u30ea\u30fc\u30b9\uff1a<a href=\"https:\/\/blog.google\/technology\/developers\/google-gemma-2\/\">Google launches Gemma 2, its next generation of open models (blog.google)<\/a><\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\n<ul class=\"wp-block-list\">\n<li><a href=\"https:\/\/huggingface.co\/google\/gemma-2-9b-it\">google\/gemma-2-9b-it \u00b7 Hugging Face<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/huggingface.co\/google\/gemma-2-27b-it\">google\/gemma-2-27b-it \u00b7 Hugging Face<\/a><\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n<p>\u3044\u3064\u3082\u306e\u7ffb\u8a33\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067\u306f\u975e\u5e38\u306b\u9ad8\u3044\u6027\u80fd\u3092\u793a\u3057\u305f\u3002\u671f\u5f85\u5927\u3067\u3042\u308b\u3002<a href=\"https:\/\/staka.jp\/wordpress\/?p=1096\">Gemma 2 9B\u306e\u6a5f\u68b0\u7ffb\u8a33\u6027\u80fd | \u3077\u308b\u30fc\u3075\u304a\u3076\u3053\u3093\u305b\u3077\u3068 (staka.jp)<\/a><\/p>\n\n\n\n<p>OpenAI\u304b\u3089\u306fGPT-4\u306e\u9593\u9055\u3044\u3092\u898b\u3064\u3051\u4fee\u6b63\u63d0\u6848\u3059\u308bCriticGPT\u304c\u51fa\u3066\u3044\u308b\u3002\u4eca\u306f\u30b3\u30fc\u30c9\u306e\u4fee\u6b63\u304c\u5bfe\u8c61\u3002\u9650\u754c\u3082\u3042\u308b\u3088\u3046\u3060\u304c\u3053\u306e\u3088\u3046\u306a\u7814\u7a76\u306f\u91cd\u8981\u3002<a href=\"https:\/\/openai.com\/index\/finding-gpt4s-mistakes-with-gpt-4\/\">Finding GPT-4\u2019s mistakes with GPT-4 | OpenAI<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Google\u304b\u3089\u516c\u958b\u30e2\u30c7\u30eb\u3068\u3057\u3066\u306f\u898f\u6a21\u306e\u5927\u304d\u3044LLM Gemma2\u304c\u30ea\u30ea\u30fc\u30b9\u3055\u308c\u305f\u30029B\u306827B\u306e\u516c\u958b\u3002Llama3\u306a\u3069\u7af6\u5408\u3059\u308b\u516c\u958b\u30e2\u30c7\u30eb\u3092\u8d85\u3048\u308b\u6027\u80fd\u3068\u306e\u3053\u3068\u3002\u30c6\u30af\u30cb\u30ab\u30eb\u30ec\u30dd\u30fc\u30c8\uff08gemma-2-report.pdf ( &hellip; <a href=\"https:\/\/devneko.jp\/wordpress\/?p=5055\" class=\"more-link\"><span class=\"screen-reader-text\">&#8220;Gemma2, CriticGPT&#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":[84,223,293,468],"class_list":["post-5055","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-critic","tag-llm","tag-oss","tag-468"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5055","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=5055"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/5055\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5055"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5055"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5055"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}