{"id":6142,"date":"2025-02-03T06:01:00","date_gmt":"2025-02-02T21:01:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=6142"},"modified":"2025-02-03T06:01:00","modified_gmt":"2025-02-02T21:01:00","slug":"o3-mini-mistral-small-tulu-3-405b-open-r1","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=6142","title":{"rendered":"Qwen2.5-Max, Janus-Pro, o3-mini, Mistral Small, Tulu 3 405B, Open R1, BAICHUAN-OMNI-1.5"},"content":{"rendered":"\n<p>Deepseek V3\/R1\u95a2\u9023\u306e\u8a71\u984c\u304c\u76db\u308a\u4e0a\u304c\u308b\u4e2d\u3001\u5148\u9031\u3082\u69d8\u3005\u306a\u8a71\u984c\u304c\u3042\u3063\u305f\u3002Deepseek\u304b\u3089\u306f\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u30e2\u30c7\u30ebJanus-Pro\uff08<a href=\"https:\/\/github.com\/deepseek-ai\/Janus\">GitHub &#8211; deepseek-ai\/Janus: Janus-Series: Unified Multimodal Understanding and Generation Models<\/a>\uff09\u3001Alibaba\u304b\u3089\u306fDeepseek\u306b\u5bfe\u6297\u3059\u308b\u3088\u3046\u306aQwen\u306e\u6700\u5927\u30e2\u30c7\u30eb\u306e\u767a\u8868\uff08<a href=\"https:\/\/qwenlm.github.io\/blog\/qwen2.5-max\/\">Qwen2.5-Max: Exploring the Intelligence of Large-scale MoE Model | Qwen<\/a>\uff09\u304c\u767a\u8868\u3055\u308c\u305f\u3002\u30ed\u30f3\u30b0\u30b3\u30f3\u30c6\u30ad\u30b9\u5316\u306b\u95a2\u3059\u308b\u8ad6\u6587\u3082\u51fa\u3066\u3044\u308b\u70b9\u306b\u3082\u8981\u6ce8\u76ee\u3002<\/p>\n\n\n\n<p>OpenAI\u304b\u3089o3-mini\uff08<a href=\"https:\/\/openai.com\/index\/openai-o3-mini\/\">OpenAI o3-mini | OpenAI<\/a>\uff09\u304c\u51fa\u3066Humanity&#8217;s Last Exam\u3067o1\u3084r1\u3092\u8d85\u3048\u305f\u306e\u306f\u5927\u304d\u306a\u30cb\u30e5\u30fc\u30b9\u3060\u3063\u305f\uff08\u82e5\u5e72\u8a87\u5927\u5e83\u544a\u6c17\u5473\u306a\u90e8\u5206\u306f\u3042\u308b\u304c\uff09<\/p>\n\n\n\n<p>Mistral\u304b\u3089\u306f\u5c0f\u898f\u6a21\u3067\u9ad8\u6027\u80fd\u306a\u30e2\u30c7\u30ebMistral small\uff08<a href=\"https:\/\/mistral.ai\/news\/mistral-small-3\/\">Mistral Small 3 | Mistral AI | Frontier AI in your hands<\/a>\u3001<a href=\"https:\/\/huggingface.co\/mistralai\/Mistral-Small-24B-Instruct-2501\">mistralai\/Mistral-Small-24B-Instruct-2501 \u00b7 Hugging Face<\/a>\uff09\u304cApache-2\u30e9\u30a4\u30bb\u30f3\u30b9\u3067\u767a\u8868\u3055\u308c\u305f\u3002<\/p>\n\n\n\n<p>Ai2\u304b\u3089\u306f\u5927\u898f\u6a21\u9ad8\u6027\u80fd\u306aLLM Tulu3\uff08<a href=\"https:\/\/allenai.org\/blog\/tulu-3-405B\">Scaling the T\u00fclu 3 post-training recipes to surpass the performance of DeepSeek V3 | Ai2\uff09<\/a>\uff08Llama 3.1\u30d9\u30fc\u30b9\u3001405B\uff09\u304c\u3001HuggingFace\u304b\u3089\u306fDeepseek R1\u306e\u518d\u73fe\u3092\u76ee\u6307\u3059Open R1\u304c\u767a\u8868\u3055\u308c\u308b\uff08<a href=\"https:\/\/huggingface.co\/blog\/open-r1\">Open-R1: a fully open reproduction of DeepSeek-R1<\/a>\uff09\u306a\u3069\u30aa\u30fc\u30d7\u30f3\u306a\u53d6\u308a\u7d44\u307f\u3082\u76db\u308a\u4e0a\u304c\u3063\u3066\u3044\u308b\u3002<\/p>\n\n\n\n<p>\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u5316\u306e\u6d41\u308c\u3067\u3082\u300cOpen-source Omni-modal Foundation Model Supporting Text, Image, Video, and Audio Inputs as Well as Text and Audio Outputs\u300d\u306aBAICHUAN-OMNI-1.5\u306e\u30c6\u30af\u30cb\u30ab\u30eb\u30ec\u30dd\u30fc\u30c8\u304c\u51fa\u3066\u304a\u308a\u3001\u30af\u30ed\u30fc\u30ba\u30c9\u3001\u30aa\u30fc\u30d7\u30f3\u4e21\u65b9\u306e\u9663\u55b6\u3068\u3082\u7af6\u4e89\u304c\u975e\u5e38\u306b\u6fc0\u3057\u3044\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling\u00a0<\/strong>[27.1]<br>\u6211\u3005\u306f\u3001Janus-Pro\u3068\u3044\u3046\u524d\u4f5c\u306e\u5148\u9032\u7684\u306a\u30d0\u30fc\u30b8\u30e7\u30f3\u3092\u7d39\u4ecb\u3057\u307e\u3059\u3002 Janus-Pro\u306f(1)\u6700\u9069\u5316\u3055\u308c\u305f\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u6226\u7565\u3001(2)\u62e1\u5f35\u3055\u308c\u305f\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u30c7\u30fc\u30bf\u3001(3)\u3088\u308a\u5927\u304d\u306a\u30e2\u30c7\u30eb\u30b5\u30a4\u30ba\u3078\u306e\u30b9\u30b1\u30fc\u30ea\u30f3\u30b0\u3092\u53d6\u308a\u5165\u308c\u3066\u3044\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2501.17811v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2501.17811v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 29 Jan 2025 18:00:19 GMT)<\/li>\n\n\n\n<li>\u300cWe apply independent encoding methods to convert the raw inputs into features, which are then processed by an unified autoregressive transformer.\u300d\u3068\u3001Auto regressive transformer\u3001LLaVA\u306b\u6bd4\u3079\u3066\u30d1\u30e9\u30e1\u30fc\u30bf\u52b9\u7387\u304c\u9ad8\u3044<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Qwen2.5-1M Technical Report\u00a0<\/strong>[72.1]<br>\u30b3\u30f3\u30c6\u30af\u30b9\u30c8\u9577\u3092100\u4e07\u30c8\u30fc\u30af\u30f3\u307e\u3067\u62e1\u5f35\u3059\u308b\u4e00\u9023\u306e\u30e2\u30c7\u30eb\u3067\u3042\u308bQwen2.5-1M\u3092\u7d39\u4ecb\u3059\u308b\u3002 \u6211\u3005\u306e\u63a8\u8ad6\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3092\u6d3b\u7528\u3059\u308b\u3053\u3068\u3067\u3001Qwen2.5-1M\u30e2\u30c7\u30eb\u306f\u9a5a\u304f\u3079\u304d3\u500d\u304b\u30897\u500d\u306e\u30d7\u30ea\u30d5\u30a3\u30eb\u30b9\u30d4\u30fc\u30c9\u30a2\u30c3\u30d7\u3092\u9054\u6210\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2501.15383v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2501.15383v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Sun, 26 Jan 2025 03:47:25 GMT)<\/li>\n\n\n\n<li>Qwen\u306e\u30ed\u30f3\u30b0\u30b3\u30f3\u30c6\u30ad\u30b9\u5316<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Baichuan-Omni-1.5 Technical Report&nbsp;<\/strong>[78.5]<br>Baichuan-Omni-1.5\u306f\u3001Omni-modal\u306e\u7406\u89e3\u80fd\u529b\u3060\u3051\u3067\u306a\u304f\u3001\u30a8\u30f3\u30c9\u30c4\u30fc\u30a8\u30f3\u30c9\u306e\u30aa\u30fc\u30c7\u30a3\u30aa\u751f\u6210\u6a5f\u80fd\u3082\u5099\u3048\u305fOmni-modal\u30e2\u30c7\u30eb\u3067\u3042\u308b\u3002 \u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u30c7\u30fc\u30bf\u306e\u305f\u3081\u306e\u5305\u62ec\u7684\u30c7\u30fc\u30bf\u30af\u30ea\u30fc\u30cb\u30f3\u30b0\u3068\u5408\u6210\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u3092\u69cb\u7bc9\u3057,\u7d04500B\u306e\u9ad8\u54c1\u8cea\u30c7\u30fc\u30bf\u3092\u53d6\u5f97\u3059\u308b\u3002 \u7b2c\u4e8c\u306b\u3001\u97f3\u58f0\u30c8\u30b1\u30ca\u30a4\u30b6\u306f\u3001\u97f3\u58f0\u304b\u3089\u610f\u5473\u60c5\u5831\u3068\u97f3\u97ff\u60c5\u5831\u3092\u30ad\u30e3\u30d7\u30c1\u30e3\u3057\u3001\u30b7\u30fc\u30e0\u30ec\u30b9\u306a\u7d71\u5408\u3068MLLM\u3068\u306e\u4e92\u63db\u6027\u306e\u5f37\u5316\u3092\u53ef\u80fd\u306b\u3059\u308b\u3088\u3046\u306b\u8a2d\u8a08\u3055\u308c\u3066\u3044\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2501.15368v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2501.15368v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Sun, 26 Jan 2025 02:19:03 GMT)<\/li>\n\n\n\n<li>\u30aa\u30fc\u30d7\u30f3\u306aMLLM<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/baichuan-inc\/Baichuan-Omni-1.5\">GitHub &#8211; baichuan-inc\/Baichuan-Omni-1.5<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Humanity&#8217;s Last Exam&nbsp;<\/strong>[244.6]<br>Humanity\u306eLast Exam(HLE)\u306f\u3001\u4eba\u9593\u306e\u77e5\u8b58\u306e\u6700\u524d\u7dda\u306b\u304a\u3051\u308b\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067\u3042\u308b\u3002 \u6570\u5b66\u3001\u4eba\u6587\u79d1\u5b66\u3001\u81ea\u7136\u79d1\u5b66\u306a\u3069\u3001\u6570\u5341\u306e\u79d1\u76ee\u306b\u308f\u305f\u308b3000\u306e\u8cea\u554f\u3067\u69cb\u6210\u3055\u308c\u3066\u3044\u308b\u3002 \u5404\u8cea\u554f\u306b\u306f\u3001\u66d6\u6627\u3067\u5bb9\u6613\u306b\u691c\u8a3c\u3067\u304d\u308b\u65e2\u77e5\u306e\u30bd\u30ea\u30e5\u30fc\u30b7\u30e7\u30f3\u304c\u3042\u308b\u304c\u3001\u30a4\u30f3\u30bf\u30fc\u30cd\u30c3\u30c8\u691c\u7d22\u3067\u306f\u3059\u3050\u306b\u306f\u7b54\u3048\u3089\u308c\u306a\u3044\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2501.14249v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2501.14249v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Fri, 24 Jan 2025 05:27:46 GMT)<\/li>\n\n\n\n<li>\u73fe\u72b6\u306eAI\u3067\u89e3\u304f\u306e\u304c\u56f0\u96e3\u306a\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3001\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u30b5\u30a4\u30c8\u306f<a href=\"https:\/\/lastexam.ai\/\">Humanity&#8217;s Last Exam<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Tulu 3: Pushing Frontiers in Open Language Model Post-Training&nbsp;<\/strong>[94.1]<br>\u30c8\u30a5\u30eb3(Tulu 3)\u306f\u3001\u6700\u5148\u7aef\u306e\u8a13\u7df4\u5f8c\u30e2\u30c7\u30eb\u3067\u3042\u308b\u3002 Tulu 3\u306fLlama 3.1\u30d9\u30fc\u30b9\u30e2\u30c7\u30eb\u3092\u30d9\u30fc\u30b9\u306b\u3057\u3066\u304a\u308a\u3001Llama 3.1\u3001Qwen 2.5\u3001Mistral\u3001\u3055\u3089\u306bGPT-4o-mini\u3001Claude 3.5-Haiku\u3068\u3044\u3063\u305f\u30af\u30ed\u30fc\u30ba\u30c9\u30e2\u30c7\u30eb\u306b\u3082\u52dd\u3063\u3066\u3044\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2411.15124v3\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2411.15124v3\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Wed, 29 Jan 2025 18:46:59 GMT)<\/li>\n\n\n\n<li>\u3082\u3068\u3082\u3068\u306f11\u6708\u306b\u51fa\u305f\u8ad6\u6587\u3002405B\u7248\u306f\u975e\u5e38\u306b\u9ad8\u6027\u80fd\u3002<\/li>\n\n\n\n<li>\u4e0a\u8ff0\u306e\u30b5\u30a4\u30c8\u3067\u306f\u300cInterestingly, we found that our Reinforcement Learning from Verifiable Rewards (RLVR) framework improved the MATH performance more significantly at a larger scale, i.e., 405B compared to 70B and 8B, similar to the findings in the DeepSeek-R1 report. Overall, our results show a consistent edge over DeepSeek V3, especially with the inclusion of safety benchmarks.\u300d\u3068\u306e\u3053\u3068\u3002<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Deepseek V3\/R1\u95a2\u9023\u306e\u8a71\u984c\u304c\u76db\u308a\u4e0a\u304c\u308b\u4e2d\u3001\u5148\u9031\u3082\u69d8\u3005\u306a\u8a71\u984c\u304c\u3042\u3063\u305f\u3002Deepseek\u304b\u3089\u306f\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u30e2\u30c7\u30ebJanus-Pro\uff08GitHub &#8211; deepseek-ai\/Janus: Janus &hellip; <a href=\"https:\/\/devneko.jp\/wordpress\/?p=6142\" class=\"more-link\"><span class=\"screen-reader-text\">&#8220;Qwen2.5-Max, Janus-Pro, o3-mini, Mistral Small, Tulu 3 405B, Open R1, BAICHUAN-OMNI-1.5&#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],"class_list":["post-6142","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-llm"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6142","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=6142"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/6142\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=6142"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=6142"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=6142"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}