{"id":28245,"date":"2025-02-07T10:36:47","date_gmt":"2025-02-07T02:36:47","guid":{"rendered":"https:\/\/www.1ai.net\/?p=28245"},"modified":"2025-02-07T10:36:47","modified_gmt":"2025-02-07T02:36:47","slug":"%e9%98%bf%e9%87%8c%e4%ba%91%e7%a1%ae%e8%ae%a4%ef%bc%9a%e6%9d%8e%e9%a3%9e%e9%a3%9e%e5%9b%a2%e9%98%9f-s1-%e6%a8%a1%e5%9e%8b%e5%9f%ba%e4%ba%8e-qwen2-5-32b-instruct-%e6%a8%a1%e5%9e%8b%e8%ae%ad%e7%bb%83","status":"publish","type":"post","link":"https:\/\/www.1ai.net\/en\/28245.html","title":{"rendered":"AliCloud Confirms: Feifei Li's Team s1 Model Trained on Qwen2.5-32B-Instruct Model"},"content":{"rendered":"

\u674e\u98de\u98de\u7814\u7a76\u56e2\u961f<\/a>\u4ee5\u4e0d\u5230 50 \u7f8e\u5143<\/strong>\u7684\u4e91\u8ba1\u7b97\u8d39\u7528\u8bad\u7ec3\u4e86\u4e00\u4e2a\u540d\u53eb s1 \u7684\u4eba\u5de5\u667a\u80fd\u63a8\u7406\u6a21\u578b\uff0c\u8be5\u6a21\u578b\u5728\u6570\u5b66\u548c\u7f16\u7801\u80fd\u529b\u6d4b\u8bd5\u4e2d\u7684\u8868\u73b0\u4e0e OpenAI \u7684 o1 \u548c DeepSeek \u7684 R1 \u7b49\u5c16\u7aef\u63a8\u7406\u6a21\u578b\u7c7b\u4f3c\u3002<\/p>\n

\"\u8bad\u7ec3\u6210\u672c\u4e0d\u5230<\/p>\n

\u4e0d\u8fc7\u5f88\u5feb\uff0c\u8be5 s1 \u6a21\u578b\u88ab\u6307\u201c\u5e76\u975e\u4ece\u96f6\u5f00\u59cb\u8bad\u7ec3\u201d\uff0c\u5176\u57fa\u5ea7\u6a21\u578b\u4e3a\u201c\u963f\u91cc\u901a\u4e49\u5343\u95ee\uff08Qwen\uff09\u6a21\u578b\u201d\u3002\u5bf9\u6b64\uff0c\u65b0\u6d6a\u79d1\u6280\u5411\u963f\u91cc\u4e91<\/a>\u65b9\u9762\u6c42\u8bc1\uff0c\u963f\u91cc\u4e91\u65b9\u9762\u786e\u8ba4\u4e86\u8fd9\u4e00\u6d88\u606f<\/strong>\u3002<\/p>\n

\u963f\u91cc\u4e91\u56de\u5e94\u79f0\uff1a\u201c\u4ed6\u4eec\u4ee5\u963f\u91cc\u901a\u4e49\u5343\u95ee Qwen2.5-32B-Instruct \u5f00\u6e90\u6a21\u578b\u4e3a\u5e95\u5ea7\uff0c\u5728 16 \u5757 H100 GPU \u4e0a\u76d1\u7763\u5fae\u8c03 26 \u5206\u949f\uff0c\u8bad\u7ec3\u51fa\u65b0\u6a21\u578b s1-32B\uff0c\u53d6\u5f97\u4e86\u4e0e OpenAI \u7684 o1 \u548c DeepSeek \u7684 R1 \u7b49\u5c16\u7aef\u63a8\u7406\u6a21\u578b\u6570\u5b66\u53ca\u7f16\u7801\u80fd\u529b\u76f8\u5f53\u7684\u6548\u679c\uff0c\u751a\u81f3\u5728\u7ade\u8d5b\u6570\u5b66\u95ee\u9898\u4e0a\u7684\u8868\u73b0\u6bd4 o1-preview \u9ad8\u51fa 27%\u3002\u201d<\/p>\n

\u636e1AI\u6b64\u524d\u62a5\u9053\uff0cs1 \u56e2\u961f\u900f\u9732\u4ed6\u4eec\u901a\u8fc7\u201c\u84b8\u998f\u201d\u6280\u672f\u521b\u5efa\u4e86\u8be5\u4eba\u5de5\u667a\u80fd\u6a21\u578b<\/strong>\uff0c\u8be5\u6280\u672f\u65e8\u5728\u901a\u8fc7\u8bad\u7ec3\u6a21\u578b\u6765\u5b66\u4e60\u53e6\u4e00\u4e2a\u4eba\u5de5\u667a\u80fd\u6a21\u578b\u7684\u7b54\u6848\uff0c\u4ece\u800c\u63d0\u53d6\u5176\u201c\u63a8\u7406\u201d\u80fd\u529b\u3002<\/p>\n

s1 \u7684\u8bba\u6587\u8868\u660e\uff0c\u53ef\u4ee5\u4f7f\u7528\u4e00\u79cd\u79f0\u4e3a\u76d1\u7763\u5fae\u8c03\uff08SFT\uff09\u7684\u65b9\u6cd5\uff0c\u53ef\u4ee5\u4f7f\u7528\u76f8\u5bf9\u8f83\u5c0f\u7684\u6570\u636e\u96c6\u6765\u84b8\u998f\u63a8\u7406\u6a21\u578b<\/strong>\u3002\u5728 SFT \u4e2d\uff0c\u4eba\u5de5\u667a\u80fd\u6a21\u578b\u4f1a\u88ab\u660e\u786e\u6307\u793a\u5728\u6570\u636e\u96c6\u4e2d\u6a21\u4eff\u67d0\u4e9b\u884c\u4e3a\u3002SFT \u6bd4 DeepSeek \u7528\u4e8e\u8bad\u7ec3\u5176 R1 \u6a21\u578b\u7684\u5927\u89c4\u6a21\u5f3a\u5316\u5b66\u4e60\u65b9\u6cd5\u66f4\u5177\u6210\u672c\u6548\u76ca\u3002<\/p>\n

s1 \u57fa\u4e8e\u963f\u91cc\u5df4\u5df4\u65d7\u4e0b\u4e2d\u56fd\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4 Qwen \u63d0\u4f9b\u7684\u4e00\u6b3e\u5c0f\u578b\u3001\u73b0\u6210\u7684\u514d\u8d39\u4eba\u5de5\u667a\u80fd\u6a21\u578b\u3002\u4e3a\u4e86\u8bad\u7ec3 s1\uff0c\u7814\u7a76\u4eba\u5458\u521b\u5efa\u4e86\u4e00\u4e2a\u4ec5\u5305\u542b 1000 \u4e2a\u7cbe\u5fc3\u7b56\u5212\u7684\u95ee\u9898\u7684\u6570\u636e\u96c6<\/strong>\uff0c\u4ee5\u53ca\u8fd9\u4e9b\u95ee\u9898\u7684\u7b54\u6848\uff0c\u4ee5\u53ca\u8c37\u6b4c Gemini 2.0 Flash Thinking Experimental \u7ed9\u51fa\u7684\u6bcf\u4e2a\u7b54\u6848\u80cc\u540e\u7684\u201c\u601d\u8003\u201d\u8fc7\u7a0b\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

Feifei Li's research team trained an AI inference model called s1 for less than $50 in the cloud, which performed similarly to cutting-edge inference models such as OpenAI's o1 and DeepSeek's R1 in tests of mathematical and coding ability. However, it was soon revealed that the s1 model was \"not trained from scratch\" and that its base model was the \"Ali Tongyi Qwen\" model. In this regard, Sina Technology asked AliCloud for confirmation, AliCloud confirmed the news. AliCloud responded: \"They used the Ali Tongyi Qwen2.5-32B-Instruct open-source model as a base, and supervised the fine-tuning on 16 H100 GPUs for 26 minutes to train a new model!<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[146],"tags":[3586,334],"collection":[],"class_list":["post-28245","post","type-post","status-publish","format-standard","hentry","category-news","tag-3586","tag-334"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/28245","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/comments?post=28245"}],"version-history":[{"count":0,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/posts\/28245\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/media?parent=28245"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/categories?post=28245"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/tags?post=28245"},{"taxonomy":"collection","embeddable":true,"href":"https:\/\/www.1ai.net\/en\/wp-json\/wp\/v2\/collection?post=28245"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}