{"id":35648,"date":"2024-12-29T00:00:13","date_gmt":"2024-12-28T16:00:13","guid":{"rendered":"https:\/\/17aitech.com\/?p=35648"},"modified":"2024-12-29T00:00:13","modified_gmt":"2024-12-28T16:00:13","slug":"openai-o1%e8%a6%81%e8%b7%9f%ef%bc%8c%e6%80%8e%e4%b9%88%e8%b7%9f%ef%bc%9f%e8%bf%99%e4%b8%aagithub%e9%a1%b9%e7%9b%ae%e6%8a%8a%e8%a7%a3%e8%af%bb%e3%80%81%e5%8d%9a%e5%ae%a2%e3%80%81%e7%9b%b8%e5%85%b3","status":"publish","type":"post","link":"https:\/\/17aitech.com\/?p=35648","title":{"rendered":"OpenAI o1\u8981\u8ddf\uff0c\u600e\u4e48\u8ddf\uff1f\u8fd9\u4e2aGitHub\u9879\u76ee\u628a\u89e3\u8bfb\u3001\u535a\u5ba2\u3001\u76f8\u5173\u8bba\u6587\u4e00\u7f51\u6253\u5c3d"},"content":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-09-17\" target=\"_blank\">OpenAI o1\u8981\u8ddf\uff0c\u600e\u4e48\u8ddf\uff1f\u8fd9\u4e2aGitHub\u9879\u76ee\u628a\u89e3\u8bfb\u3001\u535a\u5ba2\u3001\u76f8\u5173\u8bba\u6587\u4e00\u7f51\u6253\u5c3d<\/a><\/p>\n<p>\u5728 AI \u9886\u57df\uff0cOpenAI \u5df2\u7ecf\u6210\u4e86\u6307\u8def\u660e\u706f\u4e00\u822c\u7684\u5b58\u5728\u3002\u968f\u7740 o1 \u6a21\u578b\u7684\u53d1\u5e03\uff0c\u5168\u4e16\u754c\u7684 AI \u5de5\u7a0b\u5e08\u90fd\u5f00\u59cb\u4e86\u65b0\u4e00\u8f6e\u7684\u5b66\u4e60\u3002<\/p>\n<p>\u4e3a\u4e86\u5e2e\u52a9\u5927\u5bb6\u5c3d\u5feb\u6293\u4f4f\u91cd\u70b9\uff0c\u673a\u5668\u4e4b\u5fc3\u4e00\u76f4\u5728\u8ddf\u8fdb\u62a5\u9053\u76f8\u5173\u7684\u89e3\u8bfb\uff0c\u5305\u62ec\uff1a<\/p>\n<ul>\n<li><a data-itemshowtype=\"0\" data-linktype=\"2\" href=\"http:\/\/mp.weixin.qq.com\/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2650934580&amp;idx=1&amp;sn=41c727eb05b277ea855f03962bf7e5a0&amp;chksm=84e7c94ab390405c71972e79b15858cb5270cb8058782d607112edbeedd3c7f23780c8bcf757&amp;scene=21#wechat_redirect\" target=\"_blank\">\u5317\u5927\u5bf9\u9f50\u56e2\u961f\u72ec\u5bb6\u89e3\u8bfb\uff1aOpenAI o1 \u5f00\u542f\u300c\u540e\u8bad\u7ec3\u300d\u65f6\u4ee3\u5f3a\u5316\u5b66\u4e60\u65b0\u8303\u5f0f<\/a><\/li>\n<li><a data-itemshowtype=\"0\" data-linktype=\"2\" href=\"http:\/\/mp.weixin.qq.com\/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2650934463&amp;idx=2&amp;sn=641e7849ea6e8b207d010c709166d78a&amp;chksm=84e7c8c1b39041d7a21b578a3cdeb620d783c50004c0e138e3e784a584d837f8b5a585b48658&amp;scene=21#wechat_redirect\" target=\"_blank\">\u5f20\u4fca\u6797\uff1aOpenAI o1 \u7684\u4ef7\u503c\u610f\u4e49\u53ca\u5f3a\u5316\u5b66\u4e60\u7684 Scaling Law<\/a><\/li>\n<\/ul>\n<p>\u540c\u65f6\uff0c\u6211\u4eec\u4e5f\u53d1\u73b0\u4e86\u5176\u4ed6\u4e00\u4e9b\u6709\u7528\u7684\u8d44\u6e90\uff0c\u6bd4\u5982\u5728\u4e00\u4e2a Github \u9879\u76ee\u4e2d\uff0c\u6709\u4eba\u6c47\u603b\u4e86\u6700\u8fd1\u7684\u9ad8\u8d28\u91cf\u6280\u672f\u89e3\u8bfb\u535a\u5ba2\u4ee5\u53ca\u300c\u53ef\u80fd\u300d\u4e0e o1 \u6280\u672f\u8def\u7ebf\u76f8\u5173\u7684\u8bba\u6587\u3002\u76f8\u5173\u8d44\u6e90\u5217\u8868\u4f1a\u4e00\u76f4\u66f4\u65b0\uff0c<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-b2f27343a5f79615937070bd54ce7fe8.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-b2f27343a5f79615937070bd54ce7fe8.png\"><\/a><\/p>\n<ul>\n<li>\u9879\u76ee\u94fe\u63a5\uff1ahttps:\/\/github.com\/hijkzzz\/Awesome-LLM-Strawberry<\/li>\n<\/ul>\n<p>\u8fd9\u4e9b\u535a\u5ba2\u3001\u8bba\u6587\u6709\u4e9b\u662f\u5927\u5bb6\u90fd\u8bfb\u8fc7\u7684\uff0c\u8fd8\u6709\u4e9b\u53ef\u80fd\u88ab\u6df9\u6ca1\u5728\u5e73\u65f6\u7684\u4f17\u591a\u8d44\u6e90\u4e2d\u3002\u6216\u8bb8\u4ece\u4e2d\u5927\u5bb6\u53ef\u4ee5\u627e\u5230\u590d\u73b0 OpenAI o1 \u7684\u6709\u6548\u65b9\u6cd5\u3002<\/p>\n<p><strong>\u535a\u5ba2<\/strong><\/p>\n<ul>\n<li>\u535a\u5ba2 1\uff1aLearning to Reason with LLMs<\/li>\n<li>\u4f5c\u8005\uff1aOpenAI<\/li>\n<li>\u94fe\u63a5\uff1ahttps:\/\/openai.com\/index\/learning-to-reason-with-llms\/<\/li>\n<\/ul>\n<p>\u535a\u5ba2\u6982\u8ff0\uff1a\u8fd9\u7bc7\u535a\u5ba2\u7b80\u5355\u4ecb\u7ecd\u4e86 OpenAI o1 \u6a21\u578b\u7684\u8bad\u7ec3\u65b9\u6cd5\uff0c\u6bd4\u5982\u601d\u7ef4\u94fe\u7684\u91c7\u7528\u3001\u6a21\u578b\u5b89\u5168\u6027\u7684\u63d0\u5347\u7b49\u3002\u8be6\u60c5\u8bf7\u53c2\u89c1\u673a\u5668\u4e4b\u5fc3\u62a5\u9053\uff1a\u300a<a data-itemshowtype=\"0\" data-linktype=\"2\" href=\"http:\/\/mp.weixin.qq.com\/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2650934313&amp;idx=1&amp;sn=3e89923751621fed57602ffe4ba751f4&amp;chksm=84e7c857b3904141d5ef61e834efeb469e7982498aca47cfe92cadcf3912908754e752793c2a&amp;scene=21#wechat_redirect\" target=\"_blank\">\u521a\u521a\uff0cOpenAI \u9707\u64bc\u53d1\u5e03 o1 \u5927\u6a21\u578b\uff01\u5f3a\u5316\u5b66\u4e60\u7a81\u7834 LLM \u63a8\u7406\u6781\u9650<\/a>\u300b<\/p>\n<ul>\n<li>\u535a\u5ba2 2\uff1aOpenAI o1-mini<\/li>\n<li>\u4f5c\u8005\uff1aOpenAI<\/li>\n<li>\u94fe\u63a5\uff1ahttps:\/\/openai.com\/index\/openai-o1-mini-advancing-cost-efficient-reasoning\/<\/li>\n<\/ul>\n<p>\u535a\u5ba2\u6982\u8ff0\uff1a\u8fd9\u7bc7\u535a\u5ba2\u4ecb\u7ecd\u4e86 OpenAI o1-mini \u6a21\u578b\u7684\u6982\u51b5\u3002\u8be6\u60c5\u8bf7\u53c2\u89c1\u673a\u5668\u4e4b\u5fc3\u62a5\u9053\uff1a\u300a<a data-itemshowtype=\"0\" data-linktype=\"2\" href=\"http:\/\/mp.weixin.qq.com\/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2650934313&amp;idx=1&amp;sn=3e89923751621fed57602ffe4ba751f4&amp;chksm=84e7c857b3904141d5ef61e834efeb469e7982498aca47cfe92cadcf3912908754e752793c2a&amp;scene=21#wechat_redirect\" target=\"_blank\">\u521a\u521a\uff0cOpenAI \u9707\u64bc\u53d1\u5e03 o1 \u5927\u6a21\u578b\uff01\u5f3a\u5316\u5b66\u4e60\u7a81\u7834 LLM \u63a8\u7406\u6781\u9650<\/a>\u300b<\/p>\n<ul>\n<li>\u535a\u5ba2 3\uff1aFinding GPT-4\u2019s mistakes with GPT-4<\/li>\n<li>\u4f5c\u8005\uff1aOpenAI<\/li>\n<li>\u94fe\u63a5\uff1ahttps:\/\/openai.com\/index\/finding-gpt4s-mistakes-with-gpt-4\/<\/li>\n<\/ul>\n<p>\u535a\u5ba2\u6982\u8ff0\uff1a\u8fd9\u7bc7\u535a\u5ba2\u4ecb\u7ecd\u4e86 CriticGPT\u2014\u2014OpenAI \u57fa\u4e8e GPT-4 \u8bad\u7ec3\u7684\u4e00\u4e2a\u4e13\u95e8\u7ed9 ChatGPT \u6311\u6bdb\u75c5\u7684\u65b0\u6a21\u578b\u3002\u5b83\u901a\u8fc7\u7cbe\u51c6\u5730\u5206\u6790 ChatGPT \u7684\u56de\u7b54\u5e76\u63d0\u51fa\u5efa\u8bbe\u6027\u7684\u6279\u8bc4\uff0c\u5e2e\u52a9\u4eba\u7c7b\u8bad\u7ec3\u5e08\u66f4\u51c6\u786e\u5730\u8bc4\u4f30\u6a21\u578b\u751f\u6210\u7684\u4ee3\u7801\uff0c\u5e76\u8bc6\u522b\u5176\u4e2d\u7684\u9519\u8bef\u6216\u6f5c\u5728\u95ee\u9898\u3002\u636e\u4ecb\u7ecd\uff0c\u5728 CriticGPT \u7684\u8f85\u52a9\u4e0b\uff0c\u4eba\u4eec\u5ba1\u67e5 ChatGPT \u4ee3\u7801\u7684\u51c6\u786e\u7387\u63d0\u9ad8\u4e86 60%\u3002\u7814\u7a76\u4eba\u5458\u8fd8\u53d1\u73b0\uff0cCriticGPT \u5728\u5f88\u591a\u60c5\u51b5\u4e0b\u6bd4\u4eba\u7c7b\u4e13\u5bb6\u66f4\u64c5\u957f\u53d1\u73b0\u9519\u8bef\uff0c\u5b83\u4eec\u751a\u81f3\u80fd\u5728\u4e00\u4e9b\u88ab\u8ba4\u4e3a\u662f\u300c\u5b8c\u7f8e\u65e0\u7f3a\u300d\u7684\u4efb\u52a1\u4e2d\u627e\u51fa\u95ee\u9898\uff0c\u5c3d\u7ba1\u8fd9\u4e9b\u4efb\u52a1\u5927\u591a\u6570\u5e76\u4e0d\u662f\u4ee3\u7801\u4efb\u52a1\uff0c\u5bf9 CriticGPT \u6765\u8bf4\u6709\u70b9\u8d85\u7eb2\u3002<\/p>\n<p>\u7531\u4e8e\u8be5\u7814\u7a76\u53d1\u8868\u65f6\uff0cJan Leike \u7b49 OpenAI \u8d85\u7ea7\u5bf9\u9f50\u56e2\u961f\u6210\u5458\u5df2\u7ecf\u79bb\u804c\uff0c\u56e0\u6b64\u4e5f\u88ab\u79f0\u4e3a\u5bf9\u9f50\u56e2\u961f\u300c\u9057\u4f5c\u300d\u3002\u8be6\u60c5\u8bf7\u53c2\u89c1\u673a\u5668\u4e4b\u5fc3\u62a5\u9053\uff1a\u300a<a data-itemshowtype=\"0\" data-linktype=\"2\" href=\"http:\/\/mp.weixin.qq.com\/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2650923946&amp;idx=1&amp;sn=f87d1558ff6784dc157e0c532c6819a8&amp;chksm=84e427d4b393aec2b55d0a572e588579bc1c444ff31e5a79239ed2f79f01051659fe09f807fe&amp;scene=21#wechat_redirect\" target=\"_blank\">OpenAI \u524d\u5bf9\u9f50\u56e2\u961f\u300c\u9057\u4f5c\u300d\uff1aRLHF \u4e0d\u591f\u7528\u4e86\uff01\u7528 GPT-4 \u8bad\u7ec3 GPT-4<\/a>\u300b<\/p>\n<ul>\n<li>\u535a\u5ba2 4\uff1aSummary of what we have learned during AMA hour with the OpenAI o1 team<\/li>\n<li>\u4f5c\u8005\uff1aTibor Blaho<\/li>\n<li>\u94fe\u63a5\uff1ahttps:\/\/twitter-thread.com\/t\/1834686946846597281<\/li>\n<li>\u63a8\u6587\u94fe\u63a5\uff1ahttps:\/\/x.com\/btibor91\/status\/1834686946846597281<\/li>\n<\/ul>\n<p>\u535a\u5ba2\u6982\u8ff0\uff1a\u6700\u8fd1\uff0cOpenAI o1 \u56e2\u961f\u5f00\u5c55\u4e86\u4e00\u6b21\u7b54\u7591\u6d3b\u52a8\uff0c\u8fd9\u4e2a\u5e16\u5b50\u603b\u7ed3\u4e86\u7b54\u7591\u7684\u6982\u8981\uff0c\u5305\u62ec\u6a21\u578b\u547d\u540d\u548c\u63a8\u7406\u8303\u5f0f\uff0co1 \u6a21\u578b\u7684\u5c3a\u5bf8\u548c\u6027\u80fd\uff0c\u8f93\u5165 token \u4e0a\u4e0b\u6587\u548c\u6a21\u578b\u80fd\u529b\uff0c\u5de5\u5177\u3001\u529f\u80fd\u548c\u5373\u5c06\u63a8\u51fa\u7684\u7279\u6027\uff0cCoT \u63a8\u7406\uff0cAPI \u548c\u4f7f\u7528\u9650\u5236\uff0c\u5b9a\u4ef7\u3001\u5fae\u8c03\u4e0e\u6269\u5c55\uff0c\u6a21\u578b\u5f00\u53d1\u548c\u7814\u7a76\u89c1\u89e3\uff0c\u63d0\u793a\u6280\u672f\u548c\u6700\u4f73\u5b9e\u8df5\u7b49\u51e0\u4e2a\u6a21\u5757\u3002\u6bcf\u4e2a\u6a21\u5757\u7684\u603b\u7ed3\u90fd\u6bd4\u8f83\u7b80\u77ed\uff0c\u611f\u5174\u8da3\u7684\u8bfb\u8005\u53ef\u4ee5\u53c2\u89c1\u539f\u6587\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-7ca73d5679a452ee537eef3020477e3e.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-7ca73d5679a452ee537eef3020477e3e.png\"><\/a><\/p>\n<ul>\n<li>\u535a\u5ba2 5\uff1aOpenAI\u2019s Strawberry, LM self-talk, inference scaling laws, and spending more on inference<\/li>\n<li>\u4f5c\u8005\uff1aNathan Lambert\uff08Allen AI \u7814\u7a76\u79d1\u5b66\u5bb6\uff09<\/li>\n<li>\u94fe\u63a5\uff1ahttps:\/\/www.interconnects.ai\/p\/openai-strawberry-and-inference-scaling-laws<\/li>\n<\/ul>\n<p>\u535a\u5ba2\u6982\u8ff0\uff1a\u5728\u6587\u7ae0\u4e2d\uff0c\u4f5c\u8005\u8ba8\u8bba\u4e86 OpenAI \u7684\u65b0\u65b9\u6cd5\u300cStrawberry\u300d\u53ca\u63a8\u7406 scaling law\uff0c\u5f3a\u8c03\u4e86\u63a8\u7406\u8ba1\u7b97\u7684\u6295\u5165\u5bf9 AI \u80fd\u529b\u63d0\u5347\u7684\u91cd\u8981\u6027\u3002\u4f5c\u8005\u6307\u51fa\uff0c\u6269\u5927\u63a8\u7406\u8ba1\u7b97\u6bd4\u5355\u7eaf\u6269\u5927\u6a21\u578b\u89c4\u6a21\u66f4\u6709\u6548\uff0c\u7c7b\u4f3c AlphaGo \u7684\u63a8\u7406\u6280\u672f\u80fd\u591f\u663e\u8457\u63d0\u5347\u6a21\u578b\u8868\u73b0\u3002\u6587\u7ae0\u547c\u5401\u672a\u6765 AI \u5f00\u53d1\u8981\u66f4\u591a\u5173\u6ce8\u63a8\u7406\u6280\u672f\u3002\u00a0<\/p>\n<p>\u8fd9\u7bc7\u535a\u5ba2\u53d1\u5e03\u4e8e 9 \u6708\u521d\uff0c\u5f53\u65f6 OpenAI \u8fd8\u6ca1\u6709\u53d1\u5e03 o1 \u6a21\u578b\uff0c\u56e0\u6b64\u73b0\u5728\u770b\u8d77\u6765\u975e\u5e38\u6709\u524d\u77bb\u6027\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-baf66ae53da40e5e75d0b2ecfd5d7fc5.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-baf66ae53da40e5e75d0b2ecfd5d7fc5.png\"><\/a><\/p>\n<ul>\n<li>\u535a\u5ba2 6\uff1aReverse engineering OpenAI\u2019s o1<\/li>\n<li>\u4f5c\u8005\uff1aNathan Lambert\uff08Allen AI \u7814\u7a76\u79d1\u5b66\u5bb6\uff09<\/li>\n<li>\u94fe\u63a5\uff1ahttps:\/\/www.interconnects.ai\/p\/reverse-engineering-openai-o1<\/li>\n<\/ul>\n<p>\u535a\u5ba2\u6982\u8ff0\uff1a\u8fd9\u7bc7\u535a\u5ba2\u7cfb\u7edf\u8ba8\u8bba\u4e86 OpenAI o1\u3002o1 \u901a\u8fc7\u8bad\u7ec3\u65b0\u6a21\u578b\u5904\u7406\u957f\u63a8\u7406\u94fe\uff0c\u5e76\u4f7f\u7528\u5927\u91cf\u5f3a\u5316\u5b66\u4e60\u6765\u5b9e\u73b0\u3002\u4e0e\u81ea\u56de\u5f52\u8bed\u8a00\u6a21\u578b\u4e0d\u540c\uff0co1 \u5728\u7ebf\u4e3a\u7528\u6237\u641c\u7d22\u7b54\u6848\uff0c\u5c55\u793a\u4e86\u65b0\u7684 scaling law\u2014\u2014 \u63a8\u7406 scaling law\u3002\u535a\u5ba2\u8fd8\u8ba8\u8bba\u4e86 o1 \u7684\u4e00\u4e9b\u6280\u672f\u7ec6\u8282\uff0c\u5305\u62ec\u5176\u5982\u4f55\u4f7f\u7528\u5f3a\u5316\u5b66\u4e60\u8fdb\u884c\u8bad\u7ec3\uff0c\u4ee5\u53ca\u5b83\u5728\u63a8\u7406\u65f6\u7684\u9ad8\u6210\u672c\u3002\u6b64\u5916\uff0c\u535a\u5ba2\u8fd8\u63a2\u8ba8\u4e86 o1 \u5bf9\u672a\u6765 AI \u9886\u57df\u7684\u5f71\u54cd\uff0c\u5305\u62ec\u5b83\u5982\u4f55\u6539\u53d8 AI \u4ea7\u54c1\u7684\u90e8\u7f72\u5806\u6808\u548c\u671f\u671b\uff0c\u4ee5\u53ca\u5b83\u5982\u4f55\u4f5c\u4e3a\u4e00\u4e2a\u6a21\u578b\uff0c\u901a\u8fc7\u4e0d\u540c\u7684\u751f\u6210\u7b56\u7565\u6765\u5b9e\u73b0\u590d\u6742\u7684\u4efb\u52a1\u3002\u6700\u540e\uff0c\u535a\u5ba2\u63d0\u51fa\u4e86\u4e00\u4e9b\u5173\u4e8e o1 \u7ed3\u6784\u548c\u529f\u80fd\u7684\u95ee\u9898\uff0c\u5e76\u8ba8\u8bba\u4e86\u5728\u5f00\u6e90\u9886\u57df\u590d\u5236\u8fd9\u79cd\u7cfb\u7edf\u6240\u9762\u4e34\u7684\u6311\u6218\u3002\u4f5c\u8005\u8fd8\u5bf9 AI \u672a\u6765\u7684\u53d1\u5c55\u65b9\u5411\u8868\u793a\u4e86\u671f\u5f85\uff0c\u8ba4\u4e3a AI \u7684\u8fdb\u6b65\u5c06\u7ee7\u7eed\u5956\u52b1\u90a3\u4e9b\u6562\u4e8e\u60f3\u8c61\u4e0d\u53ef\u80fd\u5f88\u5feb\u53d8\u4e3a\u53ef\u80fd\u7684\u4eba\u3002<a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-0f138d50f2c8692173d17dc900c7ea4c.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-0f138d50f2c8692173d17dc900c7ea4c.png\"><\/a><\/p>\n<p><strong>\u8bba\u6587<\/strong><\/p>\n<p><strong>OpenAI o1 \u8d21\u732e\u8005\u53c2\u4e0e\u64b0\u5199\u7684\u8bba\u6587<\/strong><\/p>\n<ul>\n<li>\u8bba\u6587 1\uff1aTraining Verifiers to Solve Math Word Problems<\/li>\n<li>\u673a\u6784\uff1aOpenAI<\/li>\n<li>\u4f5c\u8005\uff1aKarl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, John Schulman<\/li>\n<li>\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/2110.14168<\/li>\n<\/ul>\n<p>\u8bba\u6587\u6982\u8ff0\uff1a\u8fd9\u7bc7\u8bba\u6587\u53d1\u5e03\u4e8e 2021 \u5e74 10 \u6708\u3002\u8bba\u6587\u6307\u51fa\uff0c\u5c3d\u7ba1\u6700\u5148\u8fdb\u7684\u8bed\u8a00\u6a21\u578b\u5728\u5f88\u591a\u4efb\u52a1\u4e0a\u8868\u73b0\u4f18\u5f02\uff0c\u4f46\u5728\u5904\u7406\u591a\u6b65\u9aa4\u6570\u5b66\u63a8\u7406\u65f6\u4ecd\u6709\u56f0\u96be\u3002\u4e3a\u4e86\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\uff0c\u4f5c\u8005\u521b\u5efa\u4e86 GSM8K \u6570\u636e\u96c6\uff0c\u5305\u542b 8500 \u4e2a\u591a\u6837\u5316\u7684\u5c0f\u5b66\u6570\u5b66\u95ee\u9898\u3002\u7814\u7a76\u53d1\u73b0\uff0c\u5373\u4f7f\u662f\u5927\u578b Transformer \u6a21\u578b\u4e5f\u96be\u4ee5\u5728\u8fd9\u4e9b\u4efb\u52a1\u4e0a\u53d6\u5f97\u597d\u6210\u7ee9\u3002\u4e3a\u4e86\u63d0\u9ad8\u6027\u80fd\uff0c\u4f5c\u8005\u5efa\u8bae\u8bad\u7ec3\u9a8c\u8bc1\u5668\u6765\u8bc4\u4f30\u6a21\u578b\u7b54\u6848\u7684\u6b63\u786e\u6027\u3002\u901a\u8fc7\u5728\u6d4b\u8bd5\u65f6\u751f\u6210\u591a\u4e2a\u7b54\u6848\u5e76\u9009\u62e9\u9a8c\u8bc1\u5668\u8bc4\u5206\u6700\u9ad8\u7684\u7b54\u6848\uff0c\u8fd9\u79cd\u65b9\u6cd5\u663e\u8457\u63d0\u5347\u4e86\u6a21\u578b\u5728 GSM8K \u4e0a\u7684\u8868\u73b0\uff0c\u5e76\u8bc1\u660e\u4e86\u8fd9\u79cd\u65b9\u6cd5\u6bd4\u4f20\u7edf\u7684\u5fae\u8c03\u65b9\u6cd5\u66f4\u6709\u6548\u3002<\/p>\n<ul>\n<li>\u8bba\u6587 2\uff1aGenerative Language Modeling for Automated Theorem Proving<\/li>\n<li>\u673a\u6784\uff1aOpenAI<\/li>\n<li>\u4f5c\u8005\uff1aStanislas Polu, Ilya Sutskever<\/li>\n<li>\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/2009.03393<\/li>\n<\/ul>\n<p>\u8bba\u6587\u6982\u8ff0\uff1a\u8fd9\u7bc7\u8bba\u6587\u53d1\u5e03\u4e8e 2020 \u5e74 9 \u6708\uff0cIlya Sutskever \u662f\u4f5c\u8005\u4e4b\u4e00\u3002\u8bba\u6587\u63a2\u8ba8\u4e86\u57fa\u4e8e Transformer \u7684\u8bed\u8a00\u6a21\u578b\u5728\u81ea\u52a8\u5b9a\u7406\u8bc1\u660e\u4e2d\u7684\u5e94\u7528\u3002\u7814\u7a76\u7684\u52a8\u673a\u662f\uff0c\u81ea\u52a8\u5b9a\u7406\u8bc1\u660e\u5668\u4e0e\u4eba\u7c7b\u76f8\u6bd4\u7684\u4e00\u4e2a\u4e3b\u8981\u9650\u5236 \u2014\u2014 \u751f\u6210\u539f\u521b\u7684\u6570\u5b66\u672f\u8bed \u2014\u2014 \u53ef\u80fd\u53ef\u4ee5\u901a\u8fc7\u8bed\u8a00\u6a21\u578b\u7684\u751f\u6210\u6765\u89e3\u51b3\u3002\u4f5c\u8005\u4ecb\u7ecd\u4e86\u4e00\u4e2a\u540d\u4e3a GPT-f \u7684\u81ea\u52a8\u8bc1\u660e\u5668\u548c\u8bc1\u660e\u52a9\u624b\uff0c\u7528\u4e8e Metamath \u5f62\u5f0f\u5316\u8bed\u8a00\uff0c\u5e76\u5206\u6790\u4e86\u5176\u6027\u80fd\u3002GPT-f \u53d1\u73b0\u4e86\u88ab Metamath \u4e3b\u8981\u5e93\u63a5\u53d7\u7684\u65b0\u77ed\u8bc1\u660e\uff0c\u636e\u4f5c\u8005\u6240\u77e5\uff0c\u8fd9\u662f\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7cfb\u7edf\u9996\u6b21\u4e3a\u5f62\u5f0f\u6570\u5b66\u793e\u533a\u8d21\u732e\u5e76\u88ab\u91c7\u7eb3\u7684\u8bc1\u660e\u3002<\/p>\n<ul>\n<li>\u8bba\u6587 3\uff1aChain-of-Thought Prompting Elicits Reasoning in Large Language Models<\/li>\n<li>\u673a\u6784\uff1a\u8c37\u6b4c\u5927\u8111<\/li>\n<li>\u4f5c\u8005\uff1aJason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc Le, Denny Zhou<\/li>\n<li>\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/2201.11903<\/li>\n<\/ul>\n<p>\u8bba\u6587\u6982\u8ff0\uff1a\u8fd9\u7bc7\u8bba\u6587\u53d1\u5e03\u4e8e 2022 \u5e74 1 \u6708\u3002\u8bba\u6587\u63a2\u8ba8\u4e86\u5982\u4f55\u901a\u8fc7\u751f\u6210\u4e00\u7cfb\u5217\u4e2d\u95f4\u63a8\u7406\u6b65\u9aa4\uff08\u5373\u300c\u601d\u7ef4\u94fe\u300d\uff09\u6765\u663e\u8457\u63d0\u9ad8\u5927\u578b\u8bed\u8a00\u6a21\u578b\u8fdb\u884c\u590d\u6742\u63a8\u7406\u7684\u80fd\u529b\u3002\u5177\u4f53\u6765\u8bf4\uff0c\u4f5c\u8005\u63d0\u51fa\u4e86\u601d\u7ef4\u94fe\u63d0\u793a\u7684\u65b9\u6cd5\uff0c\u5373\u5728\u63d0\u793a\u4e2d\u63d0\u4f9b\u51e0\u4e2a\u601d\u7ef4\u94fe\u7684\u793a\u4f8b\uff0c\u4ee5\u6b64\u6765\u5f15\u5bfc\u6a21\u578b\u8fdb\u884c\u66f4\u6df1\u5165\u7684\u63a8\u7406\u3002\u5b9e\u9a8c\u8868\u660e\uff0c\u8fd9\u79cd\u65b9\u6cd5\u5728\u4e09\u4e2a\u5927\u578b\u8bed\u8a00\u6a21\u578b\u4e0a\u63d0\u9ad8\u4e86\u7b97\u672f\u3001\u5e38\u8bc6\u548c\u7b26\u53f7\u63a8\u7406\u4efb\u52a1\u7684\u6027\u80fd\u3002<\/p>\n<ul>\n<li>\u8bba\u6587 4\uff1aLet&#8217;s Verify Step by Step<\/li>\n<li>\u673a\u6784\uff1aOpenAI<\/li>\n<li>\u4f5c\u8005\uff1aHunter Lightman, Vineet Kosaraju, Yura Burda, Harri Edwards, Bowen Baker, Teddy Lee, Jan Leike, John Schulman, Ilya Sutskever, Karl Cobbe<\/li>\n<li>\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/2305.20050<\/li>\n<\/ul>\n<p>\u8bba\u6587\u6982\u8ff0\uff1a\u8fd9\u7bc7\u8bba\u6587\u53d1\u5e03\u4e8e 2023 \u5e74 5 \u6708\u3002\u8bba\u6587\u63a2\u8ba8\u4e86\u5927\u578b\u8bed\u8a00\u6a21\u578b\u5728\u590d\u6742\u591a\u6b65\u63a8\u7406\u4efb\u52a1\u4e2d\u7684\u8868\u73b0\u53ca\u5176\u53ef\u9760\u6027\u95ee\u9898\u3002\u4f5c\u8005\u6bd4\u8f83\u4e86\u4e24\u79cd\u8bad\u7ec3\u65b9\u6cd5\uff1a\u7ed3\u679c\u76d1\u7763\uff08outcome supervision\uff09\u548c\u8fc7\u7a0b\u76d1\u7763\uff08process supervision\uff09\uff0c\u524d\u8005\u4ec5\u5bf9\u6700\u7ec8\u7ed3\u679c\u63d0\u4f9b\u53cd\u9988\uff0c\u540e\u8005\u5219\u5bf9\u6bcf\u4e2a\u63a8\u7406\u6b65\u9aa4\u63d0\u4f9b\u53cd\u9988\u3002\u7814\u7a76\u53d1\u73b0\uff0c\u8fc7\u7a0b\u76d1\u7763\u5728\u8bad\u7ec3\u6a21\u578b\u89e3\u51b3 MATH \u6570\u636e\u96c6\u4e2d\u7684\u95ee\u9898\u65f6\uff0c\u663e\u8457\u4f18\u4e8e\u7ed3\u679c\u76d1\u7763\u3002\u5177\u4f53\u6765\u8bf4\uff0c\u91c7\u7528\u8fc7\u7a0b\u76d1\u7763\u7684\u6a21\u578b\u5728 MATH \u6d4b\u8bd5\u96c6\u7684\u4e00\u4e2a\u4ee3\u8868\u6027\u5b50\u96c6\u4e2d\u89e3\u51b3\u95ee\u9898\u7684\u6210\u529f\u7387\u4e3a 78%\u3002\u6b64\u5916\uff0c\u8bba\u6587\u8fd8\u5c55\u793a\u4e86\u4e3b\u52a8\u5b66\u4e60\uff08active learning\uff09\u5728\u63d0\u9ad8\u8fc7\u7a0b\u76d1\u7763\u6548\u7387\u65b9\u9762\u7684\u91cd\u8981\u6027\u3002\u4e3a\u4e86\u652f\u6301\u76f8\u5173\u7814\u7a76\uff0c\u4f5c\u8005\u8fd8\u53d1\u5e03\u4e86 PRM800K \u6570\u636e\u96c6\uff0c\u8fd9\u662f\u4e00\u4e2a\u5305\u542b 800,000 \u4e2a\u6b65\u9aa4\u7ea7\u4eba\u7c7b\u53cd\u9988\u6807\u7b7e\u7684\u5b8c\u6574\u6570\u636e\u96c6\uff0c\u7528\u4e8e\u8bad\u7ec3\u4ed6\u4eec\u7684\u6700\u4f73\u5956\u52b1\u6a21\u578b\u3002<\/p>\n<p>\u7531\u4e8e\u5305\u62ec Ilya \u5728\u5185\u7684\u591a\u4f4d o1 \u6838\u5fc3\u8d21\u732e\u8005\u90fd\u53c2\u4e0e\u4e86\u8fd9\u7bc7\u8bba\u6587\uff0c\u6709\u4eba\u731c\u6d4b\u8fd9\u662f o1 \u6a21\u578b\u8bad\u7ec3\u7684\u65b9\u6cd5\u8bba\u3002\u611f\u5174\u8da3\u7684\u8bfb\u8005\u53ef\u4ee5\u91cd\u70b9\u9605\u8bfb\u3002\u4e5f\u53ef\u53c2\u9605\u673a\u5668\u4e4b\u5fc3\u7684\u62a5\u9053\u300a<a data-itemshowtype=\"0\" data-linktype=\"2\" href=\"http:\/\/mp.weixin.qq.com\/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2650879157&amp;idx=1&amp;sn=eac1b887a62133667d5987ea995b49fe&amp;chksm=84e4f0cbb39379dd724f6e21344eb101980458c3dbcca5ce87f5794e24810731c31883fe31de&amp;scene=21#wechat_redirect\" target=\"_blank\">OpenAI \u8981\u4e3a GPT-4 \u89e3\u51b3\u6570\u5b66\u95ee\u9898\u4e86\uff1a\u5956\u52b1\u6a21\u578b\u6307\u9519\uff0c\u89e3\u9898\u6c34\u5e73\u8fbe\u5230\u65b0\u9ad8\u5ea6<\/a>\u300b\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-e4701ed8a1d37278004582cbd24473ea.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-e4701ed8a1d37278004582cbd24473ea.png\"><\/a><\/p>\n<ul>\n<li>\u8bba\u6587 5\uff1aLLM Critics Help Catch LLM Bugs<\/li>\n<li>\u673a\u6784\uff1aOpenAI<\/li>\n<li>\u4f5c\u8005\uff1aNat McAleese, Rai Michael Pokorny, Juan Felipe Ceron Uribe, Evgenia Nitishinskaya, Maja Trebacz, Jan Leike<\/li>\n<li>\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/2407.00215<\/li>\n<\/ul>\n<p>\u8bba\u6587\u6982\u8ff0\uff1a\u8fd9\u7bc7\u8bba\u6587\u53d1\u5e03\u4e8e 2024 \u5e74 6 \u6708\u3002\u8bba\u6587\u4ecb\u7ecd\u4e86\u4e00\u79cd\u901a\u8fc7\u8bad\u7ec3\u300c\u6279\u8bc4\u8005\u300d\u6a21\u578b\uff08\u5373\u524d\u9762\u63d0\u5230\u7684 CriticGPT \uff09\u6765\u63d0\u9ad8\u4eba\u7c7b\u8bc4\u4f30\u673a\u5668\u5b66\u4e60\u6a21\u578b\u8f93\u51fa\u7684\u65b9\u6cd5\u3002\u8fd9\u4e9b\u6279\u8bc4\u8005\u6a21\u578b\u662f\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff0c\u5b83\u4eec\u88ab\u8bad\u7ec3\u6765\u63d0\u4f9b\u81ea\u7136\u8bed\u8a00\u53cd\u9988\uff0c\u6307\u51fa\u4ee3\u7801\u4e2d\u7684\u95ee\u9898\u3002\u7814\u7a76\u8868\u660e\uff0c\u8fd9\u4e9b\u6a21\u578b\u5728\u8bc6\u522b\u4ee3\u7801\u9519\u8bef\u65b9\u9762\u6bd4\u4eba\u7c7b\u66f4\u6709\u6548\uff0c\u751a\u81f3\u80fd\u591f\u53d1\u73b0\u4eba\u7c7b\u5ba1\u67e5\u8005\u672a\u53d1\u73b0\u7684\u9519\u8bef\u3002\u5c3d\u7ba1\u5b58\u5728\u5c40\u9650\u6027\uff0c\u5982\u53ef\u80fd\u4ea7\u751f\u8bef\u5bfc\u7684\u5e7b\u89c9\u9519\u8bef\uff0c\u4f46\u7ed3\u5408\u4eba\u7c7b\u548c\u673a\u5668\u7684\u56e2\u961f\u53ef\u4ee5\u51cf\u5c11\u8fd9\u79cd\u8bef\u5bfc\uff0c\u540c\u65f6\u4fdd\u6301\u9519\u8bef\u68c0\u6d4b\u7684\u6548\u7387\u3002<\/p>\n<ul>\n<li>\u8bba\u6587 6\uff1aSelf-critiquing models for assisting human evaluators<\/li>\n<li>\u673a\u6784\uff1aOpenAI<\/li>\n<li>\u4f5c\u8005\uff1aWilliam Saunders, Catherine Yeh, Jeff Wu, Steven Bills, Long Ouyang, Jonathan Ward, Jan Leike<\/li>\n<li>\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/pdf\/2206.05802<\/li>\n<\/ul>\n<p>\u8bba\u6587\u6982\u8ff0\uff1a\u8fd9\u7bc7\u8bba\u6587\u53d1\u5e03\u4e8e 2022 \u5e74 6 \u6708\u3002\u8bba\u6587\u7814\u7a76\u4e86\u5982\u4f55\u901a\u8fc7\u5fae\u8c03\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff0c\u4f7f\u7528\u884c\u4e3a\u514b\u9686\u6765\u751f\u6210\u81ea\u7136\u8bed\u8a00\u7684\u6279\u8bc4\u6027\u8bc4\u8bba\uff0c\u4ee5\u5e2e\u52a9\u4eba\u7c7b\u53d1\u73b0\u6458\u8981\u4e2d\u7684\u7f3a\u9677\u3002\u5b9e\u9a8c\u8868\u660e\uff0c\u8fd9\u4e9b\u6a21\u578b\u751f\u6210\u7684\u8bc4\u8bba\u80fd\u591f\u63ed\u793a\u4eba\u7c7b\u548c\u673a\u5668\u751f\u6210\u6458\u8981\u4e2d\u7684\u95ee\u9898\uff0c\u5305\u62ec\u6545\u610f\u8bef\u5bfc\u7684\u9519\u8bef\u3002\u7814\u7a76\u53d1\u73b0\uff0c\u66f4\u5927\u7684\u6a21\u578b\u5728\u64b0\u5199\u6709\u5e2e\u52a9\u7684\u8bc4\u8bba\u548c\u81ea\u6211\u6279\u8bc4\u65b9\u9762\u8868\u73b0\u66f4\u597d\uff0c\u5e76\u4e14\u80fd\u591f\u5229\u7528\u81ea\u6211\u6279\u8bc4\u6765\u6539\u8fdb\u81ea\u5df1\u7684\u6458\u8981\u3002\u8bba\u6587\u8fd8\u63d0\u51fa\u4e86\u4e00\u4e2a\u6bd4\u8f83\u6279\u8bc4\u80fd\u529b\u3001\u751f\u6210\u80fd\u529b\u548c\u8fa8\u522b\u80fd\u529b\u7684\u6846\u67b6\uff0c\u5e76\u6307\u51fa\u5373\u4f7f\u662f\u5927\u578b\u6a21\u578b\u4e5f\u53ef\u80fd\u6709\u672a\u8868\u8fbe\u7684\u77e5\u8bc6\u3002\u8fd9\u9879\u7814\u7a76\u4e3a\u4f7f\u7528\u4eba\u5de5\u667a\u80fd\u8f85\u52a9\u7684\u4eba\u7c7b\u53cd\u9988\u6765\u76d1\u7763\u673a\u5668\u5b66\u4e60\u7cfb\u7edf\u63d0\u4f9b\u4e86\u6982\u5ff5\u9a8c\u8bc1\uff0c\u5e76\u516c\u5f00\u4e86\u8bad\u7ec3\u6570\u636e\u96c6\u548c\u5b9e\u9a8c\u6837\u672c\u3002<\/p>\n<p><strong>\u5176\u4ed6\u8bba\u6587<\/strong><\/p>\n<p>\u9664\u4e86\u4ee5\u4e0a\u8bba\u6587\uff0c\u4f5c\u8005\u8fd8\u6309\u7167\u5e74\u4efd\u5217\u51fa\u4e86\u4e00\u4e9b\u53ef\u80fd\u4e0e OpenAI o1 \u76f8\u5173\u7684\u8bba\u6587\uff0c\u5217\u8868\u5982\u4e0b\uff1a<\/p>\n<p>2024 \u5e74\uff1a<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-88618e0adcbb51636275cf8ca76ac0cd.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-88618e0adcbb51636275cf8ca76ac0cd.png\"><\/a><\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-ca38d5d908efdfd3c3a53a77ee9b0968.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-ca38d5d908efdfd3c3a53a77ee9b0968.png\"><\/a><\/p>\n<p>2023 \u5e74\uff1a<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-f6bcb0bffc72eca83b032bcde216d201.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-f6bcb0bffc72eca83b032bcde216d201.png\"><\/a><\/p>\n<p>2022 \u5e74\uff1a<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-c6953825b65a730047dcd4b3ce5cbf91.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-c6953825b65a730047dcd4b3ce5cbf91.png\"><\/a><\/p>\n<p>2021 \u5e74\uff1a<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-b6bfd927fd336e08937a0752aa412b40.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-b6bfd927fd336e08937a0752aa412b40.png\"><\/a><\/p>\n<p>2017 \u5e74\uff1a<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-c9498eb562bdfdd40f2fa3b42aadcd4e.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-c9498eb562bdfdd40f2fa3b42aadcd4e.png\"><\/a><\/p>\n<p>\u66f4\u591a\u4fe1\u606f\u8bf7\u53c2\u89c1\u539f GitHub \u5e93\u3002<\/p>\n<p>\u76f8\u5173\u9605\u8bfb\uff1a<\/p>\n<ul>\n<li><a data-itemshowtype=\"0\" data-linktype=\"2\" href=\"http:\/\/mp.weixin.qq.com\/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2650934313&amp;idx=1&amp;sn=3e89923751621fed57602ffe4ba751f4&amp;chksm=84e7c857b3904141d5ef61e834efeb469e7982498aca47cfe92cadcf3912908754e752793c2a&amp;scene=21#wechat_redirect\" target=\"_blank\">\u521a\u521a\uff0cOpenAI \u9707\u64bc\u53d1\u5e03 o1 \u5927\u6a21\u578b\uff01\u5f3a\u5316\u5b66\u4e60\u7a81\u7834 LLM \u63a8\u7406\u6781\u9650<\/a><\/li>\n<li><a data-itemshowtype=\"0\" data-linktype=\"2\" href=\"http:\/\/mp.weixin.qq.com\/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2650934369&amp;idx=1&amp;sn=fbe27b6d0dd333c068835a2b4204db78&amp;chksm=84e7c81fb39041092b5a8cd876698b2ce031389955db493c4656a33254fbaddb4e18a11a5fb9&amp;scene=21#wechat_redirect\" target=\"_blank\">OpenAI o1 \u6df1\u591c\u70b8\u573a\uff0c\u6838\u5fc3\u8d21\u732e\u8005\u6709\u54ea\u4e9b\uff1f\u6211\u4eec\u770b\u5230\u4e86\u5927\u91cf\u534e\u4eba\u7684\u540d\u5b57<\/a><\/li>\n<li><a data-itemshowtype=\"0\" data-linktype=\"2\" href=\"http:\/\/mp.weixin.qq.com\/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2650934463&amp;idx=3&amp;sn=ebcb6468cfdac5f21cdb1dd7ac4f8601&amp;chksm=84e7c8c1b39041d706333a7a13368770348eebf6f70d2b0d43986829a7d645e815f8416215e2&amp;scene=21#wechat_redirect\" target=\"_blank\">o1 \u6539\u53d8\u4e86 Scaling \u8303\u5f0f\uff1fSelf-Play \u8fd8\u503c\u5f97 All In \u5417\uff1f<\/a><\/li>\n<li><a data-itemshowtype=\"0\" data-linktype=\"2\" href=\"http:\/\/mp.weixin.qq.com\/s?__biz=MzA3MzI4MjgzMw==&amp;mid=2650934649&amp;idx=2&amp;sn=bc5360ff59010b92689429d5ce4f0ada&amp;chksm=84e7c907b3904011237f66077a2aefb1a683aab7a7f27f55aa1e8f08cfc41c495b2948c7ebe8&amp;scene=21#wechat_redirect\" target=\"_blank\">OpenAI o1 \u667a\u5546 120\uff0c\u8fd8\u662f\u88ab\u9676\u54f2\u8f69\u79f0\u4e3a\u300c\u5e73\u5eb8\u7684\u7814\u7a76\u751f\u300d\uff0c\u4f46\u5b9e\u529b\u7a76\u7adf\u5982\u4f55\uff1f<\/a><\/li>\n<\/ul>\n<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-09-17\" target=\"_blank\">OpenAI o1\u8981\u8ddf\uff0c\u600e\u4e48\u8ddf\uff1f\u8fd9\u4e2aGitHub\u9879\u76ee\u628a\u89e3\u8bfb\u3001\u535a\u5ba2\u3001\u76f8\u5173\u8bba\u6587\u4e00\u7f51\u6253\u5c3d<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:OpenAI o1\u8981\u8ddf [&hellip;]<\/p>\n","protected":false},"author":3,"featured_media":0,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[27],"tags":[68,55,65],"class_list":["post-35648","post","type-post","status-publish","format-standard","hentry","category-news","tag-68","tag-55","tag-65"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v26.4 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>OpenAI o1\u8981\u8ddf\uff0c\u600e\u4e48\u8ddf\uff1f\u8fd9\u4e2aGitHub\u9879\u76ee\u628a\u89e3\u8bfb\u3001\u535a\u5ba2\u3001\u76f8\u5173\u8bba\u6587\u4e00\u7f51\u6253\u5c3d - \u4e00\u8d77AI\u6280\u672f<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/17aitech.com\/?p=35648\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/17aitech.com\/?p=35648\",\"url\":\"https:\/\/17aitech.com\/?p=35648\",\"name\":\"OpenAI o1\u8981\u8ddf\uff0c\u600e\u4e48\u8ddf\uff1f\u8fd9\u4e2aGitHub\u9879\u76ee\u628a\u89e3\u8bfb\u3001\u535a\u5ba2\u3001\u76f8\u5173\u8bba\u6587\u4e00\u7f51\u6253\u5c3d - \u4e00\u8d77AI\u6280\u672f\",\"isPartOf\":{\"@id\":\"https:\/\/17aitech.com\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/17aitech.com\/?p=35648#primaryimage\"},\"image\":{\"@id\":\"https:\/\/17aitech.com\/?p=35648#primaryimage\"},\"thumbnailUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-b2f27343a5f79615937070bd54ce7fe8.png\",\"datePublished\":\"2024-12-28T16:00:13+00:00\",\"author\":{\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d\"},\"breadcrumb\":{\"@id\":\"https:\/\/17aitech.com\/?p=35648#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/17aitech.com\/?p=35648\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/17aitech.com\/?p=35648#primaryimage\",\"url\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-b2f27343a5f79615937070bd54ce7fe8.png\",\"contentUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-b2f27343a5f79615937070bd54ce7fe8.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/17aitech.com\/?p=35648#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/17aitech.com\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"OpenAI o1\u8981\u8ddf\uff0c\u600e\u4e48\u8ddf\uff1f\u8fd9\u4e2aGitHub\u9879\u76ee\u628a\u89e3\u8bfb\u3001\u535a\u5ba2\u3001\u76f8\u5173\u8bba\u6587\u4e00\u7f51\u6253\u5c3d\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/17aitech.com\/#website\",\"url\":\"https:\/\/17aitech.com\/\",\"name\":\"\u4e00\u8d77AI\u6280\u672f\",\"description\":\"\u8ba9AI\u77e5\u8bc6\u89e6\u624b\u53ef\u53ca\",\"alternateName\":\"\u4e00\u8d77AI\u6280\u672f\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/17aitech.com\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d\",\"name\":\"AI\u5c0f\u52a9\u624b\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/image\/\",\"url\":\"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png\",\"contentUrl\":\"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png\",\"caption\":\"AI\u5c0f\u52a9\u624b\"},\"description\":\"\u8fd9\u4e2a\u4eba\u5f88\u61d2\uff0c\u4ec0\u4e48\u90fd\u6ca1\u6709\u7559\u4e0b\uff5e\",\"url\":\"https:\/\/17aitech.com\/?page_id=33738&user=3\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"OpenAI o1\u8981\u8ddf\uff0c\u600e\u4e48\u8ddf\uff1f\u8fd9\u4e2aGitHub\u9879\u76ee\u628a\u89e3\u8bfb\u3001\u535a\u5ba2\u3001\u76f8\u5173\u8bba\u6587\u4e00\u7f51\u6253\u5c3d - \u4e00\u8d77AI\u6280\u672f","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/17aitech.com\/?p=35648","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/17aitech.com\/?p=35648","url":"https:\/\/17aitech.com\/?p=35648","name":"OpenAI o1\u8981\u8ddf\uff0c\u600e\u4e48\u8ddf\uff1f\u8fd9\u4e2aGitHub\u9879\u76ee\u628a\u89e3\u8bfb\u3001\u535a\u5ba2\u3001\u76f8\u5173\u8bba\u6587\u4e00\u7f51\u6253\u5c3d - \u4e00\u8d77AI\u6280\u672f","isPartOf":{"@id":"https:\/\/17aitech.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/17aitech.com\/?p=35648#primaryimage"},"image":{"@id":"https:\/\/17aitech.com\/?p=35648#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-b2f27343a5f79615937070bd54ce7fe8.png","datePublished":"2024-12-28T16:00:13+00:00","author":{"@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"breadcrumb":{"@id":"https:\/\/17aitech.com\/?p=35648#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/17aitech.com\/?p=35648"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/?p=35648#primaryimage","url":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-b2f27343a5f79615937070bd54ce7fe8.png","contentUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-b2f27343a5f79615937070bd54ce7fe8.png"},{"@type":"BreadcrumbList","@id":"https:\/\/17aitech.com\/?p=35648#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/17aitech.com\/"},{"@type":"ListItem","position":2,"name":"OpenAI o1\u8981\u8ddf\uff0c\u600e\u4e48\u8ddf\uff1f\u8fd9\u4e2aGitHub\u9879\u76ee\u628a\u89e3\u8bfb\u3001\u535a\u5ba2\u3001\u76f8\u5173\u8bba\u6587\u4e00\u7f51\u6253\u5c3d"}]},{"@type":"WebSite","@id":"https:\/\/17aitech.com\/#website","url":"https:\/\/17aitech.com\/","name":"\u4e00\u8d77AI\u6280\u672f","description":"\u8ba9AI\u77e5\u8bc6\u89e6\u624b\u53ef\u53ca","alternateName":"\u4e00\u8d77AI\u6280\u672f","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/17aitech.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d","name":"AI\u5c0f\u52a9\u624b","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/#\/schema\/person\/image\/","url":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","contentUrl":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","caption":"AI\u5c0f\u52a9\u624b"},"description":"\u8fd9\u4e2a\u4eba\u5f88\u61d2\uff0c\u4ec0\u4e48\u90fd\u6ca1\u6709\u7559\u4e0b\uff5e","url":"https:\/\/17aitech.com\/?page_id=33738&user=3"}]}},"_links":{"self":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/35648","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=35648"}],"version-history":[{"count":0,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/35648\/revisions"}],"wp:attachment":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=35648"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=35648"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=35648"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}