{"id":37859,"date":"2025-02-05T21:01:42","date_gmt":"2025-02-05T13:01:42","guid":{"rendered":"https:\/\/17aitech.com\/?p=37859"},"modified":"2025-02-05T21:01:42","modified_gmt":"2025-02-05T13:01:42","slug":"%e8%bf%bd%e5%b9%b3%e6%bb%a1%e8%a1%80%e7%89%88o1%e7%9a%84%e5%9b%bd%e4%ba%a7%e5%a4%9a%e6%a8%a1%e6%80%81%e6%a8%a1%e5%9e%8b%e7%bb%88%e4%ba%8e%e6%9d%a5%e4%ba%86%ef%bc%81%e8%ae%ad%e7%bb%83%e7%bb%86%e8%8a%82","status":"publish","type":"post","link":"https:\/\/17aitech.com\/?p=37859","title":{"rendered":"\u8ffd\u5e73\u6ee1\u8840\u7248o1\u7684\u56fd\u4ea7\u591a\u6a21\u6001\u6a21\u578b\u7ec8\u4e8e\u6765\u4e86\uff01\u8bad\u7ec3\u7ec6\u8282\u5168\u90e8\u516c\u5f00"},"content":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2025-01-21-4\" target=\"_blank\">\u8ffd\u5e73\u6ee1\u8840\u7248o1\u7684\u56fd\u4ea7\u591a\u6a21\u6001\u6a21\u578b\u7ec8\u4e8e\u6765\u4e86\uff01\u8bad\u7ec3\u7ec6\u8282\u5168\u90e8\u516c\u5f00<\/a><\/p>\n<section>\u6625\u8282\u524d\u6700\u540e\u4e00\u5468\uff0c\u80fd\u5ab2\u7f8e Open AI \u6ee1\u8840\u7248 o1\uff08Full Version\uff0c\u800c\u975e preview\uff09\u7684\u6a21\u578b\u7ec8\u4e8e\u51fa\u73b0\u4e86\uff01<\/section>\n<section><\/section>\n<section>\u521a\u521a\uff0c\u6708\u4e4b\u6697\u9762\u516c\u5e03\u4e86\u4ed6\u4eec\u7684 Kimi k \u7cfb\u5217\u6a21\u578b\u6700\u65b0\u7248\u672c \u2014\u2014<strong>k1.5 \u591a\u6a21\u6001\u601d\u8003\u6a21\u578b<\/strong>\u3002\u65b0\u6a21\u578b\u5728\u6570\u5b66\u3001\u4ee3\u7801\u3001\u591a\u6a21\u6001\u63a8\u7406\u80fd\u529b\u7b49\u65b9\u9762\u5168\u9762\u5bf9\u6807 Open AI \u6ee1\u8840\u7248 o1\uff0c\u800c\u4e14\u662f OpenAI \u4e4b\u5916\u9996\u4e2a\u591a\u6a21\u6001 o1\u3002\u5c24\u5176\u662f kimi-k1.5-short\uff0c\u6210\u4e3a SOTA short cot \u6a21\u578b\uff0c\u5e76\u5927\u5e45\u9886\u5148\u00a0GPT-4o \u548c Claude 3.5 Sonnet\uff08\u63d0\u5347\u5e45\u5ea6\u9ad8\u8fbe 550%\uff09<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-223119dffedeb8fdec331c8c3d3dc670.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-223119dffedeb8fdec331c8c3d3dc670.png\"><\/a><\/section>\n<section><strong>\u8fd9\u662f Open AI \u4e4b\u5916\uff0c\u9996\u6b21\u6709\u6a21\u578b\u5728\u6570\u5b66\u548c\u4ee3\u7801\u80fd\u529b\u4e0a\u8fbe\u5230\u6ee1\u8840 o1\uff0c\u6708\u4e4b\u6697\u9762\u4e5f\u662f\u56fd\u5185\u7b2c\u4e00\u4e2a\u8fbe\u5230\u8be5\u6c34\u5e73\u7684 AI \u516c\u53f8<\/strong>\u3002\u5728\u6b64\u4e4b\u524d\uff0c\u90e8\u5206\u6a21\u578b\u5728\u5404\u7c7b Benchmark \u4e0a\u53ef\u4ee5\u8fbe\u5230 50 \u5206\u300160 \u5206\u7684\u6c34\u5e73\uff08\u76f8\u5f53\u4e8e o1-preview\uff09\uff0c\u800c o1 \u6ee1\u8840\u7248\u662f 80 \u5206\u300190 \u5206\u6c34\u5e73\uff0cKimi k1.5 \u7684\u6210\u7ee9\u4ee4\u4eba\u773c\u524d\u4e00\u4eae\u3002<\/section>\n<section><\/section>\n<section>\u8fd9\u4e00\u5207\u662f\u600e\u4e48\u505a\u5230\u7684\u5462\uff1f\u5728 Kimi \u6280\u672f\u56e2\u961f\u540c\u6b65\u53d1\u5e03\u7684\u6280\u672f\u62a5\u544a\u4e2d\uff0c\u6211\u4eec\u53ef\u4ee5\u770b\u5230\u4ed6\u4eec\u5728\u65b0\u6280\u672f\u8303\u5f0f\u4e0b\u7684\u6a21\u578b\u8bad\u7ec3\u6280\u672f\u63a2\u7d22\u4e4b\u8def\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-3735014ae469ed4e99407f5b12323167.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-3735014ae469ed4e99407f5b12323167.png\"><\/a><\/section>\n<ul>\n<li>\n<section>\u6280\u672f\u62a5\u544a\uff1aKimi k1.5\uff1a\u501f\u52a9\u5927\u8bed\u8a00\u6a21\u578b\u5b9e\u73b0\u5f3a\u5316\u5b66\u4e60\u7684 Scaling<\/section>\n<\/li>\n<li>\n<section>\u62a5\u544a\u94fe\u63a5\uff1ahttps:\/\/github.com\/MoonshotAI\/kimi-k1.5<\/section>\n<\/li>\n<\/ul>\n<section><\/section>\n<section>\u8fd9\u79cd\u6280\u672f\u900f\u660e\u5ea6\u5728\u5f53\u524d\u7ade\u4e89\u6fc0\u70c8\u7684\u5927\u6a21\u578b\u5e02\u573a\u4e0a\u5e76\u4e0d\u591a\u89c1\u3002\u5728\u8c08\u53ca\u4e3a\u4ec0\u4e48\u8981\u8fd9\u4e48\u505a\u65f6\uff0c\u6708\u4e4b\u6697\u9762\u8868\u793a\uff0c\u300c\u56e0\u4e3a\u6211\u4eec\u610f\u8bc6\u5230\uff0cAGI \u4e4b\u65c5\u624d\u521a\u521a\u5f00\u59cb\u3002\u6211\u4eec\u60f3\u8ba9\u66f4\u591a\u6280\u672f\u4eba\u624d\u4e86\u89e3\u6211\u4eec\u5728\u505a\u7684\u4e8b\u60c5\uff0c\u52a0\u5165\u6211\u4eec\u4e00\u8d77\u505a\u5230\u66f4\u591a\u300d\u3002<\/section>\n<section><\/section>\n<section><strong>Kimi k1.5 \u591a\u9879\u6d4b\u8bd5\uff0c\u5168\u90e8 SOTA<\/strong><\/section>\n<section><\/section>\n<section>\u4ece\u6280\u672f\u62a5\u544a\u6765\u770b\uff0cKimi k1.5 \u591a\u6a21\u6001\u63a8\u7406\u6a21\u578b\u5b9e\u73b0\u4e86 SOTA \uff08state-of-the-art\uff09\u7ea7\u522b\u7684\u63a8\u7406\u548c\u901a\u7528\u80fd\u529b\uff0c\u5177\u4f53\u800c\u8a00\uff1a<\/section>\n<section>\u00a0<\/section>\n<section>\u5728 long-CoT \u6a21\u5f0f\u4e0b\uff0cKimi k1.5 \u5728\u6570\u5b66\u3001\u4ee3\u7801\u53ca\u591a\u6a21\u6001\u63a8\u7406\u80fd\u529b\u4e0a\uff0c\u8fbe\u5230\u957f\u601d\u8003 SOTA \u6a21\u578b OpenAI o1 \u6b63\u5f0f\u7248\u7684\u6c34\u5e73\u3002Kimi k1.5 \u5728 AIME \u4e0a\u8fbe\u5230 77.5 \u5206\uff0c\u5728 MATH 500 \u4e0a\u8fbe\u5230 96.2 \u5206\uff0c\u5728 Codeforces \u4e0a\u8fbe\u5230 94 \u767e\u5206\u4f4d\uff0c\u5728 MathVista \u4e0a\u8fbe\u5230 74.9 \u5206\u3002<\/section>\n<section><\/section>\n<section>\u8fd9\u5e94\u8be5\u662f\u5168\u7403\u8303\u56f4\u5185\uff0cOpenAI \u4e4b\u5916\u7684\u516c\u53f8\u9996\u6b21\u5b9e\u73b0 o1 \u6ee1\u8840\u7248\u6027\u80fd\u3002\u6b64\u524d\u7684\u6a21\u578b\u53ea\u80fd\u8fbe\u5230 o1-preview \u6216 o1-mini \u7684\u63a8\u7406\u80fd\u529b\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-12ad1b6baaef7b8a64b5facd9d02e847.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-12ad1b6baaef7b8a64b5facd9d02e847.png\"><\/a><\/section>\n<section>\u5728 short-CoT \u6a21\u5f0f\u4e0b\uff0cKimi k1.5 \u5728\u6570\u5b66\u3001\u4ee3\u7801\u3001\u89c6\u89c9\u591a\u6a21\u6001\u548c\u901a\u7528\u80fd\u529b\u4e0a\uff0c\u4e5f\u8fbe\u5230\u4e86\u5168\u7403\u8303\u56f4\u5185\u77ed\u601d\u8003 SOTA \u6a21\u578b \uff0c\u5e76\u5927\u5e45\u9886\u5148 GPT-4o \u548c Claude 3.5 Sonnet \u7684\u6c34\u5e73\u3002\u6bd4\u5982\uff0cKimi k1.5 \u5728 AIME \u4e0a\u8fbe\u5230 60.8 \u5206\uff0cMATH500 \u4e0a\u8fbe\u5230 94.6 \u5206\uff0cLiveCodeBench \u4e0a\u8fbe\u5230 47.3 \u5206\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-96570c28efc7ce059c03321dcf3227cf.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-96570c28efc7ce059c03321dcf3227cf.png\"><\/a><\/section>\n<section>\u4e0d\u4ec5\u5982\u6b64\uff0c\u4ece\u5168\u7403\u524d\u6cbf\u5927\u6a21\u578b\u6570\u5b66\u7ade\u8d5b\u548c\u7f16\u7a0b\u7ade\u8d5b\u57fa\u51c6\u6d4b\u8bd5\u6765\u770b\uff0cKimi k1.5 \u7684\u8868\u73b0\u4e5f\u76f8\u5f53\u4e0d\u9519\uff0c\u5904\u4e8e\u5168\u7403\u7b2c\u4e00\u68af\u961f\uff0c\u800c\u8fd9\u4e24\u9879\u6d4b\u8bd5\u4ee3\u8868\u4e86\u4eba\u7c7b\u667a\u5546\u5dc5\u5cf0\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-0aac08d4891fa588b22209cfd06ef097.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-0aac08d4891fa588b22209cfd06ef097.png\"><\/a><\/section>\n<section>\u603b\u4e4b\uff0c\u4ece Benchmark \u6570\u636e\u6765\u770b\uff0ck1.5 \u7684\u63a8\u7406\u80fd\u529b\u5b9e\u73b0\u4e86\u5f88\u5927\u63d0\u5347\uff0c\u53ef\u4ee5\u5e2e\u52a9\u6211\u4eec\u89e3\u9501\u66f4\u96be\u7684\u4ee3\u7801\u3001\u6570\u5b66\u3001\u751f\u6d3b\u7b49\u95ee\u9898\u3002<\/section>\n<section><\/section>\n<section><strong>Kimi k1.5 \u662f\u600e\u4e48\u7ec3\u6210\u7684\uff1f \u00a0<\/strong><\/section>\n<section><\/section>\n<section>\u968f\u7740\u6a21\u578b\u5c3a\u5bf8\u9010\u6e10\u589e\u5927\uff0c\u9884\u8bad\u7ec3\u9636\u6bb5\u53c2\u6570 scaling up \u5e26\u6765\u7684\u8fb9\u9645\u6536\u76ca\u5f00\u59cb\u9012\u51cf\uff0c\u5982\u679c\u60f3\u8981\u6df1\u5ea6\u63d0\u5347\u6a21\u578b\u63a8\u7406\u80fd\u529b\u548c\u957f\u7a0b\u95ee\u9898\u80fd\u529b\uff0c\u57fa\u4e8e\u5f3a\u5316\u5b66\u4e60\u7684 Post-Training \u5c06\u4f1a\u6210\u4e3a\u4e0b\u4e00\u4e2a\u7a81\u7834\u70b9 [1]\uff0c\u56e0\u4e3a scaling \u5f3a\u5316\u5b66\u4e60\u4e3a\u4eba\u5de5\u667a\u80fd\u7684\u6301\u7eed\u8fdb\u6b65\u5f00\u8f9f\u4e86\u65b0\u7684\u7ef4\u5ea6\uff0c\u5b83\u4f7f\u5f97\u5927\u8bed\u8a00\u6a21\u578b\u80fd\u591f\u901a\u8fc7\u5e26\u6709\u5956\u52b1\u7684\u63a2\u7d22\u5b66\u4e60\u6765\u6269\u5c55\u5176\u8bad\u7ec3\u6570\u636e\uff0c\u4ece\u800c\u4e5f\u5b9e\u73b0\u8ba1\u7b97\u89c4\u6a21\u7684\u6269\u5c55\u3002<\/section>\n<section><\/section>\n<section>\u5927\u7684\u65b9\u5411\u975e\u5e38\u660e\u786e\uff0c\u7136\u800c\uff0c\u6b64\u524d\u53d1\u8868\u7684\u7814\u7a76\u5de5\u4f5c\u5c1a\u672a\u4ea7\u751f\u5177\u6709\u7ade\u4e89\u529b\u7684\u7ed3\u679c\u3002<\/section>\n<section><\/section>\n<section>\u6709\u9274\u4e8e\u6b64\uff0cKimi \u6280\u672f\u56e2\u961f\u5728 Kimi k1.5 \u7684\u8bad\u7ec3\u5b9e\u8df5\u4e2d\u5168\u9762\u63a2\u7d22\u4e86 RL \u8bad\u7ec3\u6280\u672f\u3001\u591a\u6a21\u6001\u6570\u636e\u914d\u65b9\u548c\u57fa\u7840\u8bbe\u65bd\u4f18\u5316\u3002<\/section>\n<section><\/section>\n<section>\u96be\u5f97\u7684\u662f\uff0c\u4ed6\u4eec\u63a2\u7d22\u51fa\u7684 RL \u6846\u67b6\u7b80\u5355\u3001\u6709\u6548\uff0c\u65e0\u9700\u4f9d\u8d56\u8499\u7279\u5361\u6d1b\u6811\u641c\u7d22\u3001\u4ef7\u503c\u51fd\u6570\u548c\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b\u7b49\u66f4\u590d\u6742\u7684\u6280\u672f\u4e5f\u80fd\u53d6\u5f97\u4f18\u5f02\u7684\u6027\u80fd\u3002<\/section>\n<section><\/section>\n<section>\u6b64\u5916\uff0c\u4ed6\u4eec\u8fd8\u63d0\u51fa\u4e86\u6709\u6548\u7684 long2short \u6280\u672f\uff0c\u5229\u7528 Long-CoT \u6280\u672f\u6765\u6539\u8fdb Short-CoT \u6a21\u578b\uff0c\u4f7f\u5f97\u6a21\u578b\u5728\u77ed\u94fe\u601d\u7ef4\u63a8\u7406\u65b9\u9762\u53d6\u5f97\u4e86\u6700\u4f73\u6210\u679c\u3002<\/section>\n<section><\/section>\n<section><strong>\u7b80\u5355\u3001\u6709\u6548\u7684 RL \u6846\u67b6<\/strong><\/section>\n<section><\/section>\n<section>Kimi \u6280\u672f\u56e2\u961f\u8bbe\u8ba1\u7684\u7b80\u5355\u800c\u6709\u6548\u7684 RL \u6846\u67b6\u79bb\u4e0d\u5f00\u4e24\u4e2a\u5173\u952e\u8981\u7d20\uff1a<strong>\u957f\u4e0a\u4e0b\u6587 scaling \u548c\u6539\u8fdb\u7684\u7b56\u7565\u4f18\u5316<\/strong>\u3002<\/section>\n<section><\/section>\n<section>\u5148\u8bf4\u957f\u4e0a\u4e0b\u6587 scaling\u3002\u4ed6\u4eec\u5c06\u5f3a\u5316\u5b66\u4e60\u7684\u4e0a\u4e0b\u6587\u7a97\u53e3 scale \u5230 128k\uff0c\u5e76\u89c2\u5bdf\u5230\u968f\u7740\u4e0a\u4e0b\u6587\u957f\u5ea6\u7684\u589e\u52a0\uff0c\u6a21\u578b\u6027\u80fd\u6301\u7eed\u6539\u5584\u3002\u65b0\u65b9\u6cd5\u80cc\u540e\u7684\u4e00\u4e2a\u5173\u952e\u7406\u5ff5\u662f\u4f7f\u7528 partial rollout \u6765\u63d0\u9ad8\u8bad\u7ec3\u6548\u7387 \u2014\u2014 \u5373\u901a\u8fc7\u91cd\u7528\u5927\u91cf\u4ee5\u524d\u7684\u8f68\u8ff9\u6765\u91c7\u6837\u65b0\u7684\u8f68\u8ff9\uff0c\u907f\u514d\u4ece\u5934\u91cd\u65b0\u751f\u6210\u65b0\u8f68\u8ff9\u7684\u6210\u672c\u3002\u6280\u672f\u56e2\u961f\u7684\u89c2\u5bdf\u8868\u660e\uff0c\u4e0a\u4e0b\u6587\u957f\u5ea6\u662f\u5927\u8bed\u8a00\u6a21\u578b\u5f3a\u5316\u5b66\u4e60\u6301\u7eed scaling \u7684\u4e00\u4e2a\u5173\u952e\u7ef4\u5ea6\u3002\u00a0<\/section>\n<section><\/section>\n<section>\u518d\u6765\u770b\u7b56\u7565\u4f18\u5316\u7684\u6539\u8fdb\u3002\u4ed6\u4eec\u63a8\u5bfc\u51fa\u4e86\u4e00\u4e2a\u5177\u6709 long-CoT \u7684\u5f3a\u5316\u5b66\u4e60\u516c\u5f0f\uff0c\u5e76\u91c7\u7528\u5728\u7ebf\u955c\u50cf\u4e0b\u964d\u6cd5\u7684\u53d8\u4f53\u6765\u5b9e\u73b0\u7a33\u5065\u7684\u7b56\u7565\u4f18\u5316\u3002\u901a\u8fc7\u6709\u6548\u7684\u91c7\u6837\u7b56\u7565\u3001\u957f\u5ea6\u60e9\u7f5a\u548c\u6570\u636e\u914d\u65b9\u7684\u4f18\u5316\uff0c\u4ed6\u4eec\u8fdb\u4e00\u6b65\u6539\u8fdb\u4e86\u8be5\u7b97\u6cd5\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-98e6e3242aee9021f93ee30cd8ca1790.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-98e6e3242aee9021f93ee30cd8ca1790.png\"><\/a><\/section>\n<section>\u901a\u8fc7\u5c06\u8fd9\u4e24\u4e2a\u5173\u952e\u8981\u7d20\u7ed3\u5408\uff0cKimi \u6280\u672f\u56e2\u961f\u5efa\u7acb\u4e86\u4e00\u4e2a\u7528\u4e8e LLM \u5b66\u4e60\u7684\u7b80\u5316\u5f3a\u5316\u5b66\u4e60\u6846\u67b6\u3002\u7531\u4e8e\u8be5\u6846\u67b6\u80fd\u591f scale \u4e0a\u4e0b\u6587\u957f\u5ea6\uff0c\u5b66\u4e60\u5230\u7684 CoT \u5c55\u73b0\u51fa\u89c4\u5212\u3001\u53cd\u601d\u548c\u7ea0\u6b63\u7684\u7279\u6027\u3002\u589e\u52a0\u7684\u4e0a\u4e0b\u6587\u957f\u5ea6\u5177\u6709\u589e\u52a0\u641c\u7d22\u6b65\u9aa4\u6570\u91cf\u7684\u6548\u679c\u3002\u56e0\u6b64\uff0c\u4ed6\u4eec\u8868\u660e\u65e0\u9700\u4f9d\u8d56\u8499\u7279\u5361\u6d1b\u6811\u641c\u7d22\u3001\u4ef7\u503c\u51fd\u6570\u548c\u8fc7\u7a0b\u5956\u52b1\u6a21\u578b\u7b49\u66f4\u590d\u6742\u7684\u6280\u672f\u4e5f\u80fd\u5b9e\u73b0\u5f3a\u5927\u7684\u6027\u80fd\u3002\u00a0<\/section>\n<section><\/section>\n<section>\u6b64\u5916\uff0c\u4ed6\u4eec\u7684\u6a21\u578b\u8fd8\u5728\u6587\u672c\u548c\u89c6\u89c9\u6570\u636e\u4e0a\u8fdb\u884c\u4e86\u8054\u5408\u8bad\u7ec3\uff0c\u5177\u5907\u5bf9\u8fd9\u4e24\u79cd\u6a21\u6001\u8fdb\u884c\u8054\u5408\u63a8\u7406\u7684\u80fd\u529b\u3002\u00a0<\/section>\n<section><\/section>\n<section><strong>long2short \u6280\u672f<\/strong><\/section>\n<section><\/section>\n<section>\u5c3d\u7ba1 long-CoT \u6a21\u578b\u5728\u6027\u80fd\u4e0a\u8868\u73b0\u51fa\u8272\uff0c\u4f46\u4e0e\u6807\u51c6\u7684 short-CoT LLM \u76f8\u6bd4\uff0c\u5b83\u5728\u6d4b\u8bd5\u65f6\u6d88\u8017\u7684 token \u6570\u91cf\u66f4\u591a\u3002\u7136\u800c\uff0cKimi \u6280\u672f\u56e2\u961f\u53d1\u73b0\u5c06 long-CoT \u6a21\u578b\u7684\u601d\u7ef4\u5148\u9a8c\u8fc1\u79fb\u5230 short-CoT \u6a21\u578b\u4e2d\u662f\u53ef\u80fd\u7684\uff0c\u4ece\u800c\u5728\u6709\u9650\u7684\u6d4b\u8bd5 token \u9884\u7b97\u4e0b\u63d0\u5347\u6027\u80fd\u3002<\/section>\n<section><\/section>\n<section>\u4ed6\u4eec\u63d0\u51fa\u4e86\u51e0\u79cd\u89e3\u51b3\u8fd9\u4e00 long2short \u95ee\u9898\u7684\u65b9\u6cd5\uff0c\u5305\u62ec\u6a21\u578b\u878d\u5408\u3001\u6700\u77ed\u62d2\u7edd\u91c7\u6837\u3001DPO \u4ee5\u53ca long2short RL\u3002\u4ee5\u4e0b\u662f\u8fd9\u4e9b\u65b9\u6cd5\u7684\u8be6\u7ec6\u63cf\u8ff0\uff1a<\/section>\n<section><\/section>\n<section>\u6a21\u578b\u878d\u5408\u3002\u56e2\u961f\u4eba\u5458\u53d1\u73b0\u6a21\u578b\u878d\u5408\uff08Model Merging\uff09\u6709\u52a9\u4e8e\u4fdd\u6301\u6a21\u578b\u7684\u6cdb\u5316\u80fd\u529b\u3002\u4ed6\u4eec\u8fd8\u53d1\u73b0\uff0c\u5728\u878d\u5408 long-CoT \u6a21\u578b\u548c short-CoT \u6a21\u578b\u65f6\uff0c\u6a21\u578b\u878d\u5408\u4e5f\u80fd\u6709\u6548\u63d0\u5347 token \u6548\u7387\u3002\u8fd9\u79cd\u65b9\u6cd5\u901a\u8fc7\u5c06 long-CoT \u6a21\u578b\u4e0e short-CoT \u6a21\u578b\u7ed3\u5408\uff0c\u4ece\u800c\u5728\u4e0d\u8fdb\u884c\u8bad\u7ec3\u7684\u60c5\u51b5\u4e0b\u83b7\u5f97\u4e00\u4e2a\u65b0\u6a21\u578b\u3002\u5177\u4f53\u6765\u8bf4\uff0c\u4ed6\u4eec\u901a\u8fc7\u7b80\u5355\u5730\u5e73\u5747\u4e24\u4e2a\u6a21\u578b\u7684\u6743\u91cd\u6765\u5b9e\u73b0\u878d\u5408\u3002<\/section>\n<section><\/section>\n<section>\u6700\u77ed\u62d2\u7edd\u91c7\u6837\u3002\u7814\u7a76\u8005\u89c2\u5bdf\u5230\uff0c\u6a21\u578b\u5728\u56de\u7b54\u76f8\u540c\u95ee\u9898\u65f6\u751f\u6210\u7684\u54cd\u5e94\u957f\u5ea6\u5b58\u5728\u8f83\u5927\u5dee\u5f02\u3002\u57fa\u4e8e\u6b64\uff0c\u4ed6\u4eec\u8bbe\u8ba1\u4e86\u6700\u77ed\u62d2\u7edd\u91c7\u6837\uff08Shortest Rejection Sampling\uff09\u65b9\u6cd5\u3002\u8be5\u65b9\u6cd5\u5bf9\u540c\u4e00\u4e2a\u95ee\u9898\u91c7\u6837 n \u6b21\uff08\u5b9e\u9a8c\u4e2d\uff0cn=8\uff09\uff0c\u5e76\u9009\u62e9\u6700\u77ed\u7684\u6b63\u786e\u54cd\u5e94\u8fdb\u884c\u76d1\u7763\u5fae\u8c03\u3002<\/section>\n<section><\/section>\n<section>DPO\u3002\u4e0e\u6700\u77ed\u62d2\u7edd\u91c7\u6837\u7c7b\u4f3c\uff0c\u56e2\u961f\u4eba\u5458\u5229\u7528 Long CoT \u6a21\u578b\u751f\u6210\u591a\u4e2a\u54cd\u5e94\u6837\u672c\u3002\u5e76\u9009\u62e9\u6700\u77ed\u7684\u6b63\u786e\u89e3\u51b3\u65b9\u6848\u4f5c\u4e3a\u6b63\u6837\u672c\uff0c\u800c\u8f83\u957f\u7684\u54cd\u5e94\u5219\u88ab\u89c6\u4e3a\u8d1f\u6837\u672c\uff0c\u5305\u62ec\u9519\u8bef\u7684\u8f83\u957f\u54cd\u5e94\u548c\u6b63\u786e\u7684\u8f83\u957f\u54cd\u5e94\u3002\u8fd9\u4e9b\u6b63\u8d1f\u6837\u672c\u5bf9\u6784\u6210\u4e86\u7528\u4e8e DPO \u8bad\u7ec3\u7684\u6210\u5bf9\u504f\u597d\u6570\u636e\u3002<\/section>\n<section><\/section>\n<section>Long2short RL\u3002\u5728\u6807\u51c6\u7684 RL \u8bad\u7ec3\u9636\u6bb5\u4e4b\u540e\uff0c\u56e2\u961f\u4eba\u5458\u9009\u62e9\u4e00\u4e2a\u5728\u6027\u80fd\u548c token \u6548\u7387\u4e4b\u95f4\u8fbe\u5230\u6700\u4f73\u5e73\u8861\u7684\u6a21\u578b\u4f5c\u4e3a\u57fa\u7840\u6a21\u578b\uff0c\u5e76\u8fdb\u884c\u5355\u72ec\u7684 long2short RL \u8bad\u7ec3\u9636\u6bb5\u3002\u5728\u8fd9\u4e2a\u7b2c\u4e8c\u9636\u6bb5\u4e2d\uff0c\u4ed6\u4eec\u8fd8\u5e94\u7528\u4e86\u957f\u5ea6\u60e9\u7f5a\u673a\u5236\uff0c\u4ece\u800c\u663e\u8457\u51cf\u5c11\u6700\u5927 rollout \u957f\u5ea6\uff0c\u4ee5\u8fdb\u4e00\u6b65\u60e9\u7f5a\u90a3\u4e9b\u8d85\u51fa\u671f\u671b\u957f\u5ea6\u4f46\u53ef\u80fd\u6b63\u786e\u7684\u54cd\u5e94\u3002<\/section>\n<section><\/section>\n<section>\u9664\u4e86\u4ee5\u4e0a\u8fd9\u4e9b\uff0cKimi k1.5 \u7684\u6280\u672f\u62a5\u544a\u8fd8\u900f\u9732\u4e86\u5f88\u591a\u4fe1\u606f\u3002\u611f\u5174\u8da3\u7684\u8bfb\u8005\u53ef\u4ee5\u53bb\u9605\u8bfb\u539f\u6587\u3002<\/section>\n<section><\/section>\n<section><strong>2025\uff1a\u52a0\u901f\u5347\u7ea7 k \u7cfb\u5217\u5f3a\u5316\u5b66\u4e60\u6a21\u578b<\/strong><\/section>\n<section><\/section>\n<section>OpenAI \u4e8e 2024 \u5e74 5 \u6708\u30019 \u6708\u63a8\u51fa\u7684 GPT-4o\u3001o1 \u4e24\u4e2a\u6a21\u578b\uff0c\u5206\u522b\u4ee3\u8868\u4e86\u591a\u6a21\u6001\u7406\u89e3\u3001\u5f3a\u5316\u5b66\u4e60\u4e24\u6761\u6280\u672f\u8def\u7ebf\u3002\u5728\u8fd9\u4e24\u6761\u8def\u7ebf\u4e0a\uff0c\u56fd\u5185 AI \u516c\u53f8\u90fd\u5728\u9646\u7eed\u53d1\u529b\uff0c\u5e76\u5728\u6700\u8fd1\u5c55\u5f00\u4e86\u6fc0\u70c8\u7ade\u4e89\u3002\u5982\u4eca\uff0cKimi \u6a21\u578b\u5728\u80fd\u529b\u4e0a\u6700\u63a5\u8fd1 o1\uff0c\u8fd9\u8ba9\u5916\u754c\u5bf9\u8fd9\u5bb6\u516c\u53f8\u5728 2025 \u5e74\u7684\u8868\u73b0\u5145\u6ee1\u4e86\u671f\u5f85\u3002<\/section>\n<section><\/section>\n<section>\u6708\u4e4b\u6697\u9762\u8868\u793a\uff0c2025 \u5e74\uff0c\u4ed6\u4eec\u4f1a\u7ee7\u7eed\u52a0\u901f\u5347\u7ea7 k \u7cfb\u5217\u5f3a\u5316\u5b66\u4e60\u6a21\u578b\uff0c\u5e26\u6765\u66f4\u591a\u6a21\u6001\u3001\u66f4\u591a\u9886\u57df\u7684\u80fd\u529b\u548c\u66f4\u5f3a\u7684\u901a\u7528\u80fd\u529b\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-b646a76fbf244a2af0118599dfbb871b.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-b646a76fbf244a2af0118599dfbb871b.png\"><\/a><\/section>\n<section>\u6211\u4eec\u4e5f\u671f\u5f85\u65b0\u6a21\u578b\u7684\u65e9\u65e5\u4e0a\u7ebf\uff01<\/section>\n<section><\/section>\n<section><em><sup>\u53c2\u8003\u94fe\u63a5\uff1a[1] https:\/\/mp.weixin.qq.com\/s\/FXGdJA8OyZvLl89rXJiyAQ<\/sup><\/em><\/section>\n<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2025-01-21-4\" target=\"_blank\">\u8ffd\u5e73\u6ee1\u8840\u7248o1\u7684\u56fd\u4ea7\u591a\u6a21\u6001\u6a21\u578b\u7ec8\u4e8e\u6765\u4e86\uff01\u8bad\u7ec3\u7ec6\u8282\u5168\u90e8\u516c\u5f00<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:\u8ffd\u5e73\u6ee1\u8840\u7248o1\u7684\u56fd\u4ea7\u591a [&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":[73,68,81],"class_list":["post-37859","post","type-post","status-publish","format-standard","hentry","category-news","tag-73","tag-68","tag-81"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v26.4 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>\u8ffd\u5e73\u6ee1\u8840\u7248o1\u7684\u56fd\u4ea7\u591a\u6a21\u6001\u6a21\u578b\u7ec8\u4e8e\u6765\u4e86\uff01\u8bad\u7ec3\u7ec6\u8282\u5168\u90e8\u516c\u5f00 - \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=37859\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/17aitech.com\/?p=37859\",\"url\":\"https:\/\/17aitech.com\/?p=37859\",\"name\":\"\u8ffd\u5e73\u6ee1\u8840\u7248o1\u7684\u56fd\u4ea7\u591a\u6a21\u6001\u6a21\u578b\u7ec8\u4e8e\u6765\u4e86\uff01\u8bad\u7ec3\u7ec6\u8282\u5168\u90e8\u516c\u5f00 - \u4e00\u8d77AI\u6280\u672f\",\"isPartOf\":{\"@id\":\"https:\/\/17aitech.com\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/17aitech.com\/?p=37859#primaryimage\"},\"image\":{\"@id\":\"https:\/\/17aitech.com\/?p=37859#primaryimage\"},\"thumbnailUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-223119dffedeb8fdec331c8c3d3dc670.png\",\"datePublished\":\"2025-02-05T13:01:42+00:00\",\"author\":{\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d\"},\"breadcrumb\":{\"@id\":\"https:\/\/17aitech.com\/?p=37859#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/17aitech.com\/?p=37859\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/17aitech.com\/?p=37859#primaryimage\",\"url\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-223119dffedeb8fdec331c8c3d3dc670.png\",\"contentUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-223119dffedeb8fdec331c8c3d3dc670.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/17aitech.com\/?p=37859#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/17aitech.com\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u8ffd\u5e73\u6ee1\u8840\u7248o1\u7684\u56fd\u4ea7\u591a\u6a21\u6001\u6a21\u578b\u7ec8\u4e8e\u6765\u4e86\uff01\u8bad\u7ec3\u7ec6\u8282\u5168\u90e8\u516c\u5f00\"}]},{\"@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":"\u8ffd\u5e73\u6ee1\u8840\u7248o1\u7684\u56fd\u4ea7\u591a\u6a21\u6001\u6a21\u578b\u7ec8\u4e8e\u6765\u4e86\uff01\u8bad\u7ec3\u7ec6\u8282\u5168\u90e8\u516c\u5f00 - \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=37859","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/17aitech.com\/?p=37859","url":"https:\/\/17aitech.com\/?p=37859","name":"\u8ffd\u5e73\u6ee1\u8840\u7248o1\u7684\u56fd\u4ea7\u591a\u6a21\u6001\u6a21\u578b\u7ec8\u4e8e\u6765\u4e86\uff01\u8bad\u7ec3\u7ec6\u8282\u5168\u90e8\u516c\u5f00 - \u4e00\u8d77AI\u6280\u672f","isPartOf":{"@id":"https:\/\/17aitech.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/17aitech.com\/?p=37859#primaryimage"},"image":{"@id":"https:\/\/17aitech.com\/?p=37859#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-223119dffedeb8fdec331c8c3d3dc670.png","datePublished":"2025-02-05T13:01:42+00:00","author":{"@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"breadcrumb":{"@id":"https:\/\/17aitech.com\/?p=37859#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/17aitech.com\/?p=37859"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/?p=37859#primaryimage","url":"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-223119dffedeb8fdec331c8c3d3dc670.png","contentUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/01\/frc-223119dffedeb8fdec331c8c3d3dc670.png"},{"@type":"BreadcrumbList","@id":"https:\/\/17aitech.com\/?p=37859#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/17aitech.com\/"},{"@type":"ListItem","position":2,"name":"\u8ffd\u5e73\u6ee1\u8840\u7248o1\u7684\u56fd\u4ea7\u591a\u6a21\u6001\u6a21\u578b\u7ec8\u4e8e\u6765\u4e86\uff01\u8bad\u7ec3\u7ec6\u8282\u5168\u90e8\u516c\u5f00"}]},{"@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\/37859","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=37859"}],"version-history":[{"count":0,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/37859\/revisions"}],"wp:attachment":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=37859"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=37859"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=37859"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}