SAGE-RL\uff1a\u53d1\u6325\u5546\u4e1a\u4ef7\u503c<\/strong><\/p>\nSAGE\u7b97\u6cd5\u786e\u5b9e\u975e\u5e38\u597d\u7528\uff0c\u4f46\u5b83\u7684\u5b9e\u8d28\u4ecd\u7136\u662f\u63a8\u7406\u9636\u6bb5\u7684\u4e00\u79cd\u7b56\u7565\u3002<\/p>\n
\u6bcf\u6b21\u63a8\u7406\u4ecd\u7136\u9700\u8981\u751f\u6210\u591a\u4e2a\u601d\u7ef4\u94fe\u8def\u5206\u652f\uff0c\u7b97\u529b\u5f00\u9500\u95ee\u9898\u4ecd\u7136\u5b58\u5728\u3002<\/p>\n
\u7b97\u6cd5\u7ec8\u7a76\u662f\u8bba\u6587\u4e2d\u7684\u5185\u5bb9\uff0c\u53ea\u6709\u628a\u5b83\u5e94\u7528\u5230\u5927\u6a21\u578b\u771f\u5b9e\u7684\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0c\u624d\u4f1a\u4ea7\u751f\u5b9e\u9645\u4ef7\u503c\u3002<\/p>\n
\u76ee\u524d\uff0c\u5bf9\u4e8eDeepSeek\u7b49\u63a8\u7406\u6a21\u578b\uff0c\u4e1a\u754c\u6700\u524d\u6cbf\u7684\u8bad\u7ec3\u65b9\u6cd5\u662f\u4ee5GRPO\u4e3a\u4ee3\u8868\u7684\u57fa\u4e8e\u53ef\u9a8c\u8bc1\u5956\u52b1\u7684\u5f3a\u5316\u5b66\u4e60\u3002<\/p>\n
\u7528\u4e00\u4e2a\u8bba\u6587\u4e2d\u7684\u4f8b\u5b50\u6765\u8bf4\u660e\u6807\u51c6GRPO\u7684\u8bad\u7ec3\u8fc7\u7a0b\uff1a<\/p>\n
\u7ed9\u6a21\u578b\u4e00\u9053\u9898\uff0c\u8ba9\u6a21\u578b\u968f\u673a\u751f\u62108\u4e2a\u4e92\u4e0d\u76f8\u540c\u7684\u89e3\u7b54\uff0c\u5176\u4e2d\u6b63\u786e\u7684\u89e3\u7b54\u5c06\u4f1a\u5f97\u5230\u5956\u52b1\uff0c\u4ee5\u6b64\u66f4\u65b0\u6a21\u578b\u7684\u6743\u91cd\u3002\u884c\u4e1a\u5185\u5e38\u8bf4\u7684\u201c\u5956\u52b1\u6a21\u578b\u201d\u5c31\u662f\u7528\u6765\u505a\u8fd9\u4ef6\u4e8b\u7684\u3002<\/p>\n
\u4e0d\u8fc7\uff0c\u7531\u4e8e\u8fd98\u4e2a\u89e3\u7b54\u90fd\u662f\u901a\u8fc7\u6807\u51c6\u7684\u968f\u673a\u91c7\u6837\u800c\u4ea7\u751f\u7684\uff0c\u201c\u53cd\u590d\u601d\u8003\u201d\u7684\u73b0\u8c61\u81ea\u7136\u65e0\u6cd5\u907f\u514d\u3002\u6700\u540e\u7684\u7ed3\u679c\u5c31\u4f1a\u662f\uff1a\u6a21\u578b\u80fd\u591f\u505a\u5bf9\u9898\u76ee\uff0c\u4f46\u4e5f\u5b66\u4f1a\u4e86\u8bf4\u4e00\u5927\u5806\u5e9f\u8bdd\u3002<\/p>\n
\u4e3a\u4e86\u8ba9SAGE\u7b97\u6cd5\u6295\u5165\u4f7f\u7528\uff0c\u7814\u7a76\u56e2\u961f\u5f00\u59cb\u5bf9\u73b0\u6709\u7684GRPO\u7b97\u6cd5\u8fdb\u884c\u66ff\u6362\uff1a<\/p>\n
\u540c\u6837\u662f\u751f\u62108\u4e2a\u89e3\u7b54\uff0c\u5176\u4e2d\u76846\u4e2a\u4f7f\u7528\u539f\u6709\u7684\u968f\u673a\u91c7\u6837\uff0c\u5269\u4e0b\u76842\u4e2a\u4f7f\u7528SAGE\u7b97\u6cd5\u751f\u6210\u3002<\/p>\n
\u6211\u4eec\u5df2\u7ecf\u77e5\u9053\uff0cSAGE\u7b97\u6cd5\u7ed9\u51fa\u7684\u56de\u7b54\u65e2\u6b63\u786e\uff0c\u53c8\u7b80\u7ec3\uff0c\u5f3a\u5316\u5b66\u4e60\u7684\u5956\u52b1\u673a\u5236\u867d\u7136\u4e0d\u77e5\u9053\u54ea\u4e24\u4e2a\u89e3\u7b54\u662fSAGE\u7b97\u6cd5\u751f\u6210\u7684\uff0c\u5374\u603b\u662f\u4f1a\u7ed9\u5b83\u4eec\u6253\u9ad8\u5206\u3002<\/p>\n
\u5982\u6b64\u5faa\u73af\u4e0b\u53bb\uff0cSAGE\u7b97\u6cd5\u751f\u6210\u7684\u7b54\u6848\u9010\u6e10\u6210\u4e3a\u201c\u699c\u6837\u201d\uff0c\u6a21\u578b\u9010\u6b65\u66f4\u65b0\u6743\u91cd\u540e\u5c31\u4f1a\u4e00\u70b9\u70b9\u6539\u6b63\u81ea\u5df1\u8bf4\u5e9f\u8bdd\u7684\u6bdb\u75c5\uff0c\u601d\u7ef4\u94fe\u7684\u751f\u6210\u6a21\u5f0f\u4e5f\u9010\u6e10\u5411SAGE\u7b97\u6cd5\u7684\u9ad8\u6548\u6a21\u5f0f\u8f6c\u53d8\u3002<\/p>\n
\u8fd9\u6837\u4e00\u6765\uff0cSAGE\u7b97\u6cd5\u5c31\u6210\u529f\u70d9\u5370\u5728\u5927\u6a21\u578b\u7684DNA\u4e2d\uff0c\u6210\u4e3aSAGE-RL\u8fd0\u884c\u673a\u5236\u3002<\/p>\n
\u6700\u540e\uff0c\u7814\u7a76\u56e2\u961f\u9009\u62e9\u4e86MATH-500\u3001AIME 2025\u7b496\u4e2a\u6781\u9ad8\u96be\u5ea6\u7684AI\u6570\u5b66\u57fa\u51c6\u6d4b\u8bd5\u6765\u8bc4\u4f30SAGE-RL\u7684\u6548\u679c\uff1a<\/p>\n
\u73b0\u6709\u7684\u4e00\u4e9b\u7f29\u51cf\u63a8\u7406\u957f\u5ea6\u7684\u7b97\u6cd5\u4e3a\u4e86\u538b\u7f29token\u6d88\u8017\u91cf\u4f1a\u964d\u4f4e\u6a21\u578b\u56de\u7b54\u7684\u51c6\u786e\u7387\u3002<\/p>\n
\u800c\u57fa\u4e8eGRPO\u7684SAGE-RL\u5374\u80fd\u5b9e\u73b0\u201c\u964d\u672c\u589e\u6548\u201d\u3002<\/p>\n<\/p>\n
\u7814\u7a76\u56e2\u961f\u57287B\u7ea7\u522b\u6a21\u578b\u7684\u6d4b\u8bd5\u4e2d\u53d1\u73b0\uff0c\u52a0\u5165SAGE\u80fd\u5c06\u6a21\u578b\u7684\u5e73\u5747\u56de\u590d\u957f\u5ea6\u4ece3219\u4e2atoken\u538b\u7f29\u81f32141\u4e2a\uff0c\u5927\u7ea6\u51cf\u5c11\u4e861\/3\u7684\u8ba1\u7b97\u91cf\u3002<\/p>\n
\u540c\u65f6\uff0c\u6a21\u578b\u5728MATH-500\u4e0a\u7684Pass@1\u51c6\u786e\u7387\u4ece92%\u63d0\u5347\u5230\u4e8693%\uff0cToken\u6548\u7387\u66f4\u662f\u65ad\u5d16\u5f0f\u9886\u5148\u3002<\/p>\n
\u800c\u6b64\u524d\u4ecb\u7ecd\u7684\u65b0\u5b9a\u4e49\u6307\u6807\uff1a\u9996\u6b21\u6b63\u786e\u6b65\u9aa4\u6bd4\u7387\uff08RFCS\uff09\u4e5f\u663e\u793a\uff0c\u65e0\u6548\u5197\u4f59\u6b65\u9aa4\u7684\u6bd4\u4f8b\u5927\u5e45\u4e0b\u964d\uff0c\u6a21\u578b\u7ec8\u4e8e\u6539\u6389\u4e86\u201c\u8fc7\u5ea6\u601d\u8003\u201d\u7684\u4e60\u60ef\uff0c\u5b66\u4f1a\u4e86\u201c\u70b9\u5230\u4e3a\u6b62\u201d\u3002<\/p>\n
\u76ee\u524d\uff0c\u5e02\u573a\u4e0a\u7684AI\u4f01\u4e1a\u666e\u904d\u91c7\u7528API\u8ba1\u8d39\u6a21\u5f0f\u3002\u8868\u9762\u4e0a\uff0c\u589e\u52a0token\u7684\u8f93\u51fa\u770b\u4f3c\u80fd\u591f\u589e\u52a0\u6536\u5165\u3002<\/p>\n
\u5b9e\u9645\u4e0a\uff0c\u5934\u90e8AI\u5382\u5546\u505a\u68a6\u90fd\u60f3\u5b9e\u73b0\u964d\u4f4e\u5197\u4f59token\u7684\u6d88\u8017\u3002\u80fd\u8ba9\u670d\u52a1\u5668\u63a5\u6536\u66f4\u591a\u7684\u5e76\u53d1\u8bf7\u6c42\u83b7\u5f97\u7684\u5229\u6da6\uff0c\u8fdc\u6bd4\u8ba9\u5355\u4e2a\u8bf7\u6c42\u591a\u8bf4\u51e0\u53e5\u5e9f\u8bdd\u8981\u591a\u3002<\/p>\n
\u5728AI\u667a\u80fd\u4ee3\u7406\u6295\u5165\u5b9e\u9645\u5de5\u4f5c\u6d41\u7684\u4eca\u5929\uff0c\u5ef6\u8fdf\u5df2\u7ecf\u6210\u4e3a\u4ea7\u54c1\u7684\u201c\u65a9\u6740\u7ebf\u201d\uff0c\u54cd\u5e94\u901f\u5ea6\u6162\u7684\u6a21\u578b\u65e0\u6cd5\u5b9e\u65f6\u6ee1\u8db3\u7528\u6237\u7684\u9700\u6c42\uff0c\u5fc5\u7136\u4f1a\u88ab\u7528\u6237\u65e0\u60c5\u5730\u629b\u5f03\u3002<\/p>\n
DeepSeek\u5728\u4e00\u5e74\u524d\u9707\u60ca\u7845\u8c37<\/a>\uff0c\u4e0d\u4ec5\u662f\u56e0\u4e3a\u5f00\u6e90\u4e86\u5f3a\u5927\u7684\u6280\u672f\u548c\u5148\u8fdb\u7684\u7b97\u6cd5\uff0c\u66f4\u662f\u56e0\u4e3a\u4ed6\u4eec\u628a\u63a8\u7406\u6210\u672c\u6253\u5230\u4e86\u4e0d\u53ef\u601d\u8bae\u7684\u4f4e\u4ef7\u3002<\/p>\n\u73b0\u6709\u7684\u6a21\u578b\u82e5\u662f\u65e0\u6cd5\u6539\u53d8\u751f\u6210\u7b54\u6848\u65f6\u4f34\u968f\u5927\u91cf\u5197\u4f59token\u7684\u73b0\u72b6\uff0cAPI\u7684\u7efc\u5408\u6210\u672c\u5c31\u6c38\u8fdc\u964d\u4e0d\u4e0b\u6765\u3002<\/p>\n
\u800c\u638c\u63e1\u4e86\u7c7b\u4f3cSAGE\u8fd9\u79cd\u9ad8\u6548\u63a8\u7406\u6280\u672f\u7684\u5382\u5546\uff0c\u5c31\u53ef\u4ee5\u51ed\u501f\u6781\u5c11\u7684\u7b97\u529b\u8fbe\u5230\u76f8\u540c\u751a\u81f3\u66f4\u9ad8\u7684\u51c6\u786e\u7387\u3002\u5728\u9876\u5c16\u6a21\u578b\u63d0\u5347\u667a\u80fd\u7a0b\u5ea6\u7684\u8fb9\u9645\u6536\u76ca\u9510\u51cf\u7684\u60c5\u51b5\u4e0b\uff0c\u8fd9\u6b63\u662f\u964d\u4ef7\u6253\u4ef7\u683c\u6218\u7684\u5e95\u6c14\u3002<\/p>\n
\u4ee5\u540e\uff0c\u5927\u6a21\u578b\u4e5f\u65e0\u9700\u518d\u9760\u957f\u7bc7\u5927\u8bba\u8bc1\u660e\u81ea\u5df1\u3002\u6700\u9ad8\u7ea7\u7684\u667a\u80fd\uff0c\u5176\u5b9e\u5c31\u85cf\u5728\u90a3\u4e2a\u6070\u5230\u597d\u5904\u7684\u201c\u201d\u4e2d\u3002<\/p>\n
(\u6765\u6e90\uff1a\u65b0\u6d6a\u79d1\u6280)<\/p>\n","protected":false},"excerpt":{"rendered":"
\u8fc7\u53bb\u8fd9\u4e24\u5e74\uff0c\u5927\u8bed\u8a00\u6a21\u578b\u8d70\u5f97\u5b9e\u5728\u592a\u5feb\uff0c\u5c24\u5176\u5728\u90a3\u4e9b\u9700\u8981\u590d\u6742\u903b\u8f91\u63a8\u7406\u7684\u4efb\u52a1\u4e0a\uff0c\u80fd\u529b\u5df2\u7ecf\u9ad8\u5230\u4e86\u4e00\u4e2a\u6211\u4eec\u4ee5\u524d\u4e0d\u592a\u6562\u60f3\u7684\u6c34 […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":[],"categories":[6],"tags":[379,157,88,367,238],"views":26,"_links":{"self":[{"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=\/wp\/v2\/posts\/732236"}],"collection":[{"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=732236"}],"version-history":[{"count":0,"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=\/wp\/v2\/posts\/732236\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=732236"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=732236"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.alphaseek.cn\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=732236"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
![\"\u5b57\u8282\u8df3\u52a8\u65b0\u7b97\u6cd5\uff0c\u76f4\u63a5\u628a1\/3\u7684\u7b97\u529b\u780d\u6389\u4e86\"](\"https:\/\/x0.ifengimg.com\/ucms\/2026_09\/D8E3BA095AD9AD33360BE21A73817A05E6AD1367_size243_w692_h608.png\")
![\"\u5b57\u8282\u8df3\u52a8\u65b0\u7b97\u6cd5\uff0c\u76f4\u63a5\u628a1\/3\u7684\u7b97\u529b\u780d\u6389\u4e86\"](\"https:\/\/x0.ifengimg.com\/ucms\/2026_09\/7CBB6C79D515EF620C9D33CDDF784A5304DCB38D_size252_w692_h581.png\")
![\"\u5b57\u8282\u8df3\u52a8\u65b0\u7b97\u6cd5\uff0c\u76f4\u63a5\u628a1\/3\u7684\u7b97\u529b\u780d\u6389\u4e86\"](\"https:\/\/x0.ifengimg.com\/ucms\/2026_09\/AEDA951D5C87F3A413595ACD60EB634522849F58_size189_w1080_h330.png\")
![\"\u5b57\u8282\u8df3\u52a8\u65b0\u7b97\u6cd5\uff0c\u76f4\u63a5\u628a1\/3\u7684\u7b97\u529b\u780d\u6389\u4e86\"](\"https:\/\/x0.ifengimg.com\/ucms\/2026_09\/6E35272C653B18E830413E1210ECEE67045E1C1D_size308_w692_h515.png\")
<\/p>\n![\"\u5b57\u8282\u8df3\u52a8\u65b0\u7b97\u6cd5\uff0c\u76f4\u63a5\u628a1\/3\u7684\u7b97\u529b\u780d\u6389\u4e86\"](\"https:\/\/x0.ifengimg.com\/ucms\/2026_09\/E223E6385EBC0E9FBFE809CFEDFEA2C6F10CD922_size56_w1080_h414.png\")
<\/p>\n![\"\u5b57\u8282\u8df3\u52a8\u65b0\u7b97\u6cd5\uff0c\u76f4\u63a5\u628a1\/3\u7684\u7b97\u529b\u780d\u6389\u4e86\"](\"https:\/\/x0.ifengimg.com\/ucms\/2026_09\/7170E22DED995D6901B8215EEE0F522395378A7C_size382_w1080_h1287.png\")
<\/p>\n![\"\u5b57\u8282\u8df3\u52a8\u65b0\u7b97\u6cd5\uff0c\u76f4\u63a5\u628a1\/3\u7684\u7b97\u529b\u780d\u6389\u4e86\"](\"https:\/\/x0.ifengimg.com\/ucms\/2026_09\/80C7A7DF9173DA61BA7486163C3DB625AF64CD79_size154_w1080_h598.png\")
<\/p>\n![\"\u5b57\u8282\u8df3\u52a8\u65b0\u7b97\u6cd5\uff0c\u76f4\u63a5\u628a1\/3\u7684\u7b97\u529b\u780d\u6389\u4e86\"](\"https:\/\/x0.ifengimg.com\/ucms\/2026_09\/FD6857D8D477EC644B0E436A2A79933BEC62E5C4_size120_w1080_h317.png\")
<\/p>\n