{"id":36180,"date":"2025-01-07T03:02:13","date_gmt":"2025-01-06T19:02:13","guid":{"rendered":"https:\/\/17aitech.com\/?p=36180"},"modified":"2025-01-07T03:02:13","modified_gmt":"2025-01-06T19:02:13","slug":"%e5%8d%b7%e8%b5%b7%e6%9d%a5%ef%bc%81%e8%ae%a9%e6%99%ba%e8%83%bd%e4%bd%93%e8%af%84%e4%bc%b0%e6%99%ba%e8%83%bd%e4%bd%93%ef%bc%8cmeta%e5%8f%91%e5%b8%83agent-as-a-judge","status":"publish","type":"post","link":"https:\/\/17aitech.com\/?p=36180","title":{"rendered":"\u5377\u8d77\u6765\uff01\u8ba9\u667a\u80fd\u4f53\u8bc4\u4f30\u667a\u80fd\u4f53\uff0cMeta\u53d1\u5e03Agent-as-a-Judge"},"content":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-10-18-4\" target=\"_blank\">\u5377\u8d77\u6765\uff01\u8ba9\u667a\u80fd\u4f53\u8bc4\u4f30\u667a\u80fd\u4f53\uff0cMeta\u53d1\u5e03Agent-as-a-Judge<\/a><\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-9f302a32718088e5b3475af407ceaac1.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-9f302a32718088e5b3475af407ceaac1.png\"><\/a><\/section>\n<blockquote data-author-name=\"\" data-content-utf8-length=\"166\" data-source-title=\"\" data-type=\"2\" data-url=\"\">\n<section>\n<section>AIxiv\u4e13\u680f\u662f\u673a\u5668\u4e4b\u5fc3\u53d1\u5e03\u5b66\u672f\u3001\u6280\u672f\u5185\u5bb9\u7684\u680f\u76ee\u3002\u8fc7\u53bb\u6570\u5e74\uff0c\u673a\u5668\u4e4b\u5fc3AIxiv\u4e13\u680f\u63a5\u6536\u62a5\u9053\u4e862000\u591a\u7bc7\u5185\u5bb9\uff0c\u8986\u76d6\u5168\u7403\u5404\u5927\u9ad8\u6821\u4e0e\u4f01\u4e1a\u7684\u9876\u7ea7\u5b9e\u9a8c\u5ba4\uff0c\u6709\u6548\u4fc3\u8fdb\u4e86\u5b66\u672f\u4ea4\u6d41\u4e0e\u4f20\u64ad\u3002\u5982\u679c\u60a8\u6709\u4f18\u79c0\u7684\u5de5\u4f5c\u60f3\u8981\u5206\u4eab\uff0c\u6b22\u8fce\u6295\u7a3f\u6216\u8005\u8054\u7cfb\u62a5\u9053\u3002\u6295\u7a3f\u90ae\u7bb1\uff1aliyazhou@jiqizhixin.com\uff1bzhaoyunfeng@jiqizhixin.com<\/section>\n<\/section>\n<\/blockquote>\n<section><\/section>\n<section><strong>\u672c\u8bba\u6587\u4f5c\u8005\u4e3b\u8981\u5305\u62ec\u5fb7\u56fd\u8ba1\u7b97\u673a\u79d1\u5b66\u5bb6\u3001LSTM \u4e4b\u7236 J\u00fcrgen Schmidhuber\uff1bMeta AI \u7814\u7a76\u56e2\u961f\uff08FAIR\uff09\u7814\u7a76\u79d1\u5b66\u5bb6\u603b\u76d1\u7530\u6e0a\u680b\uff0c\u4ed6\u8d1f\u8d23\u9886\u5bfc\u5927\u8bed\u8a00\u6a21\u578b\uff08LLMs\uff09\u5728\u63a8\u7406\u3001\u89c4\u5212\u548c\u51b3\u7b56\u65b9\u9762\u7684\u7814\u7a76\u56e2\u961f\uff0c\u4e3b\u5bfc\u4e86 OpenGo\u3001StreamingLLM \u548c GaLore \u9879\u76ee\uff0c\u4e13\u6ce8\u4e8e\u63d0\u5347\u5927\u6a21\u578b\u7684\u8bad\u7ec3\u548c\u63a8\u7406\u6548\u7387\uff1bVikas Chandra\uff0cMeta Reality Lab AI \u8d1f\u8d23\u4eba\uff1b\u8bf8\u845b\u9e23\u6668\uff0cMeta \u7814\u7a76\u79d1\u5b66\u5bb6\u5b9e\u4e60\u751f\uff0c\u540c\u65f6\u5728\u6c99\u7279\u963f\u535c\u675c\u62c9\u56fd\u738b\u79d1\u6280\u5927\u5b66\uff08KAUST\uff09\u653b\u8bfb\u535a\u58eb\u4e09\u5e74\u7ea7\uff0c\u5e08\u4eceJ\u00fcrgen Schmidhuber\uff0cGPTSwarm \u7b2c\u4e00\u4f5c\u8005\uff0cMetaGPT \u5171\u540c\u7b2c\u4e00\u4f5c\u8005\uff1bZechun Li\uff0cMeta Reality Lab \u7814\u7a76\u79d1\u5b66\u5bb6\uff0cMobileLLM \u7684\u7b2c\u4e00\u4f5c\u8005\uff1bYunyang Xiong\uff0cMeta Reality Lab \u9ad8\u7ea7\u7814\u7a76\u79d1\u5b66\u5bb6\uff0cEfficientSAM \u7b2c\u4e00\u4f5c\u8005\u3002<\/strong><\/section>\n<section><\/section>\n<section>\u5982\u679c\u8bf4\u53bb\u5e74\u5927\u5382\u7684\u7ade\u4e89\u7126\u70b9\u662f LLM\uff0c\u90a3\u4e48\u4eca\u5e74\uff0c\u5404\u5927\u79d1\u6280\u516c\u53f8\u7eb7\u7eb7\u63a8\u51fa\u4e86\u5404\u81ea\u7684\u667a\u80fd\u4f53\u5e94\u7528\u3002<\/section>\n<section><\/section>\n<section>\u5fae\u8f6f\u53d1\u5e03\u4e86 Copilot\uff0cApple \u5c06 Apple Intelligence \u63a5\u5165\u4e86 OpenAI \u4ee5\u589e\u5f3a Siri\u3002\u591a\u667a\u80fd\u4f53\u4e5f\u662f OpenAI \u672a\u6765\u91cd\u8981\u7684\u7814\u7a76\u65b9\u5411\u4e4b\u4e00\uff0c\u8fd9\u5bb6\u516c\u53f8\u7684\u6700\u65b0\u6210\u679c \u2014\u2014Swarm\uff0c\u4e00\u4e2a\u5b9e\u9a8c\u6027\u8d28\u7684\u591a\u667a\u80fd\u4f53\u7f16\u6392\u6846\u67b6\u5728\u5f00\u6e90\u540e\u5f15\u8d77\u4e86\u70ed\u70c8\u8ba8\u8bba\uff0c\u6709\u7f51\u53cb\u8868\u793a\u8fd9\u80fd\u5e2e\u52a9\u7b80\u5316\u8bb8\u591a\u6f5c\u5728\u591a\u667a\u80fd\u4f53\u7528\u4f8b\u7684\u5de5\u4f5c\u6d41\u7a0b\u3002<\/section>\n<section><\/section>\n<section>\u624e\u514b\u4f2f\u683c\u66f4\u662f\u65ad\u8a00\uff1a\u300cAI \u667a\u80fd\u4f53\u7684\u6570\u91cf\u53ef\u80fd\u4f1a\u8fbe\u5230\u6570\u5341\u4ebf\uff0c\u6700\u7ec8\u751a\u81f3\u8d85\u8fc7\u4eba\u7c7b\u3002\u300d\u5728 Meta Connect 2024\u5927\u4f1a\u4e0a\uff0cMeta \u63a8\u51fa\u4e86\u63a5\u5165 Llama 3.2 \u7684\u667a\u80fd\u773c\u955c Orion \u548c\u5347\u7ea7\u7248 Quest 3S\uff0c\u663e\u793a\u51fa\u667a\u80fd\u4f53\u6b63\u5728\u8fc5\u901f\u6e17\u900f\u8fdb Meta \u7684\u5404\u4e2a\u5e94\u7528\u9886\u57df\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-5e82789cd2b564885d339bb5360a3067.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-5e82789cd2b564885d339bb5360a3067.png\"><\/a><\/section>\n<section>\u8fd1\u65e5\uff0cMeta \u63d0\u51fa\u4e86 Agent-as-a-Judge \u7684\u6982\u5ff5\uff0c\u88ab\u89c6\u4e3a\u667a\u80fd\u4f53\u4f18\u5316\u65b9\u9762\u7684\u53c8\u4e00\u91cd\u8981\u6210\u679c\u3002\u4f20\u7edf\u7684\u667a\u80fd\u4f53\u8bc4\u4f30\u65b9\u5f0f\u5f80\u5f80\u53ea\u5173\u6ce8\u6700\u7ec8\u7ed3\u679c\uff0c\u5ffd\u7565\u4e86\u6267\u884c\u8fc7\u7a0b\u4e2d\u7684\u5173\u952e\u7ec6\u8282\uff0c\u6216\u4f9d\u8d56\u5927\u91cf\u4eba\u529b\u8fdb\u884c\u8bc4\u4f30\u3002\u4e3a\u4e86\u89e3\u51b3\u8fd9\u4e00\u75db\u70b9\uff0cMeta \u63a8\u51fa\u4e86\u7528\u667a\u80fd\u4f53\u8bc4\u4f30\u667a\u80fd\u4f53\u7684\u65b0\u65b9\u6cd5\uff0c\u4f7f\u8bc4\u4f30\u8fc7\u7a0b\u66f4\u52a0\u7075\u6d3b\u4e14\u81ea\u52a8\u5316\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-28ffc52ef91aba48369722f85bfd9d67.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-28ffc52ef91aba48369722f85bfd9d67.png\"><\/a><\/section>\n<section>\u8be5\u6846\u67b6\u5728 LLM-as-a-Judge \u7684\u57fa\u7840\u4e0a\u8fdb\u884c\u4e86\u5347\u7ea7\uff0c\u589e\u52a0\u4e86\u4e2d\u95f4\u53cd\u9988\u529f\u80fd\uff0c\u786e\u4fdd\u4efb\u52a1\u7684\u6bcf\u4e2a\u73af\u8282\u90fd\u80fd\u5f97\u5230\u7cbe\u51c6\u8bc4\u4f30\u4e0e\u4f18\u5316\uff0c\u540c\u65f6\u8fd8\u80fd\u6709\u6548\u6a21\u62df\u5e76\u63a5\u8fd1\u4eba\u7c7b\u53cd\u9988\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-e261d133e853d631964620c44e5513cd.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-e261d133e853d631964620c44e5513cd.png\"><\/a><\/section>\n<ul>\n<li>\n<section>\u8bba\u6587\u6807\u9898\uff1aAgent-as-a-Judge: Evaluate Agents with Agents<\/section>\n<\/li>\n<li>\n<section>\u8bba\u6587\u5730\u5740\uff1ahttps:\/\/arxiv.org\/pdf\/2410.10934<\/section>\n<\/li>\n<li>\n<section>\u9879\u76ee\u5730\u5740\uff1ahttps:\/\/github.com\/metauto-ai\/agent-as-a-judge<\/section>\n<\/li>\n<\/ul>\n<section><\/section>\n<section>\u4e3a\u4e86\u514b\u670d\u73b0\u6709\u57fa\u51c6\u5b58\u5728\u7684\u95ee\u9898\uff0c\u5e76\u4e3a Agent-as-a-Judge \u63d0\u4f9b\u4e00\u4e2a\u6982\u5ff5\u9a8c\u8bc1\u6d4b\u8bd5\u5e73\u53f0\uff0c\u7814\u7a76\u8005\u8fd8\u63d0\u51fa\u4e86 DevAI\uff0c\u4e00\u4e2a\u5305\u542b 55 \u9879\u73b0\u5b9e\u81ea\u52a8\u4eba\u5de5\u667a\u80fd\u5f00\u53d1\u4efb\u52a1\u7684\u65b0\u57fa\u51c6\u3002\u5b83\u5305\u62ec\u4e30\u5bcc\u7684\u624b\u52a8\u6ce8\u91ca\uff0c\u5982\u603b\u5171 365 \u4e2a\u5206\u5c42\u7528\u6237\u9700\u6c42\u3002<\/section>\n<section><\/section>\n<section><strong>Agent-as-a-Judge\uff1a\u667a\u80fd\u4f53\u8bc4\u4f30\u667a\u80fd\u4f53<\/strong><\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-aa65ec8487a80fdf62a4075b47472959.gif\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-aa65ec8487a80fdf62a4075b47472959.gif\"><\/a><\/section>\n<section>Agent-as-a-Judge \u6846\u67b6\u6700\u5927\u7684\u4eae\u70b9\u5728\u4e8e\u5176\u4e0e\u4eba\u7c7b\u8bc4\u4f30\u8005\u7684\u9ad8\u5ea6\u4e00\u81f4\u6027\u3002\u5728\u5b9e\u9a8c\u4e2d\uff0c\u8be5\u6846\u67b6\u7684\u8bc4\u4f30\u7ed3\u679c\u4e0e\u4eba\u7c7b\u4e13\u5bb6\u7684\u5bf9\u9f50\u7387\u9ad8\u8fbe 90.44%\uff0c\u8fdc\u8d85 LLM-as-a-Judge \u7684 70.76%\u3002\u8fd9\u4e00\u7ed3\u679c\u8868\u660e\uff0c\u667a\u80fd\u4f53\u5728\u5904\u7406\u590d\u6742\u4efb\u52a1\u65f6\uff0c\u80fd\u591f\u50cf\u4eba\u7c7b\u4e00\u6837\u7cbe\u786e\u5730\u5224\u65ad\u5e76\u4fee\u590d\u95ee\u9898\uff0c\u6781\u5927\u5730\u51cf\u5c11\u4e86\u5bf9\u4eba\u5de5\u8bc4\u4f30\u7684\u4f9d\u8d56\uff0c\u540c\u65f6\u663e\u8457\u63d0\u9ad8\u4e86\u6548\u7387\u3002<\/section>\n<section><\/section>\n<section><strong>\u663e\u8457\u7684\u6548\u7387\u63d0\u5347\u3002<\/strong>\u5b9e\u9a8c\u8868\u660e\uff0cAgent-as-a-Judge \u5728\u6548\u7387\u4e0a\u4e5f\u5177\u6709\u660e\u663e\u4f18\u52bf\u3002\u4e0e\u4eba\u7c7b\u8bc4\u4f30\u8005\u8017\u8d39 86.5 \u5c0f\u65f6\u76f8\u6bd4\uff0cAgent-as-a-Judge \u4ec5\u9700 118.43 \u5206\u949f\u5c31\u80fd\u5b8c\u6210\u76f8\u540c\u4efb\u52a1\uff0c\u5927\u5e45\u8282\u7701\u4e86\u65f6\u95f4\u548c\u6210\u672c\u3002\u8bc4\u4f30 55 \u4e2a\u4efb\u52a1\u7684\u603b\u6210\u672c\u4ec5\u4e3a 30.58 \u7f8e\u5143\uff0c\u5e73\u5747\u6bcf\u4e2a\u4efb\u52a1\u7684\u8bc4\u4f30\u8d39\u7528\u4ec5\u4e3a 0.55 \u7f8e\u5143\uff0c\u663e\u793a\u4e86\u6781\u9ad8\u7684\u6027\u4ef7\u6bd4\u548c\u5de5\u4f5c\u6548\u7387\u3002<\/section>\n<section><\/section>\n<section><strong>\u586b\u8865\u8bc4\u4f30\u4e2d\u7684\u53cd\u9988\u7a7a\u767d\u3002<\/strong>\u5f53\u524d\uff0c\u667a\u80fd\u4f53\u8bc4\u4f30\u65b9\u6cd5\u666e\u904d\u7f3a\u4e4f\u4e2d\u95f4\u53cd\u9988\u673a\u5236\uff0c\u53ea\u5173\u6ce8\u6700\u7ec8\u7ed3\u679c\uff0c\u5ffd\u89c6\u4e86\u4efb\u52a1\u6267\u884c\u4e2d\u7684\u5173\u952e\u6b65\u9aa4\u3002\u667a\u80fd\u4f53\u5728\u89e3\u51b3\u590d\u6742\u95ee\u9898\u65f6\uff0c\u901a\u5e38\u50cf\u4eba\u7c7b\u4e00\u6837\uff0c\u9010\u6b65\u601d\u8003\u5e76\u89e3\u51b3\u95ee\u9898\u3002\u56e0\u6b64\uff0c\u8bc4\u4f30\u4e0d\u4ec5\u5e94\u8be5\u5173\u6ce8\u7ed3\u679c\uff0c\u8fd8\u9700\u8003\u5bdf\u6bcf\u4e2a\u6b65\u9aa4\u7684\u601d\u7ef4\u8fc7\u7a0b\u548c\u884c\u4e3a\u8f68\u8ff9\u3002Agent-as-a-Judge \u901a\u8fc7\u63d0\u4f9b\u4e2d\u95f4\u53cd\u9988\uff0c\u586b\u8865\u4e86\u8fd9\u4e00\u7a7a\u767d\uff0c\u6807\u5fd7\u7740\u667a\u80fd\u4f53\u8bc4\u4f30\u8fdb\u5165\u4e86\u4e00\u4e2a\u65b0\u7684\u9636\u6bb5\u3002<\/section>\n<section><\/section>\n<section><strong>\u6570\u636e\u96c6\u6311\u6218\u4e0e\u7cfb\u7edf\u8868\u73b0\u3002<\/strong>\u5b9e\u9a8c\u8fd8\u63ed\u793a\uff0c\u5373\u4f7f\u662f\u8868\u73b0\u8f83\u597d\u7684\u667a\u80fd\u4f53\u7cfb\u7edf\uff08\u5982 GPT-Pilot \u548c OpenHands\uff09\u4e5f\u4ec5\u80fd\u6ee1\u8db3 DevAI \u6570\u636e\u96c6\u4e2d\u7ea6 29% \u7684\u4efb\u52a1\u9700\u6c42\uff0c\u4efb\u52a1\u5b8c\u6210\u7387\u6709\u9650\uff0c\u51f8\u663e\u4e86\u8be5\u6570\u636e\u96c6\u7684\u6311\u6218\u6027\u3002\u5728\u4e0e\u4eba\u7c7b\u4e13\u5bb6\u8bc4\u4f30\u7684\u5bf9\u6bd4\u4e2d\uff0cAgent-as-a-Judge \u8868\u73b0\u51fa\u8272\uff0c\u8fbe\u5230\u4e86 90% \u7684\u5bf9\u9f50\u7387\uff0c\u800c LLM-as-a-Judge \u4ec5\u4e3a 70%\u3002\u66f4\u503c\u5f97\u6ce8\u610f\u7684\u662f\uff0cAgent-as-a-Judge \u7684\u8868\u73b0\u751a\u81f3\u4f18\u4e8e\u5355\u4e2a\u4e13\u5bb6\u8bc4\u4f30\u8005\uff0c\u610f\u5473\u7740\u5728\u67d0\u4e9b\u60c5\u51b5\u4e0b\uff0c\u8be5\u6846\u67b6\u4e0d\u4ec5\u80fd\u591f\u66ff\u4ee3\u4eba\u7c7b\u8bc4\u4f30\uff0c\u8fd8\u53ef\u80fd\u66f4\u52a0\u6709\u6548\u3002<\/section>\n<section><\/section>\n<section><strong>\u9ad8\u6027\u4ef7\u6bd4\u4e0e\u6f5c\u529b\u3002<\/strong>\u901a\u8fc7\u8282\u7701 97.72% \u7684\u65f6\u95f4\u548c 97.64% \u7684\u6210\u672c\uff0cAgent-as-a-Judge \u5c55\u793a\u4e86\u5176\u5728 AI \u8bc4\u4f30\u4e2d\u7684\u5de8\u5927\u6f5c\u529b\u3002\u5b83\u4e3a\u667a\u80fd\u4f53\u6280\u672f\u7684\u53d1\u5c55\u63d0\u4f9b\u4e86\u5f3a\u6709\u529b\u7684\u652f\u6301\uff0c\u6807\u5fd7\u7740 AI \u8bc4\u4f30\u5de5\u5177\u8fc8\u5411\u4e86\u66f4\u9ad8\u6548\u548c\u4f4e\u6210\u672c\u7684\u65b0\u7eaa\u5143\u3002<\/section>\n<section><\/section>\n<section><strong>\u884c\u4e1a\u8d8b\u52bf\u4e0e Cognition AI\u3002<\/strong>\u503c\u5f97\u6ce8\u610f\u7684\u662f\uff0c\u8fd1\u671f\u83b7\u5f97\u878d\u8d44\u7684 Cognition AI \u4e5f\u91c7\u53d6\u4e86\u7c7b\u4f3c\u601d\u8def\uff0c\u5373\u4f7f\u7528\u667a\u80fd\u4f53\u6765\u8bc4\u4f30\u667a\u80fd\u4f53\uff0c\u8fd9\u663e\u793a\u51fa\u8fd9\u4e00\u6982\u5ff5\u6b63\u5728\u6210\u4e3a\u4e1a\u754c\u7684\u4e00\u4e2a\u91cd\u8981\u8d8b\u52bf\uff08\u66f4\u591a\u4fe1\u606f\u8bf7\u53c2\u89c1\uff1ahttps:\/\/www.cognition.ai\/blog\/evaluating-coding-agents\uff09\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-04d27872b80d45630accb7495d1ea3ac.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-04d27872b80d45630accb7495d1ea3ac.png\"><\/a><\/section>\n<section>\u7efc\u4e0a\uff0cAgent-as-a-Judge \u7684\u63d0\u51fa\u6709\u5982\u4e0b\u4ef7\u503c\uff1a<\/section>\n<section><\/section>\n<section><strong>\uff081\uff09\u667a\u80fd\u4f53\u81ea\u6211\u6539\u8fdb\u7684\u4e2d\u95f4\u53cd\u9988\u673a\u5236<\/strong><\/section>\n<section><\/section>\n<section>Agent-as-a-Judge \u7684\u4e00\u4e2a\u6838\u5fc3\u4f18\u52bf\u5728\u4e8e\u5176\u63d0\u4f9b\u7684\u4e2d\u95f4\u53cd\u9988\uff0c\u8fd9\u5bf9\u4e8e\u5b9e\u73b0\u667a\u80fd\u4f53\u7684\u9ad8\u6548\u4f18\u5316\u81f3\u5173\u91cd\u8981\u3002\u5c3d\u7ba1\u5728\u672c\u7814\u7a76\u4e2d\u8fd9\u4e00\u6f5c\u529b\u5c1a\u672a\u88ab\u5145\u5206\u53d1\u6398\uff0c\u4f46\u5b83\u7684\u4f5c\u7528\u5df2\u7ecf\u521d\u89c1\u7aef\u502a\u3002\u901a\u8fc7\u5b66\u4e60\u8f85\u52a9\u5956\u52b1\u51fd\u6570\uff0c\u80fd\u591f\u89e3\u51b3\u5f3a\u5316\u5b66\u4e60\u4e2d\u7684\u7a00\u758f\u5956\u52b1\u95ee\u9898\uff0c\u63d0\u4f9b\u5173\u952e\u7684\u4e2d\u95f4\u53cd\u9988\u3002Agent-as-a-Judge \u6846\u67b6\u7684\u4eae\u70b9\u5728\u4e8e\uff0c\u5b83\u4f7f\u667a\u80fd\u4f53\u5728\u5904\u7406\u590d\u6742\u3001\u591a\u9636\u6bb5\u95ee\u9898\u65f6\uff0c\u80fd\u591f\u5b9e\u65f6\u53d1\u73b0\u5e76\u4fee\u590d\u89e3\u51b3\u65b9\u6848\u4e2d\u7684\u95ee\u9898\uff0c\u800c\u4f20\u7edf\u7684\u5ef6\u8fdf\u53cd\u9988\u673a\u5236\u96be\u4ee5\u505a\u5230\u8fd9\u4e00\u70b9\u3002\u5f15\u5165 Agent-as-a-Judge \u540e\uff0c\u4e3a\u6784\u5efa\u667a\u80fd\u4f53\u7248\u672c\u7684\u8fc7\u7a0b\u76d1\u7763\u5956\u52b1\u6a21\u578b\uff08PRM\uff09\u6253\u5f00\u4e86\u5927\u95e8\uff0c\u4ece\u800c\u8fdb\u4e00\u6b65\u63d0\u5347\u667a\u80fd\u4f53\u7684\u4f18\u5316\u6548\u7387\u3002<\/section>\n<section><\/section>\n<section><strong>\uff082\uff09\u7531 Agent-as-a-Judge \u9a71\u52a8\u7684\u98de\u8f6e\u6548\u5e94<\/strong><\/section>\n<section><\/section>\n<section>Agent-as-a-Judge \u548c\u88ab\u8bc4\u4f30\u667a\u80fd\u4f53\u4e4b\u95f4\u7684\u76f8\u4e92\u6539\u8fdb\uff0c\u901a\u8fc7\u4e0d\u65ad\u7684\u8fed\u4ee3\u53cd\u9988\u9010\u6b65\u6f14\u8fdb\uff0c\u8fd9\u4e00\u5faa\u73af\u5c55\u793a\u4e86\u5e7f\u9614\u7684\u53d1\u5c55\u524d\u666f\u3002\u901a\u8fc7\u5c06 Agent-as-a-Judge \u4f5c\u4e3a\u6838\u5fc3\u673a\u5236\uff0c\u6216\u8bb8\u80fd\u591f\u50ac\u751f\u51fa\u4e00\u79cd\u667a\u80fd\u4f53\u81ea\u6211\u535a\u5f08\u7cfb\u7edf\u3002\u968f\u7740 Agent-as-a-Judge \u4e0e\u88ab\u8bc4\u4f30\u667a\u80fd\u4f53\u7684\u6301\u7eed\u4ea4\u4e92\uff0c\u8fd9\u79cd\u8fc7\u7a0b\u53ef\u80fd\u4f1a\u4ea7\u751f\u98de\u8f6e\u6548\u5e94 \u2014\u2014 \u6bcf\u6b21\u6539\u8fdb\u76f8\u4e92\u5f3a\u5316\uff0c\u4ece\u800c\u4e0d\u65ad\u63a8\u52a8\u6027\u80fd\u7684\u63d0\u5347\u3002\u8fd9\u79cd\u8fed\u4ee3\u4e0d\u4ec5\u80fd\u589e\u5f3a\u667a\u80fd\u4f53\u7cfb\u7edf\u7684\u80fd\u529b\uff0c\u8fd8\u53ef\u80fd\u6210\u4e3a LLM \u63a8\u7406\u6570\u636e\u7684\u91cd\u8981\u8865\u5145\uff0c\u6709\u52a9\u4e8e\u5c06\u667a\u80fd\u4f53\u7684\u80fd\u529b\u66f4\u597d\u5730\u5d4c\u5165\u57fa\u7840\u6a21\u578b\u4e2d\uff0c\u8fdb\u4e00\u6b65\u62d3\u5c55\u667a\u80fd\u4f53\u7cfb\u7edf\u7684\u6f5c\u529b\u3002<\/section>\n<section><\/section>\n<section><strong>DevAI\uff1a\u4ece\u7528\u6237\u89d2\u5ea6\u51fa\u53d1\u7684 AI \u81ea\u52a8\u5316\u6570\u636e\u96c6<\/strong><\/section>\n<section><\/section>\n<section>\u8fc7\u53bb\u4e00\u5e74\u4e2d\uff0cLLM \u667a\u80fd\u4f53\u7cfb\u7edf\u7684\u80fd\u529b\u663e\u8457\u63d0\u5347\uff0c\u4ece\u89e3\u51b3\u7b80\u5355\u7684\u300c\u73a9\u5177\u95ee\u9898\u300d\u9010\u6b65\u6269\u5c55\u5230\u5904\u7406\u590d\u6742\u7684\u5b9e\u9645\u4efb\u52a1\u3002\u7136\u800c\uff0c\u5927\u591a\u6570\u73b0\u6709\u7684\u8bc4\u4f30\u65b9\u6cd5\u548c\u6570\u636e\u96c6\u4ecd\u7136\u57fa\u4e8e\u4e3a\u57fa\u7840\u6a21\u578b\u8bbe\u8ba1\u7684\u6807\u51c6\uff0c\u96be\u4ee5\u5168\u9762\u53cd\u6620\u667a\u80fd\u4f53\u5728\u73b0\u5b9e\u4efb\u52a1\u4e2d\u7684\u8868\u73b0\u548c\u6311\u6218\u3002\u4ee5 HumanEval \u548c MBPP \u7b49\u6570\u636e\u96c6\u4e3a\u4f8b\uff0c\u5c3d\u7ba1\u5b83\u4eec\u5728\u8bc4\u4f30\u57fa\u7840\u6a21\u578b\u7684\u7b97\u6cd5\u80fd\u529b\u65b9\u9762\u6709\u6548\uff0c\u4f46\u5728\u4ee3\u7801\u751f\u6210\u7b49\u9886\u57df\uff0c\u73b0\u6709\u65b9\u6cd5\u8fc7\u4e8e\u4f9d\u8d56\u6700\u7ec8\u7ed3\u679c\uff0c\u65e0\u6cd5\u6355\u6349\u5f00\u53d1\u8005\u5728\u73b0\u5b9e\u4efb\u52a1\u4e2d\u9047\u5230\u7684\u590d\u6742\u6027\u548c\u52a8\u6001\u8fc7\u7a0b\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-c2b959f7c21ea1300e5423edbb4c9d76.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-c2b959f7c21ea1300e5423edbb4c9d76.png\"><\/a><\/section>\n<section>\u5c3d\u7ba1 SWE-Bench \u5c1d\u8bd5\u5f15\u5165\u66f4\u63a5\u8fd1\u73b0\u5b9e\u7684\u8bc4\u4f30\u6807\u51c6\uff0c\u4f46\u5b83\u4f9d\u7136\u4e3b\u8981\u4f9d\u8d56\u300c\u89e3\u51b3\u7387\u300d\uff08resolve rate\uff09\uff0c\u8fd9\u4e00\u6307\u6807\u672a\u80fd\u63d0\u4f9b\u5f00\u53d1\u8fc7\u7a0b\u4e2d\u6bcf\u4e2a\u9636\u6bb5\u7684\u5177\u4f53\u53cd\u9988\uff0c\u4e5f\u96be\u4ee5\u6355\u6349\u667a\u80fd\u4f53\u7cfb\u7edf\u7684\u52a8\u6001\u8868\u73b0\u3002\u56e0\u6b64\uff0c\u8fd9\u7c7b\u8bc4\u4f30\u6807\u51c6\u65e0\u6cd5\u51c6\u786e\u53cd\u6620\u667a\u80fd\u4f53\u5728\u5b9e\u9645\u4efb\u52a1\u4e2d\u7684\u771f\u5b9e\u80fd\u529b\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-b3835c8ebd23cf37bbf3e21d6fd8f09f.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-b3835c8ebd23cf37bbf3e21d6fd8f09f.png\"><\/a><\/section>\n<section>\u76f8\u5173\u7814\u7a76\u751a\u81f3\u8868\u660e\uff0c\u5373\u4f7f\u4e0d\u5177\u5907\u667a\u80fd\u4f53\u7279\u6027\uff0c\u6a21\u578b\u4e5f\u80fd\u8fbe\u5230 27% \u7684\u89e3\u51b3\u7387\u3002\u6b64\u5916\uff0c\u6fc0\u70c8\u7684\u7ade\u4e89\u8fd8\u5f15\u53d1\u4e86\u5bf9 SWE-Bench \u5f97\u5206\u771f\u5b9e\u6027\u7684\u62c5\u5fe7\uff0c\u8bb8\u591a\u9ad8\u5206\u53ef\u80fd\u901a\u8fc7\u5bf9\u72ec\u7acb\u4efb\u52a1\u7684\u8fc7\u62df\u5408\u83b7\u5f97\uff0c\u65e0\u6cd5\u771f\u5b9e\u53cd\u6620\u667a\u80fd\u4f53\u7684\u5b9e\u9645\u80fd\u529b\u3002<\/section>\n<section><\/section>\n<section>\u9274\u4e8e\u6b64\uff0c\u8feb\u5207\u9700\u8981\u65b0\u7684\u8bc4\u4f30\u65b9\u6cd5\u6765\u5f25\u8865\u8fd9\u4e9b\u4e0d\u8db3\u3002\u4e3a\u6b64\uff0cMeta \u63a8\u51fa\u4e86\u4e13\u95e8\u4e3a\u667a\u80fd\u4f53\u7cfb\u7edf\u8bbe\u8ba1\u7684 DevAI \u6570\u636e\u96c6\u3002DevAI \u6db5\u76d6\u4e86 55 \u4e2a AI \u5f00\u53d1\u4efb\u52a1\uff0c\u6d89\u53ca\u76d1\u7763\u5b66\u4e60\u3001\u5f3a\u5316\u5b66\u4e60\u3001\u8ba1\u7b97\u673a\u89c6\u89c9\u548c\u81ea\u7136\u8bed\u8a00\u5904\u7406\u7b49\u9886\u57df\u3002\u6bcf\u4e2a\u4efb\u52a1\u5305\u542b\u7528\u6237\u67e5\u8be2\u3001365 \u4e2a\u4efb\u52a1\u9700\u6c42\u548c 125 \u4e2a\u504f\u597d\u6807\u51c6\u3002<\/section>\n<section><\/section>\n<section>\u4e0e\u4f20\u7edf\u8bc4\u4f30\u65b9\u6cd5\u4e0d\u540c\uff0cDevAI \u4e0d\u4ec5\u5173\u6ce8\u4efb\u52a1\u7684\u6700\u7ec8\u7ed3\u679c\uff0c\u8fd8\u8ddf\u8e2a\u5e76\u8bc4\u4f30\u4efb\u52a1\u6267\u884c\u8fc7\u7a0b\u4e2d\u7684\u6bcf\u4e2a\u9636\u6bb5\uff0c\u4ece\u800c\u63d0\u4f9b\u66f4\u5168\u9762\u7684\u53cd\u9988\uff08\u56fe 4 \u6240\u793a\uff09\u3002\u867d\u7136\u8fd9\u4e9b\u4efb\u52a1\u89c4\u6a21\u76f8\u5bf9\u8f83\u5c0f\uff0c\u4f46\u5b83\u4eec\u771f\u5b9e\u53cd\u6620\u4e86\u5f00\u53d1\u4e2d\u7684\u5b9e\u9645\u95ee\u9898\uff0c\u4e14\u8ba1\u7b97\u6210\u672c\u8f83\u4f4e\uff0c\u9002\u5408\u5e7f\u6cdb\u5e94\u7528\u3002\u503c\u5f97\u6ce8\u610f\u7684\u662f\uff0cDevAI \u4e0d\u5173\u6ce8\u300c\u73a9\u5177\u300d\u6570\u636e\u96c6\uff08\u5982 FashionMNIST\uff09\u4e0a\u7684\u9ad8\u5206\u8868\u73b0\uff0c\u800c\u66f4\u6ce8\u91cd\u667a\u80fd\u4f53\u5728\u5904\u7406\u73b0\u5b9e\u4efb\u52a1\u4e2d\u7684\u80fd\u529b\u3002\u6b64\u5916\uff0cDevAI \u91c7\u7528\u6709\u5411\u65e0\u73af\u56fe\uff08DAG\uff09\u7ed3\u6784\u6392\u5217\u4efb\u52a1\u9700\u6c42\uff0c\u786e\u4fdd\u8bc4\u4f30\u5177\u5907\u5c42\u6b21\u6027\uff0c\u4e0d\u518d\u4f9d\u8d56\u7b80\u5355\u7684\u6210\u529f\u6216\u5931\u8d25\u5224\u65ad\uff0c\u800c\u662f\u8981\u6c42\u667a\u80fd\u4f53\u5177\u5907\u66f4\u6df1\u5165\u7684\u89e3\u51b3\u95ee\u9898\u80fd\u529b\u3002\u672a\u6765\uff0c\u4ee3\u7801\u751f\u6210\u9886\u57df\u7684\u6807\u51c6\u8bc4\u4f30\u65b9\u6cd5\u53ef\u80fd\u4f1a\u91c7\u7528\u7c7b\u4f3c DevAI \u8fd9\u6837\u7684\u6570\u636e\u96c6\uff0c\u63d0\u4f9b\u4e2d\u95f4\u53cd\u9988\uff0c\u4ee5\u6a21\u5757\u5316\u63d0\u5347\u667a\u80fd\u4f53\u7684\u80fd\u529b\uff1b\u5728\u80fd\u529b\u63d0\u5347\u540e\uff0c\u4f7f\u7528 OpenAI \u7684 MLE-Bench \u8fdb\u4e00\u6b65\u8bc4\u4f30\u667a\u80fd\u4f53\u89e3\u51b3\u590d\u6742\u95ee\u9898\u7684\u80fd\u529b\u3002<\/section>\n<section><\/section>\n<section><strong>\u6536\u96c6\u4eba\u7c7b\u4e13\u5bb6\u8bc4\u4f30<\/strong><\/section>\n<section><\/section>\n<section><strong>\u4eba\u7c7b\u8bc4\u4f30\u8bbe\u7f6e<\/strong><\/section>\n<section><\/section>\n<section>\u5728\u5b8c\u6210\u57fa\u7ebf\u6267\u884c\u7ed3\u679c\u548c\u57fa\u672c\u7edf\u8ba1\u5206\u6790\u540e\uff0c\u7814\u7a76\u56e2\u961f\u9080\u8bf7\u4e86\u4e09\u4f4d\u5177\u5907 5 \u5e74\u4ee5\u4e0a AI \u5f00\u53d1\u7ecf\u9a8c\u4eba\u7c7b\u4e13\u5bb6\u8bc4\u4f30\u5458\uff08\u533f\u540d\u4e3a 231a\u300138bb \u548c cn90\uff09\u5bf9 AI \u5f00\u53d1\u8005\u7684\u57fa\u7ebf\u8f93\u51fa\u8fdb\u884c\u5ba1\u67e5\uff0c\u8bc4\u4f30\u6bcf\u9879\u9700\u6c42\u662f\u5426\u5f97\u5230\u4e86\u6ee1\u8db3\u3002\u8bc4\u4f30\u5206\u4e3a\u4e24\u8f6e\u3002\u4e3a\u4e86\u5c3d\u91cf\u6355\u6349\u4eba\u7c7b\u8bc4\u4f30\u4e2d\u5e38\u89c1\u7684\u504f\u5dee\uff08\u6a21\u62df\u5b9e\u9645\u90e8\u7f72\u573a\u666f\uff09\uff0c\u5728\u7b2c\u4e00\u8f6e\u4e2d\uff0c\u8bc4\u4f30\u5458\u8ba8\u8bba\u4e86\u57fa\u672c\u6807\u51c6\u3002\u867d\u7136\u5141\u8bb8\u8bc4\u4f30\u5458\u5e26\u6709\u4e2a\u4eba\u504f\u597d\uff0c\u4f46\u8bc4\u4f30\u8fc7\u7a0b\u9700\u57fa\u4e8e\u7edf\u4e00\u7684\u6807\u51c6\u8fdb\u884c\u3002\u5728\u7b2c\u4e00\u8f6e\u8bc4\u4f30\u5b8c\u6210\u540e\uff08\u603b\u8ba1\u7ea6 58 \u5c0f\u65f6\uff09\uff0c\u8bc4\u4f30\u5458\u518d\u6b21\u8fdb\u884c\u8ba8\u8bba\uff0c\u8fdb\u4e00\u6b65\u4fee\u6b63\u548c\u8fbe\u6210\u4e00\u81f4\u610f\u89c1\uff0c\u786e\u4fdd\u8bc4\u4f30\u7ed3\u679c\u66f4\u52a0\u7edf\u4e00\u548c\u4e00\u81f4\u3002\u8fd9\u4e00\u8fc7\u7a0b\u5171\u8017\u65f6 28.5 \u5c0f\u65f6\uff0c\u6700\u7ec8\u7684\u5171\u8bc6\u4f5c\u4e3a\u6bcf\u79cd\u65b9\u6cd5\u7684\u6700\u7ec8\u4eba\u7c7b\u8bc4\u4f30\u7ed3\u679c\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-34c7f123b879b5b2910846a2740ee1ce.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-34c7f123b879b5b2910846a2740ee1ce.png\"><\/a><\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-ee1770300d15a67f49e60acad4aeac36.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-ee1770300d15a67f49e60acad4aeac36.png\"><\/a><\/section>\n<section><strong>\u6027\u80fd\u5206\u6790\u3002<\/strong>\u5b9e\u9a8c\u7ed3\u679c\u663e\u793a\uff08\u5982\u8868 2\uff09\uff0c\u8868\u73b0\u6700\u597d\u7684\u4e24\u79cd\u65b9\u6cd5 \u2014\u2014GPT-Pilot \u548c OpenHands\u2014\u2014 \u4ec5\u80fd\u6ee1\u8db3\u7ea6 29% \u7684\u9700\u6c42\uff08\u5ffd\u7565\u524d\u63d0\u6761\u4ef6\u540e\u4e3a 44%\uff09\uff0c\u4e14\u4ec5\u5728\u4e00\u4e2a\u4efb\u52a1\u4e2d\u6ee1\u8db3\u4e86\u6240\u6709\u8981\u6c42\u3002\u8fd9\u8868\u660e DevAI \u4e3a\u5f53\u524d\u53ca\u672a\u6765\u7684\u667a\u80fd\u4f53\u65b9\u6cd5\u8bbe\u5b9a\u4e86\u8f83\u9ad8\u7684\u6311\u6218\u6027\u3002\u6b64\u5916\uff0c\u6b63\u5982\u7b2c 2 \u8282\u6240\u8ba8\u8bba\u7684\uff0cDevAI \u4e0d\u4ec5\u63ed\u793a\u4e86\u4efb\u52a1\u6700\u7ec8\u7ed3\u679c\uff0c\u8fd8\u901a\u8fc7\u53cd\u9988\u63ed\u793a\u4e86\u667a\u80fd\u4f53\u5728\u4efb\u52a1\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u7684\u95ee\u9898\uff0c\u4e3a\u8bc4\u4f30\u63d0\u4f9b\u4e86\u66f4\u4e30\u5bcc\u7684\u5c42\u6b21\u3002<\/section>\n<section><\/section>\n<section><strong>\u9519\u8bef\u5206\u6790\u3002<\/strong>\u5728\u5b9e\u9a8c\u4e2d\uff0c\u8bc4\u4f30\u5458\u5728\u521d\u6b65\u8bc4\u4f30\u540e\u8fdb\u884c\u4e86\u6df1\u5165\u8fa9\u8bba\uff0c\u76f4\u5230\u4ed6\u4eec\u5bf9\u6bcf\u4e2a\u4efb\u52a1\u7684\u9700\u6c42\u8fbe\u6210\u4e00\u81f4\u610f\u89c1\u3002\u5171\u8bc6\u8bc4\u4f30\uff08consensus\uff09\u901a\u8fc7\u8fd9\u79cd\u65b9\u5f0f\u6a21\u62df\u5b9e\u9645\u60c5\u51b5\uff0c\u51cf\u5c11\u4e86\u4e2a\u4f53\u8bc4\u4f30\u4e2d\u7684\u504f\u5dee\u3002\u5728 Human-as-a-Judge \u6846\u67b6\u4e0b\uff0c\u8bc4\u4f30\u5458\u53ef\u4ee5\u901a\u8fc7\u8ba8\u8bba\u548c\u8bc1\u636e\u4fee\u6b63\u81ea\u5df1\u7684\u5224\u65ad\uff0c\u4ece\u800c\u8c03\u6574\u8bc4\u4f30\u7ed3\u679c\u3002\u8fd9\u79cd\u65b9\u5f0f\u4e5f\u7528\u6765\u8fd1\u4f3c\u4f30\u8ba1\u4e2a\u4f53\u7684\u9519\u8bef\u7387\u3002\u7406\u8bba\u4e0a\uff0c\u96c6\u4f53\u8ba8\u8bba\u8fbe\u6210\u7684\u5171\u8bc6\u5e94\u6bd4\u4efb\u4f55\u4e2a\u4f53\u8bc4\u4f30\u66f4\u63a5\u8fd1\u771f\u5b9e\u7ed3\u679c\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-2fca7e5c617743cc8818d832051bbbe7.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-2fca7e5c617743cc8818d832051bbbe7.png\"><\/a><\/section>\n<section>\u867d\u7136\u5171\u8bc6\u8bc4\u4f30\u53ef\u80fd\u5e76\u4e0d\u5b8c\u7f8e\uff08\u67d0\u4e9b\u9519\u8bef\u4ecd\u7136\u5b58\u5728\uff09\uff0c\u4f46\u76f8\u8f83\u4e8e\u4e2a\u4f53\u8bc4\u4f30\uff0c\u7406\u8bba\u4e0a\u5171\u8bc6\u8bc4\u4f30\u5e94\u66f4\u63a5\u8fd1\u771f\u5b9e\u7ed3\u679c\u3002\u5982\u56fe 5 \u6240\u793a\uff0c\u5b9e\u9a8c\u7ed3\u679c\u8bc1\u5b9e\u4e86\u8fd9\u4e00\u5047\u8bbe\u3002\u5c3d\u7ba1\u8bc4\u4f30\u5458\u4e4b\u95f4\u7684\u9519\u8bef\u7387\u6709\u6240\u4e0d\u540c\uff0c\u4f46\u591a\u6570\u6295\u7968\u6709\u6548\u5730\u4fee\u6b63\u4e86\u5927\u90e8\u5206\u9519\u8bef\u3002\u4f8b\u5982\uff0c\u8bc4\u4f30\u5458 cn90 \u5728\u8bc4\u4f30 GPT-Pilot \u65f6\u72af\u4e0b\u4e86\u6700\u591a\u7684\u9519\u8bef\uff08\u9519\u8bef\u7387\u8fbe 23.77%\uff09\u3002\u7136\u800c\uff0c\u901a\u8fc7\u591a\u6570\u6295\u7968\uff08majority vote\uff09\uff0c\u4e09\u4f4d\u8bc4\u4f30\u5458\u7684\u6574\u4f53\u9519\u8bef\u7387\u964d\u4f4e\u81f3 6.01%\uff0c\u663e\u793a\u4e86\u591a\u6570\u6295\u7968\u5728\u51cf\u5c11\u8bc4\u4f30\u504f\u5dee\u65b9\u9762\u7684\u4f18\u52bf\u3002<\/section>\n<section><\/section>\n<section><strong>\u7ed3\u8bba\u3002<\/strong>\u4eba\u7c7b\u8bc4\u4f30\u4e2d\u7684\u9519\u8bef\u662f\u4e0d\u53ef\u907f\u514d\u7684\u3002\u4e3a\u51cf\u5c11\u8fd9\u4e9b\u9519\u8bef\uff0c\u7814\u7a76\u63d0\u51fa\u4e86\u4e24\u79cd\u7b56\u7565\u3002\u7b2c\u4e00\uff0c\u50cf\u672c\u7814\u7a76\u4e00\u6837\uff0c\u5728\u6bcf\u6b21\u8bc4\u4f30\u540e\u5f15\u5165\u8ba8\u8bba\u73af\u8282\uff0c\u8bc4\u4f30\u5458\u53ef\u4ee5\u6839\u636e\u65b0\u7684\u8bc1\u636e\u8c03\u6574\u4ed6\u4eec\u7684\u5224\u65ad\u3002\u8fd9\u4e00\u65b9\u6cd5\u5728\u8bc4\u4f30\u5458\u6570\u91cf\u8f83\u5c11\u65f6\u5c24\u5176\u6709\u6548\uff0c\u56e0\u4e3a\u5c0f\u7ec4\u8bc4\u4f30\u4e2d\u7684\u591a\u6570\u6295\u7968\u4ecd\u53ef\u80fd\u4ea7\u751f\u4e00\u5b9a\u8bef\u5dee\uff08\u5982\u56fe 5 \u6240\u793a\uff0c\u76f8\u6bd4\u5171\u8bc6\u8bc4\u4f30\u5927\u7ea6\u6709 5% \u7684\u9519\u8bef\u7387\uff09\u3002\u7b2c\u4e8c\uff0c\u7ec4\u5efa\u66f4\u5927\u7684\u4e13\u5bb6\u56e2\u961f\u6765\u63d0\u9ad8\u8bc4\u4f30\u7684\u51c6\u786e\u6027\u3002\u7814\u7a76\u8868\u660e\uff0c\u5f53\u8bc4\u4f30\u5458\u4eba\u6570\u8d85\u8fc7 5 \u4eba\u65f6\uff0c\u8bc4\u4f30\u51c6\u786e\u6027\u6709\u671b\u8d85\u8fc7 50%\u3002\u7136\u800c\uff0c\u7531\u4e8e\u52a8\u5458\u66f4\u591a\u4e13\u5bb6\u7684\u6210\u672c\u8f83\u9ad8\uff0c\u5b9e\u8df5\u4e2d\u8fd9\u79cd\u65b9\u6cd5\u5e76\u4e0d\u603b\u662f\u53ef\u884c\u3002\u56e0\u6b64\uff0c\u672c\u7814\u7a76\u66f4\u503e\u5411\u4e8e\u901a\u8fc7\u8ba8\u8bba\u548c\u5171\u8bc6\u6295\u7968\u6765\u51cf\u5c11\u8bc4\u4f30\u4e2d\u7684\u504f\u5dee\u3002<\/section>\n<section><\/section>\n<section><strong>\u4f7f\u7528<\/strong><\/section>\n<section><\/section>\n<section>\u76ee\u524d\uff0c\u4f5c\u8005\u5df2\u5728 GitHub \u4e0a\u63d0\u4f9b\u4e86\u5f00\u6e90\u4ee3\u7801\uff0c\u652f\u6301\u5bf9\u4efb\u610f\u5de5\u4f5c\u533a\uff08workspace\uff09\u8fdb\u884c\u63d0\u95ee\uff0c\u5e76\u901a\u8fc7 Agent-as-a-Judge \u529f\u80fd\u5728 DevAI \u6570\u636e\u96c6\u4e0a\u8fdb\u884c\u81ea\u52a8\u8bc4\u4f30\u3002\u672a\u6765\u7684\u5f00\u6e90\u667a\u80fd\u4f53\u8bc4\u4f30\u5c06\u8fdb\u4e00\u6b65\u6539\u8fdb\uff0c\u9996\u5148\u5229\u7528 DevAI \u5bf9\u4e2d\u95f4\u8fc7\u7a0b\u8fdb\u884c\u5224\u65ad\u548c\u4f18\u5316\uff0c\u6700\u7ec8\u901a\u8fc7\u7c7b\u4f3c MLE-Bench \u7684\u5de5\u5177\u6d4b\u8bd5\u667a\u80fd\u4f53\u7684\u6574\u4f53\u6027\u80fd\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-d995647d57f5800ff62528d400df0494.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-d995647d57f5800ff62528d400df0494.png\"><\/a><\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-999c2a188bbf25f3f03691d43d28a685.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-999c2a188bbf25f3f03691d43d28a685.png\"><\/a><\/section>\n<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-10-18-4\" target=\"_blank\">\u5377\u8d77\u6765\uff01\u8ba9\u667a\u80fd\u4f53\u8bc4\u4f30\u667a\u80fd\u4f53\uff0cMeta\u53d1\u5e03Agent-as-a-Judge<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:\u5377\u8d77\u6765\uff01\u8ba9\u667a\u80fd\u4f53\u8bc4\u4f30\u667a [&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":[70,68,67,66],"class_list":["post-36180","post","type-post","status-publish","format-standard","hentry","category-news","tag-agent","tag-68","tag-67","tag-66"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v26.4 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>\u5377\u8d77\u6765\uff01\u8ba9\u667a\u80fd\u4f53\u8bc4\u4f30\u667a\u80fd\u4f53\uff0cMeta\u53d1\u5e03Agent-as-a-Judge - \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=36180\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/17aitech.com\/?p=36180\",\"url\":\"https:\/\/17aitech.com\/?p=36180\",\"name\":\"\u5377\u8d77\u6765\uff01\u8ba9\u667a\u80fd\u4f53\u8bc4\u4f30\u667a\u80fd\u4f53\uff0cMeta\u53d1\u5e03Agent-as-a-Judge - \u4e00\u8d77AI\u6280\u672f\",\"isPartOf\":{\"@id\":\"https:\/\/17aitech.com\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/17aitech.com\/?p=36180#primaryimage\"},\"image\":{\"@id\":\"https:\/\/17aitech.com\/?p=36180#primaryimage\"},\"thumbnailUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-9f302a32718088e5b3475af407ceaac1.png\",\"datePublished\":\"2025-01-06T19:02:13+00:00\",\"author\":{\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d\"},\"breadcrumb\":{\"@id\":\"https:\/\/17aitech.com\/?p=36180#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/17aitech.com\/?p=36180\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/17aitech.com\/?p=36180#primaryimage\",\"url\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-9f302a32718088e5b3475af407ceaac1.png\",\"contentUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-9f302a32718088e5b3475af407ceaac1.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/17aitech.com\/?p=36180#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/17aitech.com\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u5377\u8d77\u6765\uff01\u8ba9\u667a\u80fd\u4f53\u8bc4\u4f30\u667a\u80fd\u4f53\uff0cMeta\u53d1\u5e03Agent-as-a-Judge\"}]},{\"@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":"\u5377\u8d77\u6765\uff01\u8ba9\u667a\u80fd\u4f53\u8bc4\u4f30\u667a\u80fd\u4f53\uff0cMeta\u53d1\u5e03Agent-as-a-Judge - \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=36180","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/17aitech.com\/?p=36180","url":"https:\/\/17aitech.com\/?p=36180","name":"\u5377\u8d77\u6765\uff01\u8ba9\u667a\u80fd\u4f53\u8bc4\u4f30\u667a\u80fd\u4f53\uff0cMeta\u53d1\u5e03Agent-as-a-Judge - \u4e00\u8d77AI\u6280\u672f","isPartOf":{"@id":"https:\/\/17aitech.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/17aitech.com\/?p=36180#primaryimage"},"image":{"@id":"https:\/\/17aitech.com\/?p=36180#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-9f302a32718088e5b3475af407ceaac1.png","datePublished":"2025-01-06T19:02:13+00:00","author":{"@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"breadcrumb":{"@id":"https:\/\/17aitech.com\/?p=36180#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/17aitech.com\/?p=36180"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/?p=36180#primaryimage","url":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-9f302a32718088e5b3475af407ceaac1.png","contentUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/10\/frc-9f302a32718088e5b3475af407ceaac1.png"},{"@type":"BreadcrumbList","@id":"https:\/\/17aitech.com\/?p=36180#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/17aitech.com\/"},{"@type":"ListItem","position":2,"name":"\u5377\u8d77\u6765\uff01\u8ba9\u667a\u80fd\u4f53\u8bc4\u4f30\u667a\u80fd\u4f53\uff0cMeta\u53d1\u5e03Agent-as-a-Judge"}]},{"@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\/36180","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=36180"}],"version-history":[{"count":0,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/36180\/revisions"}],"wp:attachment":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=36180"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=36180"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=36180"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}