{"id":1142564,"date":"2025-06-19T03:17:24","date_gmt":"2025-06-19T10:17:24","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-blog-post&p=1142564"},"modified":"2025-06-19T03:17:26","modified_gmt":"2025-06-19T10:17:26","slug":"fea-bench","status":"publish","type":"msr-blog-post","link":"https:\/\/www.microsoft.com\/en-us\/research\/articles\/fea-bench\/","title":{"rendered":"FEA-Bench\uff1a\u9996\u4e2a\u4ed3\u5e93\u7ea7\u65b0\u529f\u80fd\u5b9e\u73b0\u57fa\u51c6\uff0c\u8ba9\u5927\u6a21\u578b\u66f4\u61c2\u8f6f\u4ef6\u5f00\u53d1"},"content":{"rendered":"\n
\u7f16\u8005\u6309\uff1a\u5f53\u524d\uff0c\u5927\u8bed\u8a00\u6a21\u578b\u5728\u4ee3\u7801\u751f\u6210\u9886\u57df\u5df2\u5c55\u73b0\u51fa\u60ca\u4eba\u7684\u80fd\u529b\uff0c\u4f46\u80fd\u5426\u80dc\u4efb\u771f\u5b9e\u8f6f\u4ef6\u5f00\u53d1\u4e2d\u7684\u201c\u65b0\u589e\u529f\u80fd\u5b9e\u73b0\u201d\u4efb\u52a1\uff0c\u4ecd\u662f\u4e00\u4e2a\u5173\u952e\u672a\u89e3\u7684\u95ee\u9898\u3002\u5bf9\u6b64\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u4e0e\u5317\u4eac\u5927\u5b66\u8054\u5408\u53d1\u5e03\u4e86\u9996\u4e2a\u4e13\u6ce8\u4e8e\u4ed3\u5e93\u7ea7\u65b0\u529f\u80fd\u5b9e\u73b0\u7684\u57fa\u51c6\u6d4b\u8bd5 FEA-Bench\uff0c\u586b\u8865\u4e86\u8bc4\u4f30\u4f53\u7cfb\u4e2d\u7684\u91cd\u8981\u7a7a\u767d\u3002\u8be5\u6d4b\u8bd5\u96c6\u6784\u5efa\u4e8e\u771f\u5b9e\u5f00\u6e90\u9879\u76ee\u7684 pull request \uff08\u5408\u5e76\u8bf7\u6c42\uff09\uff0c\u8986\u76d61400\u591a\u4e2a\u9ad8\u8d28\u91cf\u4efb\u52a1\uff0c\u7cfb\u7edf\u8bc4\u4f30\u4e86\u4e3b\u6d41\u5927\u6a21\u578b\u5728\u590d\u6742\u5de5\u7a0b\u4efb\u52a1\u4e2d\u7684\u8868\u73b0\u3002FEA-Bench \u4e0d\u4ec5\u4e3a\u63a8\u52a8\u4ee3\u7801\u751f\u6210\u7cfb\u7edf\u8fc8\u5411\u771f\u5b9e\u4e16\u754c\u7684\u65b0\u529f\u80fd\u5b9e\u73b0\u4efb\u52a1\u63d0\u4f9b\u4e86\u575a\u5b9e\u652f\u6491\uff0c\u4e5f\u4e3a\u76f8\u5173\u7814\u7a76\u62d3\u5c55\u4e86\u66f4\u5e7f\u9614\u7684\u63a2\u7d22\u7a7a\u95f4\u3002<\/p>\n\n\n\n
\u968f\u7740\u4eba\u5de5\u667a\u80fd\u7684\u5feb\u901f\u53d1\u5c55\uff0c\u5927\u8bed\u8a00\u6a21\u578b\u5728\u4ee3\u7801\u751f\u6210\u9886\u57df\u5c55\u73b0\u51fa\u8d8a\u6765\u8d8a\u591a\u7684\u53ef\u80fd\u6027\uff0c\u4ece\u51fd\u6570\u7ea7\u522b\u7684\u8865\u5168\u5230\u590d\u6742\u95ee\u9898\u7684\u4fee\u590d\uff0cAI \u5df2\u9010\u6e10\u6e17\u900f\u81f3\u5f00\u53d1\u8005\u7684\u5de5\u4f5c\u6d41\u7a0b\u3002\u7136\u800c\uff0c\u5728\u771f\u5b9e\u7684\u8f6f\u4ef6\u5de5\u7a0b\u573a\u666f\u4e2d\uff0c\u5927\u8bed\u8a00\u6a21\u578b\u662f\u5426\u5177\u5907\u5b9e\u73b0\u65b0\u529f\u80fd\u7684\u80fd\u529b\uff0c\u4ecd\u662f\u4e00\u4e2a\u4e9f\u5f85\u9a8c\u8bc1\u7684\u91cd\u8981\u8bfe\u9898\u3002\u4e0e\u6b64\u540c\u65f6\uff0c\u73b0\u6709\u7684\u57fa\u51c6\u6d4b\u8bd5\u591a\u805a\u7126\u4e8e\u5927\u6a21\u578b\u7684\u72ec\u7acb\u7f16\u7a0b\u95ee\u9898\u6216\u4fee\u590d\u4efb\u52a1\uff0c\u7f3a\u4e4f\u5bf9\u201c\u65b0\u529f\u80fd\u5b9e\u73b0\u201d\u8fd9\u4e00\u6838\u5fc3\u80fd\u529b\u7684\u7cfb\u7edf\u6027\u8bc4\u4f30\u3002<\/p>\n\n\n\n
\u5bf9\u6b64\uff0c\u6765\u81ea\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u4e0e\u5317\u4eac\u5927\u5b66\u7684\u7814\u7a76\u56e2\u961f\u8054\u5408\u63a8\u51fa\u4e86\u9996\u4e2a\u9762\u5411\u4ed3\u5e93\u7ea7\u65b0\u529f\u80fd\u5b9e\u73b0\u7684\u57fa\u51c6\u6d4b\u8bd5 FEA-Bench<\/strong>\uff0c\u586b\u8865\u4e86\u5f53\u524d\u4ee3\u7801\u751f\u6210\u8bc4\u4f30\u4f53\u7cfb\u4e2d\u7684\u5173\u952e\u4e00\u73af\u3002\u8be5\u57fa\u51c6\u6d4b\u8bd5\u901a\u8fc71401\u4e2a\u9ad8\u8d28\u91cf\u4efb\u52a1\u5b9e\u4f8b\uff0c\u5168\u9762\u8bc4\u4f30\u4e86\u5927\u8bed\u8a00\u6a21\u578b\u5728\u65b0\u589e\u529f\u80fd\u65f6\u6240\u9700\u7684\u4ee3\u7801\u751f\u6210\u4e0e\u7f16\u8f91\u80fd\u529b\uff0c\u63ed\u793a\u4e86\u5f53\u524d\u6a21\u578b\u5728\u590d\u6742\u8f6f\u4ef6\u5de5\u7a0b\u4efb\u52a1\u4e2d\u7684\u5c40\u9650\u6027\uff0c\u4e3a\u672a\u6765\u7684\u7814\u7a76\u63d0\u4f9b\u4e86\u91cd\u8981\u65b9\u5411\u3002<\/p>\n\n\n\n \u5c3d\u7ba1\u5f53\u524d\u7684\u5927\u8bed\u8a00\u6a21\u578b\u5982 GPT-4\u3001DeepSeek-Coder \u548c Qwen2.5-Coder \u7b49\u5728\u4ee3\u7801\u8865\u5168\u3001\u51fd\u6570\u7ea7\u751f\u6210\u4e43\u81f3\u4fee\u590d GitHub \u8bae\u9898\uff08issue\uff09\u7b49\u4efb\u52a1\u4e2d\u8868\u73b0\u51fa\u8272\uff0c\u4f46\u5728\u5b9e\u9645\u7684\u8f6f\u4ef6\u5f00\u53d1\u4e2d\uff0c\u4e3a\u4ee3\u7801\u4ed3\u5e93\u589e\u6dfb\u65b0\u529f\u80fd\u662f\u4e00\u4e2a\u66f4\u91cd\u8981\u4e14\u96be\u5ea6\u66f4\u5927\u7684\u9700\u6c42\u3002\u5f00\u53d1\u8005\u5f80\u5f80\u9700\u8981\u5728\u5df2\u6709\u4ee3\u7801\u5e93\u4e2d\u6dfb\u52a0\u65b0\u7ec4\u4ef6\uff08\u5982\u51fd\u6570\u3001\u7c7b\uff09\uff0c\u5e76\u76f8\u5e94\u4fee\u6539\u5176\u4ed6\u76f8\u5173\u4ee3\u7801\u4ee5\u786e\u4fdd\u6574\u4f53\u7684\u4e00\u81f4\u6027\u3002\u8fd9\u79cd\u4ed3\u5e93\u7ea7\u589e\u91cf\u5f00\u53d1\u80fd\u529b\u662f\u8861\u91cf\u5927\u8bed\u8a00\u6a21\u578b\u80fd\u5426\u80dc\u4efb\u771f\u5b9e\u8f6f\u4ef6\u5de5\u7a0b\u4efb\u52a1\u7684\u6838\u5fc3\u6307\u6807\u3002\u800c\u4e14\uff0c\u76ee\u524d\u5c1a\u65e0\u4e13\u95e8\u9488\u5bf9\u8be5\u80fd\u529b\u8bbe\u8ba1\u7684\u8bc4\u4f30\u57fa\u51c6\u3002<\/p>\n\n\n\n FEA-Bench \u7684\u51fa\u73b0\u586b\u8865\u4e86\u8fd9\u4e00\u7a7a\u767d\u3002\u8be5\u6570\u636e\u96c6\u57fa\u4e8e83\u4e2a\u5f00\u6e90 Python \u9879\u76ee\u7684\u771f\u5b9e pull request \uff08\u5408\u5e76\u8bf7\u6c42\uff09\u6570\u636e\u6784\u5efa\u800c\u6210\uff0c\u5305\u542b1401\u4e2a\u65b0\u589e\u529f\u80fd\u7684\u4efb\u52a1\u5b9e\u4f8b\uff0c\u5e76\u914d\u5957\u5b8c\u6574\u7684\u5355\u5143\u6d4b\u8bd5\u9a8c\u8bc1\u673a\u5236\uff0c\u4ece\u800c\u4fdd\u8bc1\u751f\u6210\u7ed3\u679c\u7684\u53ef\u6267\u884c\u6027\u4e0e\u51c6\u786e\u6027\u3002\u6bcf\u4e2a\u4efb\u52a1\u5b9e\u4f8b\u5747\u8981\u6c42\u6a21\u578b\u4e0d\u4ec5\u8981\u751f\u6210\u65b0\u7684\u4ee3\u7801\u7ec4\u4ef6\uff0c\u8fd8\u9700\u7406\u89e3\u6574\u4e2a\u4ed3\u5e93\u7ed3\u6784\u5e76\u5728\u5fc5\u8981\u65f6\u8fdb\u884c\u8054\u52a8\u4fee\u6539\uff0c\u4ece\u800c\u5b9e\u73b0\u7aef\u5230\u7aef\u7684\u529f\u80fd\u5b9e\u73b0\u3002<\/p>\n\n\n\n \u4e0d\u540c\u4e8e\u4ee5\u5f80\u4e3b\u8981\u5173\u6ce8 bug \u4fee\u590d\uff08\u5982 SWE-bench\uff09\u6216\u72ec\u7acb\u7f16\u7a0b\uff08\u5982 HumanEval\uff09\u4efb\u52a1\u7684\u57fa\u51c6\u6d4b\u8bd5\uff0cFEA-Bench \u5c06\u76ee\u6807\u805a\u7126\u4e8e\u65b0\u529f\u80fd\u7684\u5b9e\u73b0\u3002\u5176\u4efb\u52a1\u5b9e\u4f8b\u5168\u90e8\u6765\u81ea\u771f\u5b9e\u7684 GitHub pull request\uff0c\u6db5\u76d6\u591a\u4e2a\u884c\u4e1a\u573a\u666f\uff0c\u5177\u6709\u8f83\u9ad8\u7684\u4ee3\u8868\u6027\u548c\u5b9e\u7528\u4ef7\u503c\u3002\u8fd9\u6837\u7684\u8bbe\u8ba1\u4e0d\u4ec5\u4f7f\u6570\u636e\u96c6\u66f4\u8d34\u8fd1\u771f\u5b9e\u7684\u8f6f\u4ef6\u5f00\u53d1\u60c5\u5883\uff0c\u4e5f\u4e3a\u5927\u6a21\u578b\u5728\u5de5\u4e1a\u573a\u666f\u4e2d\u7684\u843d\u5730\u63d0\u4f9b\u4e86\u6307\u5bfc\u610f\u4e49\u3002<\/p>\n\n\n\n \u4e3a\u4e86\u786e\u4fdd\u4efb\u52a1\u8bbe\u7f6e\u7684\u771f\u5b9e\u6027\u4e0e\u6311\u6218\u6027\uff0cFEA-Bench \u6784\u5efa\u4e86\u4e00\u5957\u591a\u7ef4\u5ea6\u7684\u96be\u5ea6\u63a7\u5236\u4e0e\u8d28\u91cf\u4fdd\u969c\u673a\u5236\u3002\u5728\u6570\u636e\u7b5b\u9009\u9636\u6bb5\uff0c\u7cfb\u7edf\u4f1a\u7ed3\u5408\u89c4\u5219\u8fc7\u6ee4\u4e0e\u610f\u56fe\u8bc6\u522b\u6280\u672f\uff0c\u7cbe\u51c6\u7504\u522b\u51fa\u786e\u5b9e\u6d89\u53ca\u65b0\u529f\u80fd\u5f00\u53d1\u7684 pull request\u3002\u6bcf\u4e2a\u4efb\u52a1\u5b9e\u4f8b\u5747\u914d\u5907\u8be6\u5c3d\u7684\u529f\u80fd\u8bf7\u6c42\u63cf\u8ff0\u3001\u65b0\u589e\u7ec4\u4ef6\u5b9a\u4e49\uff08\u5305\u62ec\u51fd\u6570\u7b7e\u540d\u3001\u7c7b\u7ed3\u6784\u7b49\uff09\u3001\u73af\u5883\u914d\u7f6e\u4fe1\u606f\u53ca\u5355\u5143\u6d4b\u8bd5\u7528\u4f8b\uff0c\u5e76\u6807\u6ce8\u9ec4\u91d1\u8865\u4e01\uff08gold patch\uff09\u4f5c\u4e3a\u53c2\u8003\u5b9e\u73b0\u3002\u6240\u6709\u5b9e\u4f8b\u5728\u53d1\u5e03\u524d\u5747\u901a\u8fc7\u81ea\u52a8\u5355\u5143\u6d4b\u8bd5\u9a8c\u8bc1\uff0c\u786e\u4fdd\u6570\u636e\u96c6\u5177\u6709\u826f\u597d\u7684\u53ef\u9a8c\u8bc1\u6027\u548c\u53ef\u590d\u73b0\u6027\uff0c\u4e3a\u8bc4\u4f30\u6a21\u578b\u6027\u80fd\u63d0\u4f9b\u4e86\u7a33\u5b9a\u53ef\u9760\u7684\u57fa\u7840\u3002<\/p>\n\n\n\n \u6b64\u5916\uff0cFEA-Bench \u8fd8\u63d0\u4f9b\u4e86\u4e00\u4e2a\u8f7b\u91cf\u7248\u672c FEA-Bench Lite\uff0c\u7528\u4e8e\u9002\u914d\u4e0d\u540c\u5b9e\u9a8c\u6761\u4ef6\u4e0b\u7684\u8ba1\u7b97\u8d44\u6e90\u9700\u6c42\u3002\u8be5\u5b50\u96c6\u5728\u4fdd\u7559\u4ee3\u8868\u6027\u7684\u524d\u63d0\u4e0b\uff0c\u7b5b\u9009\u51fa\u4e86\u590d\u6742\u5ea6\u76f8\u5bf9\u8f83\u4f4e\u7684\u4efb\u52a1\u5b9e\u4f8b\uff0c\u65b9\u4fbf\u79d1\u7814\u4eba\u5458\u5728\u8d44\u6e90\u53d7\u9650\u7684\u73af\u5883\u4e2d\u8fdb\u884c\u6a21\u578b\u7684\u5feb\u901f\u8fed\u4ee3\u4e0e\u8bc4\u4f30\u3002<\/p>\n\n\n\n \u7814\u7a76\u56e2\u961f\u4f7f\u7528 FEA-Bench \u7cfb\u7edf\u5730\u8bc4\u4f30\u4e86 GPT-4\u3001DeepSeek-R1\u3001Qwen2.5-Coder \u7b49\u76ee\u524d\u4e3b\u6d41\u7684\u5927\u8bed\u8a00\u6a21\u578b\u3002\u7ed3\u679c\u663e\u793a\uff0c\u5728\u4e0d\u4f7f\u7528\u4ee3\u7406\uff08agent\uff09\u6846\u67b6\u7684\u60c5\u51b5\u4e0b\uff0c\u5f53\u524d\u6700\u4f18\u6a21\u578b DeepSeek-R1 \u5728 Oracle \u8bbe\u7f6e\u4e2d\uff08\u7ed9\u5b9a\u9700\u8981\u7f16\u8f91\u7684\u4ee3\u7801\u6587\u4ef6\u4f5c\u4e3a\u4e0a\u4e0b\u6587\u8f93\u5165\uff09\u7684\u89e3\u51b3\u7387\u4ec5\u4e3a9.92%\u3002\u76f8\u6bd4\u4e4b\u4e0b\uff0cSWE-bench \u4e2d\u6a21\u578b\u7684 bug \u4fee\u590d\u6210\u529f\u7387\u666e\u904d\u572830%\u4ee5\u4e0a\u3002\u8fd9\u4e00\u5dee\u8ddd\u5145\u5206\u8bf4\u660e\uff0c\u201c\u65b0\u589e\u529f\u80fd\u201d\u4efb\u52a1\u5728\u903b\u8f91\u590d\u6742\u5ea6\u3001\u8de8\u6587\u4ef6\u4e00\u81f4\u6027\u7ef4\u62a4\u4e0e\u903b\u8f91\u6574\u5408\u7b49\u65b9\u9762\uff0c\u90fd\u5bf9\u5927\u8bed\u8a00\u6a21\u578b\u63d0\u51fa\u4e86\u66f4\u9ad8\u7684\u8981\u6c42\u3002<\/p>\n\n\n\n FEA-Bench \u7684\u53d1\u5e03\u4e3a\u4ee3\u7801\u751f\u6210\u7814\u7a76\u6253\u5f00\u4e86\u4e00\u4e2a\u5168\u65b0\u7684\u7ef4\u5ea6\u3002\u5728\u4e3a\u5b66\u672f\u754c\u63d0\u4f9b\u5168\u65b0\u7684\u6d4b\u8bd5\u5e73\u53f0\u7684\u540c\u65f6\uff0cFEA-Bench \u4e5f\u4e3a\u5de5\u4e1a\u754c\u8bc4\u4f30 AI \u8f85\u52a9\u5f00\u53d1\u8f6f\u4ef6\u65b0\u529f\u80fd\u63d0\u4f9b\u4f9d\u636e\u6807\u51c6\u3002\u7814\u7a76\u56e2\u961f\u5e0c\u671b\u901a\u8fc7 FEA-Bench\uff0c\u53ef\u4ee5\u8ba9\u66f4\u591a\u4eba\u5173\u6ce8 AI \u5728\u9762\u5bf9\u81ea\u7531\u5ea6\u8f83\u9ad8\u7684\u65b0\u529f\u80fd\u5b9e\u73b0\u4e2d\u7684\u8868\u73b0\uff0c\u6fc0\u53d1\u66f4\u591a\u79d1\u7814\u4eba\u5458\u63a2\u7d22\u5982\u4f55\u8ba9 AI \u771f\u6b63\u7406\u89e3\u5e76\u6df1\u5ea6\u53c2\u4e0e\u73b0\u4ee3\u8f6f\u4ef6\u5de5\u7a0b\u7684\u6838\u5fc3\u3002<\/p>\n\n\n\n \u672a\u6765\uff0c\u7814\u7a76\u56e2\u961f\u5c06\u6301\u7eed\u62d3\u5c55\u4efb\u52a1\u7684\u8986\u76d6\u8303\u56f4\u3001\u4f18\u5316\u68c0\u7d22\u673a\u5236\u3001\u63d0\u5347\u751f\u6210\u683c\u5f0f\u7684\u517c\u5bb9\u6027\uff0c\u8fdb\u4e00\u6b65\u63a8\u52a8\u5927\u8bed\u8a00\u6a21\u578b\u5411\u81ea\u4e3b\u8f6f\u4ef6\u5f00\u53d1\u8fc8\u8fdb\u3002<\/p>\n","protected":false},"excerpt":{"rendered":" \u7f16\u8005\u6309\uff1a\u5f53\u524d\uff0c\u5927\u8bed\u8a00\u6a21\u578b\u5728\u4ee3\u7801\u751f\u6210\u9886\u57df\u5df2\u5c55\u73b0\u51fa\u60ca\u4eba\u7684\u80fd\u529b\uff0c\u4f46\u80fd\u5426\u80dc\u4efb\u771f\u5b9e\u8f6f\u4ef6\u5f00\u53d1\u4e2d\u7684\u201c\u65b0\u589e\u529f\u80fd\u5b9e\u73b0\u201d\u4efb\u52a1\uff0c\u4ecd\u662f\u4e00\u4e2a\u5173\u952e\u672a\u89e3\u7684\u95ee\u9898\u3002\u5bf9\u6b64\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u4e0e\u5317\u4eac\u5927\u5b66\u8054\u5408\u53d1\u5e03\u4e86\u9996\u4e2a\u4e13\u6ce8\u4e8e\u4ed3\u5e93\u7ea7\u65b0\u529f\u80fd\u5b9e\u73b0\u7684\u57fa\u51c6\u6d4b\u8bd5 FEA-Bench\uff0c\u586b\u8865\u4e86\u8bc4\u4f30\u4f53\u7cfb\u4e2d\u7684\u91cd\u8981\u7a7a\u767d\u3002\u8be5\u6d4b\u8bd5\u96c6\u6784\u5efa\u4e8e\u771f\u5b9e\u5f00\u6e90\u9879\u76ee\u7684 pull request \uff08\u5408\u5e76\u8bf7\u6c42\uff09\uff0c\u8986\u76d61400\u591a\u4e2a\u9ad8\u8d28\u91cf\u4efb\u52a1\uff0c\u7cfb\u7edf\u8bc4\u4f30\u4e86\u4e3b\u6d41\u5927\u6a21\u578b\u5728\u590d\u6742\u5de5\u7a0b\u4efb\u52a1\u4e2d\u7684\u8868\u73b0\u3002FEA-Bench \u4e0d\u4ec5\u4e3a\u63a8\u52a8\u4ee3\u7801\u751f\u6210\u7cfb\u7edf\u8fc8\u5411\u771f\u5b9e\u4e16\u754c\u7684\u65b0\u529f\u80fd\u5b9e\u73b0\u4efb\u52a1\u63d0\u4f9b\u4e86\u575a\u5b9e\u652f\u6491\uff0c\u4e5f\u4e3a\u76f8\u5173\u7814\u7a76\u62d3\u5c55\u4e86\u66f4\u5e7f\u9614\u7684\u63a2\u7d22\u7a7a\u95f4\u3002 \u968f\u7740\u4eba\u5de5\u667a\u80fd\u7684\u5feb\u901f\u53d1\u5c55\uff0c\u5927\u8bed\u8a00\u6a21\u578b\u5728\u4ee3\u7801\u751f\u6210\u9886\u57df\u5c55\u73b0\u51fa\u8d8a\u6765\u8d8a\u591a\u7684\u53ef\u80fd\u6027\uff0c\u4ece\u51fd\u6570\u7ea7\u522b\u7684\u8865\u5168\u5230\u590d\u6742\u95ee\u9898\u7684\u4fee\u590d\uff0cAI \u5df2\u9010\u6e10\u6e17\u900f\u81f3\u5f00\u53d1\u8005\u7684\u5de5\u4f5c\u6d41\u7a0b\u3002\u7136\u800c\uff0c\u5728\u771f\u5b9e\u7684\u8f6f\u4ef6\u5de5\u7a0b\u573a\u666f\u4e2d\uff0c\u5927\u8bed\u8a00\u6a21\u578b\u662f\u5426\u5177\u5907\u5b9e\u73b0\u65b0\u529f\u80fd\u7684\u80fd\u529b\uff0c\u4ecd\u662f\u4e00\u4e2a\u4e9f\u5f85\u9a8c\u8bc1\u7684\u91cd\u8981\u8bfe\u9898\u3002\u4e0e\u6b64\u540c\u65f6\uff0c\u73b0\u6709\u7684\u57fa\u51c6\u6d4b\u8bd5\u591a\u805a\u7126\u4e8e\u5927\u6a21\u578b\u7684\u72ec\u7acb\u7f16\u7a0b\u95ee\u9898\u6216\u4fee\u590d\u4efb\u52a1\uff0c\u7f3a\u4e4f\u5bf9\u201c\u65b0\u529f\u80fd\u5b9e\u73b0\u201d\u8fd9\u4e00\u6838\u5fc3\u80fd\u529b\u7684\u7cfb\u7edf\u6027\u8bc4\u4f30\u3002 \u5bf9\u6b64\uff0c\u6765\u81ea\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u4e0e\u5317\u4eac\u5927\u5b66\u7684\u7814\u7a76\u56e2\u961f\u8054\u5408\u63a8\u51fa\u4e86\u9996\u4e2a\u9762\u5411\u4ed3\u5e93\u7ea7\u65b0\u529f\u80fd\u5b9e\u73b0\u7684\u57fa\u51c6\u6d4b\u8bd5 FEA-Bench\uff0c\u586b\u8865\u4e86\u5f53\u524d\u4ee3\u7801\u751f\u6210\u8bc4\u4f30\u4f53\u7cfb\u4e2d\u7684\u5173\u952e\u4e00\u73af\u3002\u8be5\u57fa\u51c6\u6d4b\u8bd5\u901a\u8fc71401\u4e2a\u9ad8\u8d28\u91cf\u4efb\u52a1\u5b9e\u4f8b\uff0c\u5168\u9762\u8bc4\u4f30\u4e86\u5927\u8bed\u8a00\u6a21\u578b\u5728\u65b0\u589e\u529f\u80fd\u65f6\u6240\u9700\u7684\u4ee3\u7801\u751f\u6210\u4e0e\u7f16\u8f91\u80fd\u529b\uff0c\u63ed\u793a\u4e86\u5f53\u524d\u6a21\u578b\u5728\u590d\u6742\u8f6f\u4ef6\u5de5\u7a0b\u4efb\u52a1\u4e2d\u7684\u5c40\u9650\u6027\uff0c\u4e3a\u672a\u6765\u7684\u7814\u7a76\u63d0\u4f9b\u4e86\u91cd\u8981\u65b9\u5411\u3002 \u5c3d\u7ba1\u5f53\u524d\u7684\u5927\u8bed\u8a00\u6a21\u578b\u5982 GPT-4\u3001DeepSeek-Coder \u548c Qwen2.5-Coder \u7b49\u5728\u4ee3\u7801\u8865\u5168\u3001\u51fd\u6570\u7ea7\u751f\u6210\u4e43\u81f3\u4fee\u590d GitHub \u8bae\u9898\uff08issue\uff09\u7b49\u4efb\u52a1\u4e2d\u8868\u73b0\u51fa\u8272\uff0c\u4f46\u5728\u5b9e\u9645\u7684\u8f6f\u4ef6\u5f00\u53d1\u4e2d\uff0c\u4e3a\u4ee3\u7801\u4ed3\u5e93\u589e\u6dfb\u65b0\u529f\u80fd\u662f\u4e00\u4e2a\u66f4\u91cd\u8981\u4e14\u96be\u5ea6\u66f4\u5927\u7684\u9700\u6c42\u3002\u5f00\u53d1\u8005\u5f80\u5f80\u9700\u8981\u5728\u5df2\u6709\u4ee3\u7801\u5e93\u4e2d\u6dfb\u52a0\u65b0\u7ec4\u4ef6\uff08\u5982\u51fd\u6570\u3001\u7c7b\uff09\uff0c\u5e76\u76f8\u5e94\u4fee\u6539\u5176\u4ed6\u76f8\u5173\u4ee3\u7801\u4ee5\u786e\u4fdd\u6574\u4f53\u7684\u4e00\u81f4\u6027\u3002\u8fd9\u79cd\u4ed3\u5e93\u7ea7\u589e\u91cf\u5f00\u53d1\u80fd\u529b\u662f\u8861\u91cf\u5927\u8bed\u8a00\u6a21\u578b\u80fd\u5426\u80dc\u4efb\u771f\u5b9e\u8f6f\u4ef6\u5de5\u7a0b\u4efb\u52a1\u7684\u6838\u5fc3\u6307\u6807\u3002\u800c\u4e14\uff0c\u76ee\u524d\u5c1a\u65e0\u4e13\u95e8\u9488\u5bf9\u8be5\u80fd\u529b\u8bbe\u8ba1\u7684\u8bc4\u4f30\u57fa\u51c6\u3002 FEA-Bench \u7684\u51fa\u73b0\u586b\u8865\u4e86\u8fd9\u4e00\u7a7a\u767d\u3002\u8be5\u6570\u636e\u96c6\u57fa\u4e8e83\u4e2a\u5f00\u6e90 Python \u9879\u76ee\u7684\u771f\u5b9e pull request \uff08\u5408\u5e76\u8bf7\u6c42\uff09\u6570\u636e\u6784\u5efa\u800c\u6210\uff0c\u5305\u542b1401\u4e2a\u65b0\u589e\u529f\u80fd\u7684\u4efb\u52a1\u5b9e\u4f8b\uff0c\u5e76\u914d\u5957\u5b8c\u6574\u7684\u5355\u5143\u6d4b\u8bd5\u9a8c\u8bc1\u673a\u5236\uff0c\u4ece\u800c\u4fdd\u8bc1\u751f\u6210\u7ed3\u679c\u7684\u53ef\u6267\u884c\u6027\u4e0e\u51c6\u786e\u6027\u3002\u6bcf\u4e2a\u4efb\u52a1\u5b9e\u4f8b\u5747\u8981\u6c42\u6a21\u578b\u4e0d\u4ec5\u8981\u751f\u6210\u65b0\u7684\u4ee3\u7801\u7ec4\u4ef6\uff0c\u8fd8\u9700\u7406\u89e3\u6574\u4e2a\u4ed3\u5e93\u7ed3\u6784\u5e76\u5728\u5fc5\u8981\u65f6\u8fdb\u884c\u8054\u52a8\u4fee\u6539\uff0c\u4ece\u800c\u5b9e\u73b0\u7aef\u5230\u7aef\u7684\u529f\u80fd\u5b9e\u73b0\u3002 \u4e0d\u540c\u4e8e\u4ee5\u5f80\u4e3b\u8981\u5173\u6ce8 bug \u4fee\u590d\uff08\u5982 SWE-bench\uff09\u6216\u72ec\u7acb\u7f16\u7a0b\uff08\u5982 HumanEval\uff09\u4efb\u52a1\u7684\u57fa\u51c6\u6d4b\u8bd5\uff0cFEA-Bench \u5c06\u76ee\u6807\u805a\u7126\u4e8e\u65b0\u529f\u80fd\u7684\u5b9e\u73b0\u3002\u5176\u4efb\u52a1\u5b9e\u4f8b\u5168\u90e8\u6765\u81ea\u771f\u5b9e\u7684 GitHub pull request\uff0c\u6db5\u76d6\u591a\u4e2a\u884c\u4e1a\u573a\u666f\uff0c\u5177\u6709\u8f83\u9ad8\u7684\u4ee3\u8868\u6027\u548c\u5b9e\u7528\u4ef7\u503c\u3002\u8fd9\u6837\u7684\u8bbe\u8ba1\u4e0d\u4ec5\u4f7f\u6570\u636e\u96c6\u66f4\u8d34\u8fd1\u771f\u5b9e\u7684\u8f6f\u4ef6\u5f00\u53d1\u60c5\u5883\uff0c\u4e5f\u4e3a\u5927\u6a21\u578b\u5728\u5de5\u4e1a\u573a\u666f\u4e2d\u7684\u843d\u5730\u63d0\u4f9b\u4e86\u6307\u5bfc\u610f\u4e49\u3002 \u4e3a\u4e86\u786e\u4fdd\u4efb\u52a1\u8bbe\u7f6e\u7684\u771f\u5b9e\u6027\u4e0e\u6311\u6218\u6027\uff0cFEA-Bench \u6784\u5efa\u4e86\u4e00\u5957\u591a\u7ef4\u5ea6\u7684\u96be\u5ea6\u63a7\u5236\u4e0e\u8d28\u91cf\u4fdd\u969c\u673a\u5236\u3002\u5728\u6570\u636e\u7b5b\u9009\u9636\u6bb5\uff0c\u7cfb\u7edf\u4f1a\u7ed3\u5408\u89c4\u5219\u8fc7\u6ee4\u4e0e\u610f\u56fe\u8bc6\u522b\u6280\u672f\uff0c\u7cbe\u51c6\u7504\u522b\u51fa\u786e\u5b9e\u6d89\u53ca\u65b0\u529f\u80fd\u5f00\u53d1\u7684 pull request\u3002\u6bcf\u4e2a\u4efb\u52a1\u5b9e\u4f8b\u5747\u914d\u5907\u8be6\u5c3d\u7684\u529f\u80fd\u8bf7\u6c42\u63cf\u8ff0\u3001\u65b0\u589e\u7ec4\u4ef6\u5b9a\u4e49\uff08\u5305\u62ec\u51fd\u6570\u7b7e\u540d\u3001\u7c7b\u7ed3\u6784\u7b49\uff09\u3001\u73af\u5883\u914d\u7f6e\u4fe1\u606f\u53ca\u5355\u5143\u6d4b\u8bd5\u7528\u4f8b\uff0c\u5e76\u6807\u6ce8\u9ec4\u91d1\u8865\u4e01\uff08gold patch\uff09\u4f5c\u4e3a\u53c2\u8003\u5b9e\u73b0\u3002\u6240\u6709\u5b9e\u4f8b\u5728\u53d1\u5e03\u524d\u5747\u901a\u8fc7\u81ea\u52a8\u5355\u5143\u6d4b\u8bd5\u9a8c\u8bc1\uff0c\u786e\u4fdd\u6570\u636e\u96c6\u5177\u6709\u826f\u597d\u7684\u53ef\u9a8c\u8bc1\u6027\u548c\u53ef\u590d\u73b0\u6027\uff0c\u4e3a\u8bc4\u4f30\u6a21\u578b\u6027\u80fd\u63d0\u4f9b\u4e86\u7a33\u5b9a\u53ef\u9760\u7684\u57fa\u7840\u3002 \u6b64\u5916\uff0cFEA-Bench \u8fd8\u63d0\u4f9b\u4e86\u4e00\u4e2a\u8f7b\u91cf\u7248\u672c FEA-Bench Lite\uff0c\u7528\u4e8e\u9002\u914d\u4e0d\u540c\u5b9e\u9a8c\u6761\u4ef6\u4e0b\u7684\u8ba1\u7b97\u8d44\u6e90\u9700\u6c42\u3002\u8be5\u5b50\u96c6\u5728\u4fdd\u7559\u4ee3\u8868\u6027\u7684\u524d\u63d0\u4e0b\uff0c\u7b5b\u9009\u51fa\u4e86\u590d\u6742\u5ea6\u76f8\u5bf9\u8f83\u4f4e\u7684\u4efb\u52a1\u5b9e\u4f8b\uff0c\u65b9\u4fbf\u79d1\u7814\u4eba\u5458\u5728\u8d44\u6e90\u53d7\u9650\u7684\u73af\u5883\u4e2d\u8fdb\u884c\u6a21\u578b\u7684\u5feb\u901f\u8fed\u4ee3\u4e0e\u8bc4\u4f30\u3002 \u7814\u7a76\u56e2\u961f\u4f7f\u7528 FEA-Bench \u7cfb\u7edf\u5730\u8bc4\u4f30\u4e86 GPT-4\u3001DeepSeek-R1\u3001Qwen2.5-Coder \u7b49\u76ee\u524d\u4e3b\u6d41\u7684\u5927\u8bed\u8a00\u6a21\u578b\u3002\u7ed3\u679c\u663e\u793a\uff0c\u5728\u4e0d\u4f7f\u7528\u4ee3\u7406\uff08agent\uff09\u6846\u67b6\u7684\u60c5\u51b5\u4e0b\uff0c\u5f53\u524d\u6700\u4f18\u6a21\u578b DeepSeek-R1 \u5728 Oracle \u8bbe\u7f6e\u4e2d\uff08\u7ed9\u5b9a\u9700\u8981\u7f16\u8f91\u7684\u4ee3\u7801\u6587\u4ef6\u4f5c\u4e3a\u4e0a\u4e0b\u6587\u8f93\u5165\uff09\u7684\u89e3\u51b3\u7387\u4ec5\u4e3a9.92%\u3002\u76f8\u6bd4\u4e4b\u4e0b\uff0cSWE-bench \u4e2d\u6a21\u578b\u7684 bug \u4fee\u590d\u6210\u529f\u7387\u666e\u904d\u572830%\u4ee5\u4e0a\u3002\u8fd9\u4e00\u5dee\u8ddd\u5145\u5206\u8bf4\u660e\uff0c\u201c\u65b0\u589e\u529f\u80fd\u201d\u4efb\u52a1\u5728\u903b\u8f91\u590d\u6742\u5ea6\u3001\u8de8\u6587\u4ef6\u4e00\u81f4\u6027\u7ef4\u62a4\u4e0e\u903b\u8f91\u6574\u5408\u7b49\u65b9\u9762\uff0c\u90fd\u5bf9\u5927\u8bed\u8a00\u6a21\u578b\u63d0\u51fa\u4e86\u66f4\u9ad8\u7684\u8981\u6c42\u3002 FEA-Bench […]<\/p>\n","protected":false},"author":34512,"featured_media":1142573,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-content-parent":1012650,"msr_hide_image_in_river":null,"footnotes":""},"research-area":[13556],"msr-locale":[268881],"msr-post-option":[],"class_list":["post-1142564","msr-blog-post","type-msr-blog-post","status-publish","has-post-thumbnail","hentry","msr-research-area-artificial-intelligence","msr-locale-zh_cn"],"msr_assoc_parent":{"id":1012650,"type":"lab"},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1142564","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-blog-post"}],"author":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/users\/34512"}],"version-history":[{"count":4,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1142564\/revisions"}],"predecessor-version":[{"id":1142574,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1142564\/revisions\/1142574"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/1142573"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1142564"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1142564"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1142564"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1142564"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}\n
FEA-Bench\u63a8\u52a8\u5927\u8bed\u8a00\u6a21\u578b\u8d70\u5411\u771f\u6b63\u7684\u8f6f\u4ef6\u5de5\u7a0b\u81ea\u52a8\u5316<\/h2>\n\n\n\n





\u5f00\u542f\u65b0\u9636\u6bb5\uff1a\u52a9\u529bAI\u771f\u6b63\u7406\u89e3\u8f6f\u4ef6\u5de5\u7a0b<\/h2>\n\n\n\n