{"id":1501,"date":"2025-06-18T21:20:00","date_gmt":"2025-06-18T13:20:00","guid":{"rendered":"http:\/\/www.applicationsofllm.com\/?p=1501"},"modified":"2025-09-17T21:21:40","modified_gmt":"2025-09-17T13:21:40","slug":"%e6%89%93%e7%a0%b4%e5%a4%a7%e6%a8%a1%e5%9e%8b%e7%bc%96%e7%a8%8b%e3%80%8c%e6%95%b0%e6%8d%ae%e6%b1%a1%e6%9f%93%e3%80%8d%e4%b8%8e%e3%80%8c%e8%83%bd%e5%8a%9b%e8%99%9a%e8%83%96%e3%80%8d%e5%9b%b0%e5%a2%83","status":"publish","type":"post","link":"http:\/\/www.applicationsofllm.com\/index.php\/2025\/06\/18\/%e6%89%93%e7%a0%b4%e5%a4%a7%e6%a8%a1%e5%9e%8b%e7%bc%96%e7%a8%8b%e3%80%8c%e6%95%b0%e6%8d%ae%e6%b1%a1%e6%9f%93%e3%80%8d%e4%b8%8e%e3%80%8c%e8%83%bd%e5%8a%9b%e8%99%9a%e8%83%96%e3%80%8d%e5%9b%b0%e5%a2%83\/","title":{"rendered":"\u6253\u7834\u5927\u6a21\u578b\u7f16\u7a0b\u300c\u6570\u636e\u6c61\u67d3\u300d\u4e0e\u300c\u80fd\u529b\u865a\u80d6\u300d\u56f0\u5883\uff0cMeituan-M17\u56e2\u961f\u6784\u5efa\u65b0\u4e00\u4ee3AI\u7f16\u7a0b\u8bc4\u6d4b\u65b0\u6807\u51c6\u2014\u2014OIBench"},"content":{"rendered":"\n<p>\u5f53\u524d\uff0c\u5927\u8bed\u8a00\u6a21\u578b\uff08LLMs\uff09\u5728\u7f16\u7a0b\u9886\u57df\u7684\u80fd\u529b\u53d7\u5230\u5e7f\u6cdb\u5173\u6ce8\uff0c\u76f8\u5173\u8bba\u65ad\u5728\u5e02\u573a\u4e2d\u666e\u904d\u5b58\u5728\uff0c\u4f8b\u5982 DeepMind \u7684 AlphaCode \u66fe\u5ba3\u79f0\u8fbe\u5230\u4eba\u7c7b\u7ade\u6280\u7f16\u7a0b\u9009\u624b\u7684\u6c34\u5e73\uff1bOpenAI \u7684\u9876\u5c16\u6a21\u578b\u5c61\u5c61\u88ab\u62a5\u9053\u80fd\u901a\u8fc7\u8c37\u6b4c\u9ad8\u7ea7\u7f16\u7a0b\u9762\u8bd5\uff0c\u5e76\u5728 LeetCode \u6311\u6218\u4e2d\u8868\u73b0\u51fa\u8f83\u9ad8\u80fd\u529b\u3002<\/p>\n\n\n\n<p>\u7136\u800c\uff0c\u5c06\u8fd9\u4e9b\u80fd\u529b\u5ba3\u79f0\u4e0e\u5b9e\u9645\u8bc4\u6d4b\u7ed3\u679c\u8fdb\u884c\u5bf9\u6bd4\u65f6\uff0c<strong>\u5f53\u524d\u8bc4\u4f30\u4f53\u7cfb\u7684\u6df1\u5c42\u95ee\u9898\u4fbf\u968f\u4e4b\u663e\u73b0\uff1a<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u6bd4\u5982\u5c3d\u7ba1 GPT-4o \u6a21\u578b\u88ab\u51a0\u4ee5 \u201c\u7ade\u8d5b\u7ea7\u201d \u5934\u8854\uff0c\u751a\u81f3\u6709\u58f0\u97f3\u79f0\u5176\u7b97\u6cd5\u6c34\u5e73\u63a5\u8fd1 ACM \u533a\u57df\u8d5b\u91d1\u724c\u9009\u624b\uff0c\u4f46\u5b9e\u9645\u5728\u9762\u5bf9\u672a\u7ecf\u5927\u91cf\u516c\u5f00\u6570\u636e\u8bad\u7ec3\u7684\u3001\u66f4\u9ad8\u96be\u5ea6\u7684\u4fe1\u606f\u5b66\u5965\u8d5b\u7ea7\u522b\u95ee\u9898\u65f6\uff0c\u5176\u901a\u8fc7\u7387\u5374\u5f80\u5f80\u4f4e\u81f3\u4e2a\u4f4d\u6570\uff0c\u4e0e 985 \u7ea7\u522b\u9ad8\u6821 ACM \u6821\u961f\u6210\u5458\u7684\u5e73\u5747\u901a\u8fc7\u7387\u5b58\u5728\u663e\u8457\u5dee\u8ddd\u3002<\/li>\n\n\n\n<li>\u5f53\u90e8\u5206\u8bc4\u6d4b\u5ba3\u79f0 Claude 3.5 Sonnet \u53ef\u66ff\u4ee3\u4e2d\u7ea7\u5f00\u53d1\u4eba\u5458\u65f6\uff0c\u5b83\u5728\u52a8\u6001\u89c4\u5212\u7b49\u9ad8\u96be\u5ea6\u9898\u578b\u4e2d\u9519\u8bef\u7387\u5374\u9ad8\u8fbe 80% \u4ee5\u4e0a\uff0c\u4e14\u65e0\u6cd5\u72ec\u7acb\u5b8c\u6210\u9700\u6570\u5b66\u5efa\u6a21\u7684\u590d\u6742\u7ade\u8d5b\u9898\u3002<\/li>\n\n\n\n<li>\u8bf8\u5982\u6587\u5fc3\u4e00\u8a00\u3001\u901a\u4e49\u5343\u95ee\u7b49\u6a21\u578b\u5728 MBPP \u57fa\u7840\u9898\u5e93\u4e2d\u901a\u8fc7\u7387\u53ef\u8fbe 90% \u4ee5\u4e0a\uff0c\u4f46\u79fb\u690d\u81f3 Codeforces Div.2 C \u7ea7\u9898\u76ee\u65f6\uff0c\u901a\u8fc7\u7387\u5374\u4e0d\u8db3 15%\uff0c\u8fdc\u4f4e\u4e8e\u4eba\u7c7b\u9009\u624b\u5e73\u5747 70% \u7684\u6c34\u5e73\u3002<\/li>\n<\/ul>\n\n\n\n<p>\u8fd9\u4e9b\u9c9c\u660e\u7684\u5bf9\u6bd4\uff0c\u5171\u540c\u6307\u5411\u4e00\u4e2a\u6838\u5fc3<strong>\u95ee\u9898<\/strong>\uff1a\u5f53\u524d\u5bf9 LLM \u7f16\u7a0b\u80fd\u529b\u7684\u8bc4\u4f30\uff0c\u5f80\u5f80\u5b58\u5728 \u201c\u5ba3\u4f20\u4e0e\u73b0\u5b9e\u7684\u8ba4\u77e5\u9e3f\u6c9f\u201d\u3002\u8fd9\u79cd\u5dee\u5f02\u4e0d\u4ec5\u6e90\u4e8e\u6a21\u578b\u80fd\u529b\u8fb9\u754c\u7684\u590d\u6742\u6027\uff0c\u4e5f\u66b4\u9732\u51fa\u73b0\u6709\u8bc4\u4f30\u4f53\u7cfb\u7684\u8bf8\u591a\u5c40\u9650\u6027\u3002\u5177\u4f53\u8868\u73b0\u4e3a\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u8bc4\u6d4b\u96c6 \u201c\u9971\u548c\u201d \u4e0e\u533a\u5206\u5ea6\u4e0d\u8db3\uff1a\u00a0<\/strong>\u4f20\u7edf\u8bc4\u6d4b\u96c6\uff08\u5982 HumanEval\u3001MBPP\uff09\u7531\u4e8e\u6a21\u578b\u80fd\u529b\u7684\u5feb\u901f\u63d0\u5347\uff0c\u901a\u8fc7\u7387\u666e\u904d\u8d85\u8fc7 90%\uff0c\u5df2\u65e0\u6cd5\u6709\u6548\u533a\u5206\u6700\u5148\u8fdb\u6a21\u578b\u7684\u7ec6\u5fae\u4f18\u52a3\u3002<\/li>\n\n\n\n<li><strong>\u6570\u636e \u201c\u6cc4\u6f0f\u201d \u98ce\u9669\uff1a<\/strong>\u00a0\u5c3d\u7ba1\u4e00\u4e9b\u65b0\u8bc4\u6d4b\u96c6\uff08\u5982 Codeforces\u3001USACO\u3001LeetCode\uff09\u5f15\u5165\u4e86\u9ad8\u96be\u5ea6\u9898\u76ee\uff0c\u4f46\u7531\u4e8e\u5927\u6a21\u578b\u9884\u8bad\u7ec3\u6570\u636e\u5305\u542b\u5927\u91cf\u4e92\u8054\u7f51\u516c\u5f00\u5185\u5bb9\uff0c\u8fd9\u4e9b\u9898\u76ee\u53ef\u80fd\u5df2\u88ab\u6a21\u578b \u201c\u89c1\u8fc7\u201d\uff0c\u5bfc\u81f4\u8bc4\u6d4b\u7ed3\u679c\u865a\u9ad8\uff0c\u65e0\u6cd5\u771f\u5b9e\u53cd\u6620\u5176\u63a8\u7406\u80fd\u529b\u3002<\/li>\n\n\n\n<li><strong>\u4eba\u673a\u5bf9\u6bd4\u7684\u5c40\u9650\u6027\uff1a<\/strong>\u00a0\u73b0\u6709\u57fa\u4e8e Elo \u8bc4\u5206\u4f53\u7cfb\u7684\u6a21\u578b\u4e0e\u771f\u4eba\u9009\u624b\u5bf9\u6bd4\u65b9\u6cd5\uff0c\u5b58\u5728\u5468\u671f\u957f\u3001\u9009\u624b\u6c34\u5e73\u6ce2\u52a8\u5927\u3001\u590d\u73b0\u6027\u5dee\u7b49\u95ee\u9898\uff0c\u96be\u4ee5\u63d0\u4f9b\u7cbe\u786e\u4e14\u53ef\u9760\u7684\u8bc4\u4f30\u3002<\/li>\n\n\n\n<li><strong>\u6548\u7387\u6307\u6807\u7684\u7c97\u7565\u6027\uff1a\u00a0<\/strong>\u90e8\u5206\u8bc4\u6d4b\u867d\u5f15\u5165\u8fd0\u884c\u65f6\u95f4\u3001\u5185\u5b58\u7b49\u6548\u7387\u6307\u6807\uff0c\u4f46\u901a\u5e38\u4ec5\u4e3a\u7c97\u7565\u7684\u5e73\u5747\u5206\uff0c\u65e0\u6cd5\u7ec6\u81f4\u53cd\u6620\u6a21\u578b\u5728\u4e0d\u540c\u7c7b\u578b\u9898\u76ee\u4e0a\u7684\u6027\u80fd\u5dee\u5f02\u3002<\/li>\n<\/ul>\n\n\n\n<p>\u4e3a\u4e86\u89e3\u51b3\u4e0a\u8ff0\u8fd9\u4e9b\u8bc4\u4f30\u56f0\u5883\u3001\u8bc4\u6d4b\u51fa\u5168\u7403\u9876\u5c16\u6a21\u578b\u771f\u5b9e\u7684\u7f16\u7a0b\u80fd\u529b\uff0c&nbsp;<strong>Meituan-M17<\/strong><strong>\u56e2\u961f \u63a8\u51fa\u4e86\u66f4\u771f\u5b9e\u3001\u66f4\u5177\u533a\u5206\u5ea6\u7684\u8bc4\u4f30\u57fa\u51c6 OIBench \u6570\u636e\u96c6\uff0c\u5e76\u6258\u7ba1\u4e8e AGI-Eval \u8bc4\u6d4b\u793e\u533a\u3002<\/strong>\u57fa\u4e8e\u6b64\u6570\u636e\u96c6\uff0c\u6211\u4eec\u5bf9\u5168\u7403 18 \u4e2a\u4e3b\u6d41\u5927\u6a21\u578b\u7684\u7b97\u6cd5\u7f16\u7a0b\u80fd\u529b\u8fdb\u884c\u4e86\u7cfb\u7edf\u8bc4\u6d4b\u5e76\u91cf\u5316\u5f97\u5206\uff0c\u8be6\u7ec6\u8bc4\u5206\u699c\u5355\u5982\u4e0b\u6240\u793a\uff0c\u53ef\u4ee5\u770b\u5230\u5168\u7403\u9876\u5c16\u5927\u6a21\u578b\u8ddd\u79bb\u4ee5\u5f80\u6240\u5ba3\u79f0\u7684\u7f16\u7a0b\u80fd\u529b\u8fd8\u5b58\u5728\u5f88\u5927\u5dee\u8ddd\uff0c\u54ea\u6015\u662f\u6700\u9ad8\u5206\u7684 o4-mini-high \u4e5f\u4ec5\u4ec5\u53ea\u6709 36.35 \u5206\uff0c\u8ddd\u79bb\u4eba\u7c7b\u7ade\u8d5b\u9009\u624b\u7684\u6c34\u5e73\u8fd8\u76f8\u5dee\u751a\u8fdc\uff0c\u751a\u81f3\u5f88\u591a\u6a21\u578b\u53ea\u6709\u4e2a\u4f4d\u6570\u7684\u5f97\u5206\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"427\" height=\"1024\" src=\"http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-601-427x1024.jpg\" alt=\"\" class=\"wp-image-1506\" srcset=\"http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-601-427x1024.jpg 427w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-601-125x300.jpg 125w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-601-768x1842.jpg 768w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-601-640x1536.jpg 640w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-601.jpg 776w\" sizes=\"auto, (max-width: 427px) 100vw, 427px\" \/><\/figure>\n\n\n\n<p>OIBench \u7684\u8bc4\u6d4b\u699c\u5355\u672a\u6765\u5c06\u7531 AGI-Eval \u8bc4\u6d4b\u793e\u533a\u957f\u671f\u7ef4\u62a4\u66f4\u65b0\uff0c\u6b22\u8fce\u6301\u7eed\u5173\u6ce8\u3002\u699c\u5355\u5730\u5740\u5982\u4e0b<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u7f51\u9875\u7aef\u5730\u5740\uff1ahttps:\/\/agi-eval.cn\/evaluation\/detail?id=60<\/li>\n\n\n\n<li>\u5fae\u4fe1\u5c0f\u7a0b\u5e8f\uff1a<\/li>\n\n\n\n<li>\u8bba\u6587\u5730\u5740\uff1ahttps:\/\/arxiv.org\/abs\/2506.10481<\/li>\n<\/ul>\n\n\n\n<p>\u672c\u6587\u6570\u636e\u5747\u5f15\u7528\u81ea OIBench v1.0 \u8bba\u6587\uff08arxiv:2506.10481v3\uff09\uff0c\u53d1\u5e03\u65e5\u671f 2025 \u5e74 6 \u6708 13 \u65e5<\/p>\n\n\n\n<p>\u63a5\u4e0b\u6765\u4e3a\u5927\u5bb6\u8be6\u7ec6\u4ecb\u7ecd OIBench \u6570\u636e\u96c6\u662f<strong>\u5982\u4f55\u6784\u5efa\u4ee5\u53ca\u5982\u4f55\u5bf9\u5927\u6a21\u578b\u8fdb\u884c\u8bc4\u6d4b\u7684<\/strong>\u3002<\/p>\n\n\n\n<p><strong>1. OIBench \u7684\u6784\u5efa\u4e0e\u521b\u65b0<\/strong><\/p>\n\n\n\n<p>OIBench \u662f\u4e00\u4e2a\u9ad8\u8d28\u91cf\u3001\u79c1\u6709\u4e14\u6781\u5177\u6311\u6218\u6027\u7684\u4fe1\u606f\u5b66\u5965\u8d5b\u7ea7\u522b\u7b97\u6cd5\u9898\u5e93\uff0c\u65e8\u5728\u63d0\u4f9b\u4e00\u4e2a\u66f4\u771f\u5b9e\u3001\u66f4\u5177\u533a\u5206\u5ea6\u7684\u8bc4\u4f30\u57fa\u51c6\u3002\u8be5\u6570\u636e\u96c6\u7684\u7b97\u6cd5\u9898\u4e3b\u8981\u6765\u6e90\u4e8e\u4e2d\u56fd ACM-ICPC \u961f\u4f0d\u548c\u4fe1\u606f\u5b66\u5965\u8d5b\u7684\u9ad8\u6821\u6559\u7ec3\u56e2\u961f\u7cbe\u5fc3\u7f16\u7e82\uff0c\u4ed6\u4eec\u62e5\u6709\u4e30\u5bcc\u7684\u9ad8\u96be\u5ea6\u7b97\u6cd5\u9898\u8bbe\u8ba1\u7ecf\u9a8c\u548c\u72ec\u5230\u89c1\u89e3\u3002<\/p>\n\n\n\n<p>\u4e3a\u4e86\u786e\u4fdd OIBench \u9898\u76ee\u7684\u9ad8\u8d28\u91cf\u548c\u9ad8\u6311\u6218\u6027\uff0c\u6211\u4eec\u5236\u5b9a\u4e86\u4e09\u6761\u4e25\u683c\u7684\u51c6\u5165\u6807\u51c6\uff0cOIBench \u5177\u5907\u4ee5\u4e0b\u5173\u952e\u7279\u6027\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u539f\u521b\u6027\u4e0e\u79c1\u6709\u6027\uff1a<\/strong>\u00a0OIBench \u5305\u542b 250 \u9053\u5019\u9009\u9898\u76ee\uff0c\u7ecf\u96be\u5ea6\u9a8c\u8bc1\u540e\u4fdd\u7559 212 \u9053\u9ad8\u96be\u5ea6\u3001\u9632\u6cc4\u6f0f\u7684\u4fe1\u606f\u5b66\u5965\u8d5b\u9898\u76ee\uff08IOI Level\uff09\u3002\u6240\u6709\u9898\u76ee\u5728\u53d1\u5e03\u524d\u90fd\u7ecf\u8fc7\u4e25\u683c\u68c0\u7d22\uff0c\u786e\u4fdd\u672a\u5728\u4efb\u4f55\u516c\u5f00\u5e73\u53f0\u51fa\u73b0\uff0c\u6700\u5927\u7a0b\u5ea6\u907f\u514d\u6570\u636e\u6c61\u67d3\u98ce\u9669\u3002<\/li>\n\n\n\n<li><strong>\u96be\u5ea6\u5206\u7ea7\u4e0e\u628a\u63a7\uff1a<\/strong>\u00a0\u6bcf\u9053\u9898\u76ee\u90fd\u53c2\u7167\u4fe1\u606f\u5b66\u7ade\u8d5b\u548c Codeforces \u96be\u5ea6\u8bc4\u7ea7\u8fdb\u884c\u6807\u6ce8\u3002\u540c\u65f6\uff0c\u4e3a\u907f\u514d\u4e3b\u89c2\u504f\u5dee\uff0c\u6211\u4eec\u5f15\u5165\u4e86\u81ea\u52a8\u5316\u9a8c\u8bc1\u673a\u5236 \u2014\u2014 \u53ea\u6709\u5f53 GPT-4o\u3001Qwen2.5-Coder-32B\u3001Doubao-32k-pro\u3001Llama3.1-405B \u8fd9\u51e0\u4e2a\u6807\u6746\u5927\u6a21\u578b\u4e2d \u201c\u6700\u591a\u53ea\u6709\u4e00\u4e2a\u6a21\u578b\u80fd\u89e3\u51fa\u201d \u65f6\uff0c\u8be5\u9898\u624d\u4f1a\u88ab\u6536\u5f55\uff0c\u4ece\u800c\u786e\u4fdd\u4e86\u9898\u76ee\u7684 \u201c\u786c\u6838\u201d \u96be\u5ea6\u3002<\/li>\n\n\n\n<li><strong>\u9ad8\u6807\u51c6\u6d4b\u8bd5\u7528\u4f8b\u4e0e\u6807\u51c6\u89e3\u7b54\uff1a<\/strong>\u00a0\u6bcf\u9053\u9898\u90fd\u914d\u5907\u8986\u76d6\u5927\u6570\u636e\u91cf\u3001\u8fb9\u754c\u60c5\u51b5\u7b49\u591a\u6837\u7684\u6d4b\u8bd5\u7528\u4f8b\uff0c\u529b\u6c42\u66b4\u9732\u4ee3\u7801\u5728\u65f6\u95f4\u548c\u7a7a\u95f4\u4e0a\u7684\u6f5c\u5728\u74f6\u9888\u3002\u540c\u65f6\uff0c\u6bcf\u9053\u9898\u90fd\u5fc5\u987b\u914d\u5907\u7ecf\u8fc7\u6240\u6709\u6d4b\u8bd5\u7528\u4f8b\u4e25\u683c\u9a8c\u8bc1\u7684 C++ \u6807\u51c6\u89e3\u7b54\uff0c\u4ee5\u786e\u4fdd\u9898\u76ee\u672c\u8eab\u7684\u51c6\u786e\u6027\u53ca\u8bc4\u6d4b\u7684\u516c\u6b63\u6027\u3002<\/li>\n\n\n\n<li><strong>\u4e2d\u82f1\u6587\u53cc\u8bed\u652f\u6301\uff1a<\/strong>\u00a0\u6570\u636e\u96c6\u63d0\u4f9b\u4e2d\u82f1\u6587\u53cc\u8bed\u7248\u672c\uff0c\u65b9\u4fbf\u5168\u7403\u5927\u6a21\u578b\u4ece\u4e1a\u8005\u4f7f\u7528\u3002<\/li>\n<\/ul>\n\n\n\n<p>\u6211\u4eec\u8fd8\u5728\u8bba\u6587\u4e2d\u5c55\u793a\u4e86 OIBench \u4e0e\u5176\u4ed6\u4e3b\u6d41\u8bc4\u6d4b\u96c6\u7684\u5bf9\u6bd4\uff08\u89c1\u4e0b\u8868\uff09\uff0c\u53ef\u4ee5\u770b\u5230 OIBench \u5728\u9898\u76ee\u96be\u5ea6\u548c\u6d4b\u8bd5\u7528\u4f8b\u89c4\u6a21\u4e0a\u90fd\u76f8\u5bf9\u66f4\u9ad8\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"249\" src=\"http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-602-1024x249.jpg\" alt=\"\" class=\"wp-image-1507\" srcset=\"http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-602-1024x249.jpg 1024w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-602-300x73.jpg 300w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-602-768x186.jpg 768w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-602.jpg 1108w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<p>OIBench \u5728\u9898\u76ee\u96be\u5ea6\u548c\u6d4b\u8bd5\u7528\u4f8b\u89c4\u6a21\u4e0a\u663e\u8457\u9886\u5148\u4e8e\u5176\u4ed6\u4e3b\u6d41\u8bc4\u6d4b\u96c6\u3002\u4f8b\u5982\uff0c\u5728\u5176\u4ed6\u699c\u5355\u4e0a\u8868\u73b0\u8f83\u597d\u7684 GPT-4o \u6a21\u578b\u5728 OIBench \u4e0a\u4ec5\u80fd\u7b54\u5bf9 2.6% \u7684\u9898\u76ee\uff0c\u540c\u65f6 OIBench \u7684\u6d4b\u8bd5\u7528\u4f8b\u6570\u91cf\u5927\u5e45\u8d85\u8fc7\u4e86\u5176\u4ed6\u7b97\u6cd5\u7ade\u8d5b\u57fa\u51c6\uff0c<strong>\u5bf9\u6807\u771f\u5b9e\u7684\u7ade\u8d5b\u73af\u5883<\/strong>\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"568\" height=\"360\" src=\"http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-234.png\" alt=\"\" class=\"wp-image-1502\" srcset=\"http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-234.png 568w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-234-300x190.png 300w\" sizes=\"auto, (max-width: 568px) 100vw, 568px\" \/><\/figure>\n\n\n\n<p><strong>\u5f3a\u6297\u6570\u636e\u6c61\u67d3\u80fd\u529b\uff1a&nbsp;<\/strong>\u5728\u8bc4\u6d4b\u96c6\u8bbe\u8ba1\u4e2d\uff0c\u201c\u540c\u6e90\u6c61\u67d3\u201d \u662f\u4e00\u4e2a\u91cd\u8981\u6311\u6218\u3002\u7531\u4e8e\u5927\u6a21\u578b\u7684\u9884\u8bad\u7ec3\u548c\u5fae\u8c03\u6570\u636e\u5f80\u5f80\u4f1a\u722c\u53d6\u5927\u91cf\u4e92\u8054\u7f51\u5185\u5bb9\uff0c\u5bb9\u6613\u51fa\u73b0\u6a21\u578b\u5728\u8bad\u7ec3\u9636\u6bb5\u5c31\u89c1\u8fc7\u7c7b\u4f3c\u9898\u76ee\u7684\u60c5\u51b5\uff0c\u4ece\u800c\u5bfc\u81f4\u8bc4\u6d4b\u5206\u6570\u865a\u9ad8\uff0c\u65e0\u6cd5\u771f\u5b9e\u53cd\u6620\u6a21\u578b\u5b9e\u9645\u80fd\u529b\u3002\u867d\u7136 OIBench \u5728\u6570\u636e\u6784\u9020\u65f6\u6781\u529b\u907f\u514d\u4f7f\u7528\u4e92\u8054\u7f51\u53ef\u516c\u5f00\u68c0\u7d22\u7684\u9898\u76ee\uff0c\u4f46\u4e00\u4e9b\u76f8\u8fd1\u7684\u9898\u76ee\u4ecd\u53ef\u80fd\u5728\u5927\u6a21\u578b\u7684\u9884\u8bad\u7ec3\u6216\u5fae\u8c03\u9636\u6bb5\u5e26\u6765\u6570\u636e\u6c61\u67d3\u3002\u4e3a\u6b64\uff0c\u6211\u4eec\u4e13\u95e8\u8bbe\u8ba1\u4e86\u5b9e\u9a8c\u6765\u9a8c\u8bc1 OIBench \u7684\u6297\u6c61\u67d3\u80fd\u529b\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u5177\u4f53\u505a\u6cd5\uff1a<\/strong>\u00a0\u6211\u4eec\u4ece OIBench \u4e2d\u62bd\u53d6\u90e8\u5206\u9898\u76ee\uff0c\u6a21\u62df\u5b83\u4eec\u5728\u6a21\u578b\u8bad\u7ec3\u6570\u636e\u4e2d \u201c\u6cc4\u6f0f\u201d \u7684\u573a\u666f\uff0c\u5e76\u4e0e\u5e38\u89c4\u8bad\u7ec3\u6570\u636e\u6df7\u5408\uff0c\u5bf9\u6bd4\u6a21\u578b\u5728 OIBench \u4e0a\u7684\u8868\u73b0\u63d0\u5347\u3002<\/li>\n\n\n\n<li><strong>\u5b9e\u9a8c\u8bc1\u660e\uff1a<\/strong>\u00a0\u5373\u4f7f\u6a21\u62df\u5c11\u91cf\u9898\u76ee \u201c\u6cc4\u6f0f\u201d \u5230\u6a21\u578b\u7684\u8bad\u7ec3\u6570\u636e\u4e2d\uff0cOIBench \u7684\u5f97\u5206\u63d0\u5347\u4e5f\u6781\u4e3a\u6709\u9650\uff0c\u98ce\u9669\u5206\u6570\u51e0\u4e4e\u4e3a\u96f6\uff0c\u8868\u660e\u5176\u5bf9\u6570\u636e\u6c61\u67d3\u5177\u6709\u5f88\u5f3a\u7684\u9c81\u68d2\u6027\u3002<\/li>\n<\/ul>\n\n\n\n<p><strong>2. OIBench \u8bc4\u6d4b\u7ed3\u679c\u4e0e\u53d1\u73b0<\/strong><\/p>\n\n\n\n<p><strong>\u53c2\u8bc4\u6a21\u578b\u4e0e\u8bc4\u6d4b\u65b9\u5f0f<\/strong><\/p>\n\n\n\n<p>OIBench \u5bf9 18 \u4e2a\u4e3b\u6d41\u5927\u6a21\u578b\uff08\u5305\u62ec 14 \u4e2a\u6307\u4ee4\u5fae\u8c03\u6a21\u578b\u548c 4 \u4e2a\u57fa\u7840\u6a21\u578b\uff09\u8fdb\u884c\u4e86 zero-shot \u8bc4\u6d4b\uff0c\u6db5\u76d6 C++\u3001Python\u3001Java\u3001JavaScript \u56db\u79cd\u8bed\u8a00\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"685\" src=\"http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-600-1024x685.jpg\" alt=\"\" class=\"wp-image-1505\" srcset=\"http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-600-1024x685.jpg 1024w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-600-300x201.jpg 300w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-600-768x514.jpg 768w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-600.jpg 1108w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<p><strong>\u4e3b\u699c\u5355\u7ed3\u679c<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u63a8\u7406\u6a21\u578b\u8868\u73b0\u7a81\u51fa\uff1a<\/strong>\u00a0\u63a8\u7406\u7c7b\u6a21\u578b\uff08\u5982 o4-mini-high\uff09\u5728 OIBench \u4e0a\u7684\u5e73\u5747\u5f97\u5206\u9ad8\u8fbe 21.4%\uff0c\u8fdc\u9ad8\u4e8e\u666e\u901a\u6307\u4ee4\u5fae\u8c03\u6a21\u578b\uff08\u7ea6 3.6%\uff09\u3002\u8fd9\u8868\u660e OIBench \u80fd\u6709\u6548\u533a\u5206\u6a21\u578b\u7684\u63a8\u7406\u548c\u94fe\u5f0f\u601d\u8003\u80fd\u529b\uff0c\u4e14 o4-mini-high \u5728\u6240\u6709\u8bed\u8a00\u548c\u4efb\u52a1\u4e0a\u8868\u73b0\u6700\u4f18\u3002<\/li>\n\n\n\n<li><strong>\u95ed\u6e90\u6a21\u578b\u4f18\u52bf\u660e\u663e\uff1a\u00a0<\/strong>\u95ed\u6e90\u6a21\u578b\u5e73\u5747\u5f97\u5206 14.5%\uff0c\u663e\u8457\u9ad8\u4e8e\u5f00\u6e90\u6a21\u578b\uff086.3%\uff09\uff0c\u8fd9\u4e3b\u8981\u5f97\u76ca\u4e8e\u95ed\u6e90\u6a21\u578b\u5728\u7b97\u529b\u548c\u6570\u636e\u8d28\u91cf\u4e0a\u7684\u4f18\u52bf\u3002<\/li>\n\n\n\n<li><strong>\u57fa\u7840\u6a21\u578b\u51b3\u5b9a\u4e0a\u9650\uff1a<\/strong>\u00a0\u6307\u4ee4\u5fae\u8c03\u6a21\u578b\u5728 OIBench \u4e0a\u7684\u8868\u73b0\u9ad8\u5ea6\u4f9d\u8d56\u5176\u57fa\u7840\u6a21\u578b\u7684\u80fd\u529b\uff0c\u8bf4\u660e\u57fa\u7840\u6a21\u578b\u7684\u9884\u8bad\u7ec3\u8d28\u91cf\u662f\u51b3\u5b9a\u4ee3\u7801\u80fd\u529b\u7684\u5173\u952e\u3002<\/li>\n\n\n\n<li><strong>DeepSeek-V3-0324 \u7684\u4eae\u70b9\uff1a\u00a0<\/strong>\u4f5c\u4e3a\u975e\u63a8\u7406\u6a21\u578b\uff0cDeepSeek-V3-0324 \u8868\u73b0\u7a81\u51fa\uff0c\u5f97\u76ca\u4e8e\u5176\u91c7\u7528\u4e86 DeepSeek-R1 \u7684\u94fe\u5f0f\u63a8\u7406\u84b8\u998f\u65b9\u6848\uff0c\u63a8\u7406\u80fd\u529b\u5927\u5e45\u63d0\u5347\u3002<\/li>\n<\/ul>\n\n\n\n<p><strong>\u8bed\u8a00\u504f\u597d\u4e0e\u4e2d\u82f1\u6587\u5dee\u5f02\uff1a<\/strong>&nbsp;\u6a21\u578b\u5728 JavaScript \u548c Python \u4e0a\u7684\u8868\u73b0\u5e73\u5747\u6bd4 C++ \u548c Java \u4f4e 10% \u4ee5\u4e0a\uff0c\u53ef\u80fd\u4e0e\u8bad\u7ec3\u6570\u636e\u5206\u5e03\u6709\u5173\uff1b\u4e2d\u82f1\u6587\u9898\u76ee\u8868\u73b0\u5dee\u5f02\u6781\u5c0f\uff0c\u751a\u81f3\u4e2d\u6587\u7565\u4f18\u3002<\/p>\n\n\n\n<p><strong>\u4f2a\u4ee3\u7801\uff08Pseudocode\uff09\u63d0\u793a\u7684\u79ef\u6781\u4f5c\u7528<\/strong><\/p>\n\n\n\n<p>&nbsp;OIBench \u7684\u9ad8\u96be\u5ea6\u5bf9\u666e\u901a\u6a21\u578b\u6765\u8bf4\u6311\u6218\u5de8\u5927\u3002\u4e3a\u4e86\u66f4\u7ec6\u81f4\u5730\u5206\u6790\u6a21\u578b\u7684\u80fd\u529b\uff0c\u6211\u4eec\u8fd8\u5f15\u5165\u4e86 \u201c\u4f2a\u4ee3\u7801\u63d0\u793a\u201d \u8bc4\u6d4b\uff1a\u5c06\u6807\u51c6\u89e3\u7b54\u8f6c\u4e3a\u4f2a\u4ee3\u7801\u5e76\u4f5c\u4e3a\u63d0\u793a\u8f93\u5165\uff0c\u8003\u67e5\u6a21\u578b\u7406\u89e3\u548c\u590d\u73b0\u89e3\u9898\u601d\u8def\u7684\u80fd\u529b\u3002<\/p>\n\n\n\n<p>\u7ed3\u679c\u663e\u793a\uff0c\u6240\u6709\u6a21\u578b\u5728\u6709\u4f2a\u4ee3\u7801\u63d0\u793a\u65f6\u8868\u73b0\u5747\u6709\u660e\u663e\u63d0\u5347\uff0c\u5c24\u5176\u662f\u5f3a\u63a8\u7406\u6a21\u578b\uff08\u5982 o3-mini-high \u548c o4-mini-high\uff09\u63d0\u5347\u5c24\u4e3a\u663e\u8457\u3002\u8fd9\u8bf4\u660e\u4f2a\u4ee3\u7801\u6781\u5927\u964d\u4f4e\u4e86\u9898\u76ee\u7684\u63a8\u7406\u96be\u5ea6\uff0c\u66f4\u80fd\u8003\u67e5\u6a21\u578b\u7684\u4ee3\u7801\u7406\u89e3\u4e0e\u751f\u6210\u80fd\u529b\u3002\u540c\u65f6\uff0c\u63a8\u7406\u6a21\u578b\u5728\u7406\u89e3\u89e3\u9898\u601d\u8def\u65b9\u9762\u4f9d\u7136\u5177\u5907\u4f18\u52bf\u3002\u8fdb\u4e00\u6b65\u5206\u6790\u53d1\u73b0\uff0c\u6307\u4ee4\u5fae\u8c03\u6a21\u578b\u7684\u8868\u73b0\u4e0e\u5176\u57fa\u7840\u6a21\u578b\u9ad8\u5ea6\u76f8\u5173\uff0c\u8bf4\u660e\u4ee3\u7801\u751f\u6210\u80fd\u529b\u4e3b\u8981\u53d6\u51b3\u4e8e\u9884\u8bad\u7ec3\u6c34\u5e73\u3002<\/p>\n\n\n\n<p>\u5728\u63d0\u4f9b\u4f2a\u4ee3\u7801\u63d0\u793a\u540e\uff0c\u6240\u6709\u6a21\u578b\u8868\u73b0\u5747\u6709\u660e\u663e\u63d0\u5347\uff0c\u5c24\u5176\u662f\u5f3a\u63a8\u7406\u6a21\u578b\uff0c\u8fd9\u8bf4\u660e\u4f2a\u4ee3\u7801\u80fd\u6709\u6548\u964d\u4f4e\u63a8\u7406\u96be\u5ea6\uff0c\u66f4\u80fd\u8003\u67e5\u6a21\u578b\u7684\u4ee3\u7801\u7406\u89e3\u4e0e\u751f\u6210\u80fd\u529b\u3002<\/p>\n\n\n\n<p><strong>\u63a8\u7406\u6548\u7387\uff1a<\/strong>&nbsp;\u968f\u7740 \u201c\u6d4b\u8bd5\u65f6\u63a8\u7406\u201d \u6210\u4e3a\u63d0\u5347\u5927\u6a21\u578b\u80fd\u529b\u7684\u91cd\u8981\u624b\u6bb5\uff0c OpenAI-o1\u3001DeepSeek-R1 \u7b49\u6a21\u578b\u5728\u89e3\u9898\u65f6\u4f1a\u751f\u6210\u5927\u91cf\u63a8\u7406\u5185\u5bb9\u3002\u6211\u4eec\u7edf\u8ba1\u4e86\u5404\u6a21\u578b\u63a8\u7406\u65f6\u7684 Token \u6d88\u8017\u4e0e\u901a\u8fc7\u7387\u7684\u5173\u7cfb\uff0c\u53d1\u73b0 o4-mini-high \u80fd\u4ee5\u66f4\u5c11\u7684 Token \u89e3\u51fa\u66f4\u591a\u9898\u76ee\uff0c\u63a8\u7406\u6548\u7387\u6700\u9ad8\uff1bDeepSeek-V3-0324 \u867d\u7136\u8868\u73b0\u4e0d\u4fd7\uff0c\u4f46\u63a8\u7406 Token \u6570\u91cf\u4e5f\u6700\u591a\uff0c\u4f53\u73b0\u5176\u957f\u94fe\u63a8\u7406\u7684\u7279\u70b9\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"555\" src=\"http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-598-1024x555.jpg\" alt=\"\" class=\"wp-image-1503\" srcset=\"http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-598-1024x555.jpg 1024w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-598-300x163.jpg 300w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-598-768x417.jpg 768w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-598.jpg 1108w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<p><strong>3. \u6a21\u578b\u4e0e\u4eba\u7c7b\u9009\u624b\u7684\u5bf9\u6bd4<\/strong><\/p>\n\n\n\n<p><strong>\u8bb8\u591a\u6280\u672f\u4eba\u5458\u90fd\u5173\u5fc3\uff1a<\/strong>\u73b0\u5728\u7684\u5927\u8bed\u8a00\u6a21\u578b\u5728\u7b97\u6cd5\u7f16\u7a0b\u9898\u4e0a\u7684\u8868\u73b0\uff0c\u548c\u771f\u6b63\u7684\u7ade\u8d5b\u9009\u624b\u76f8\u6bd4\u5230\u5e95\u5982\u4f55\uff1fOpenAI\u3001 DeepSeek \u4f1a\u7528\u7ebf\u4e0a\u7f16\u7a0b\u5e73\u53f0 Codeforces \u7684 Elo \u8bc4\u5206\u4f53\u7cfb\u6765\u505a\u6a21\u578b\u4e0e\u4eba\u7c7b\u7684\u5bf9\u6bd4\uff0c\u5e76\u62a5\u544a\u81ea\u5bb6\u6a21\u578b\u6700\u65b0\u7684 Elo \u5206\u6570\uff0c\u4f46\u8fd9\u79cd\u65b9\u5f0f\u5b58\u5728\u4e00\u4e9b\u95ee\u9898\uff1a\u6bd4\u5982\u6570\u636e\u65f6\u95f4\u8de8\u5ea6\u957f\uff08\u4e00\u822c\u9700\u8981\u534a\u5e74\u4ee5\u4e0a\u7684\u53c2\u8d5b\u8bb0\u5f55\uff09\u3001\u5728\u7ebf\u9009\u624b\u6c34\u5e73\u6ce2\u52a8\u5927\uff0c\u5bfc\u81f4\u5bf9\u6bd4\u7ed3\u679c\u4e0d\u591f\u7cbe\u786e\uff0c\u4e5f\u4e0d\u5bb9\u6613\u590d\u73b0\u3002<\/p>\n\n\n\n<p><strong>OIBench \u521b\u65b0\u6027\u5730\u91c7\u7528\u4e86\u66f4\u53ef\u63a7\u7684\u65b9\u6cd5\uff1a<\/strong>\u9080\u8bf7\u4e86\u4e2d\u56fd 985 \u7ea7\u522b\u9ad8\u6821 ACM \u6821\u961f\u9009\u624b\u53c2\u4e0e\u90e8\u5206\u9898\u76ee\u7684\u4f5c\u7b54\uff0c\u5e76\u5c06\u5176\u6210\u7ee9\u4e0e\u5927\u6a21\u578b\u76f4\u63a5\u5bf9\u6bd4\uff0c\u63d0\u4f9b\u4e86\u66f4\u7cbe\u51c6\u3001\u53ef\u590d\u73b0\u7684\u4eba\u673a\u5bf9\u6bd4\u6570\u636e\uff1b\u6211\u4eec\u7528\u5c0f\u63d0\u7434\u56fe\u5c55\u793a\u4e86\u6bcf\u4e2a\u6a21\u578b\u5728\u6240\u6709\u4eba\u7c7b\u9009\u624b\u4e2d\u7684\u6392\u540d\u5206\u5e03\uff0c\u80fd\u76f4\u89c2\u53cd\u6620\u6a21\u578b\u4e0e\u4eba\u7c7b\u5728\u4e0d\u540c\u9898\u76ee\u4e0a\u7684\u8868\u73b0\u5dee\u5f02\u3002<\/p>\n\n\n\n<p>\u6392\u540d\u89c4\u5219\u53c2\u8003\u4e86\u4fe1\u606f\u5b66\u5965\u8d5b\uff08IOI\uff09\u7684\u6807\u51c6\uff1a\u5148\u6bd4\u8f83\u901a\u8fc7\u7684\u6d4b\u8bd5\u7528\u4f8b\u6570\u91cf\uff0c\u6570\u91cf\u76f8\u540c\u5219\u6309\u8fd0\u884c\u65f6\u95f4\u6392\u5e8f\uff08\u8d8a\u5feb\u8d8a\u9ad8\uff09\uff1b<\/p>\n\n\n\n<p><strong>\u63d0\u4ea4\u6807\u51c6\uff1a<\/strong>\u4eba\u7c7b\u9009\u624b\u7684\u7b54\u6848\u4ee5\u6700\u540e\u4e00\u6b21\u63d0\u4ea4\u4e3a\u51c6\u3002<\/p>\n\n\n\n<p><strong>\u4eba\u7c7b\u89e3\u7b54\u5f00\u6e90:&nbsp;<\/strong>\u5206\u6790\u4e2d\u6240\u6d89\u53ca\u7684\u4eba\u7c7b\u89e3\u7b54\u8bb0\u5f55\u4e5f\u5c06\u533f\u540d\u5316\u5e76\u5f00\u6e90\uff0c\u4fbf\u4e8e\u540e\u7eed\u7814\u7a76\u548c\u590d\u73b0\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"684\" src=\"http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-599-1024x684.jpg\" alt=\"\" class=\"wp-image-1504\" srcset=\"http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-599-1024x684.jpg 1024w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-599-300x200.jpg 300w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-599-768x513.jpg 768w, http:\/\/www.applicationsofllm.com\/wp-content\/uploads\/2025\/09\/image-599.jpg 1108w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<p><sup>\u5728\u5c0f\u63d0\u7434\u56fe\u4e2d\uff0c\u5404\u6a21\u578b\u5728\u6bcf\u9053\u9898\u4e2d\u7684\u4eba\u7c7b\u6392\u540d\u4f4d\u7f6e\u4f1a\u4f5c\u4e3a\u4e00\u4e2a\u6570\u636e\u70b9\uff0c\u8fd9\u4e9b\u6570\u636e\u70b9\u5f62\u6210\u7684\u6982\u7387\u5bc6\u5ea6\u56fe\u5c31\u662f\u5c0f\u63d0\u7434\u56fe\u4e2d\u7684\u201c\u7434\u8eab\u201d\u3002\u201c\u7434\u8eab\u201d\u7684\u5bbd\u5ea6\u663e\u793a\u6a21\u578b\u6392\u540d\u5206\u5e03\u7684\u5bc6\u5ea6\uff0c\u8d8a\u5bbd\u8868\u793a\u6a21\u578b\u5728\u5bf9\u5e94\u7684\u6392\u540d\u533a\u95f4\u5185\u51fa\u73b0\u7684\u9891\u7387\u8d8a\u9ad8\uff0c\u4ece\u800c\u76f4\u89c2\u5730\u53cd\u6620\u51fa\u6a21\u578b\u6392\u540d\u8868\u73b0\u7684\u96c6\u4e2d\u8d8b\u52bf\u3002\u4e2d\u592e\u7684\u6846\u7ebf\u4ee3\u8868\u6392\u540d\u6570\u636e\u6700\u96c6\u4e2d\u7684\u533a\u57df\uff0c\u4ee5o4-mini-high\u4e3e\u4f8b\uff0c\u5b83\u7684\u6392\u540d\u5927\u81f4\u8d85\u8fc7\u4e8642%\u7684\u4eba\u7c7b\u9009\u624b\u3002<\/sup><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u4e09\u79cd\u7c7b\u578b\u7684\u6a21\u578b\u8868\u73b0\uff1a<\/li>\n\n\n\n<li>\u4f4e\u8c37\u578b\uff1a \u591a\u6570\u9898\u76ee\u6392\u540d\u9760\u540e\uff0c\u53ea\u80fd\u8d85\u8d8a\u4e0d\u5230 20% \u7684\u4eba\u7c7b\u9009\u624b\uff0c\u591a\u4e3a\u6ca1\u6709\u957f\u94fe\u63a8\u7406\u80fd\u529b\u7684\u6a21\u578b\u3002<\/li>\n\n\n\n<li>\u53cc\u5cf0\u578b\uff1a \u5728\u90e8\u5206\u9898\u76ee\u4e0a\u80fd\u8d85\u8d8a\u4e00\u534a\u4eba\u7c7b\u9009\u624b\uff0c\u4f46\u5728\u53e6\u4e00\u4e9b\u9898\u76ee\u4e0a\u8868\u73b0\u8f83\u5dee\uff0c\u591a\u6570\u652f\u6301\u957f\u94fe\u63a8\u7406\u7684\u6a21\u578b\u5c5e\u4e8e\u6b64\u7c7b\u578b\uff0c\u663e\u793a\u5176\u5728\u7279\u5b9a\u9898\u578b\u4e0a\u7684\u4f18\u52bf\u548c\u77ed\u677f\u3002<\/li>\n\n\n\n<li>\u6a44\u6984\u578b\uff1a \u6392\u540d\u5206\u5e03\u66f4\u5747\u5300\uff0c\u8868\u73b0\u66f4\u63a5\u8fd1\u4eba\u7c7b\u6574\u4f53\u80fd\u529b\u5206\u5e03\uff0c\u76ee\u524d\u53ea\u6709 o4-mini-high \u5177\u5907\u8fd9\u79cd\u5168\u9762\u548c\u7a33\u5b9a\u7684\u63a8\u7406\u7279\u5f81\u3002<\/li>\n<\/ul>\n\n\n\n<p><strong>4. \u603b\u7ed3\u4e0e\u5c55\u671b<\/strong><\/p>\n\n\n\n<p>\u672c\u6587\u6df1\u5165\u5206\u6790\u4e86\u5f53\u524d\u5927\u6a21\u578b\u7f16\u7a0b\u80fd\u529b\u8bc4\u4f30\u4e2d\u5b58\u5728\u7684\u8ba4\u77e5\u9e3f\u6c9f\uff0c\u63ed\u793a\u4e86 \u201c\u5ba3\u4f20\u201d \u4e0e \u201c\u73b0\u5b9e\u201d \u4e4b\u95f4\u7684\u5dee\u8ddd\u3002Meituan-M17\u56e2\u961f \u901a\u8fc7 OIBench \u8fd9\u4e00\u9ad8\u8d28\u91cf\u3001\u9ad8\u533a\u5206\u5ea6\u7684\u79c1\u6709\u6570\u636e\u96c6\uff0c\u6e05\u6670\u63ed\u793a\u4e86\u9876\u7ea7 LLM \u5728\u9762\u5bf9\u590d\u6742\u7b97\u6cd5\u6311\u6218\u65f6\uff0c\u4e0e\u4eba\u7c7b\u9876\u5c16\u6c34\u5e73\u4e4b\u95f4\u7684\u771f\u5b9e\u5dee\u8ddd\u3002\u4e0d\u4ec5\u4e3a\u5927\u8bed\u8a00\u6a21\u578b\u7684\u7b97\u6cd5\u63a8\u7406\u80fd\u529b\u8bc4\u6d4b\u6811\u7acb\u4e86\u4e00\u4e2a\u5168\u65b0\u6807\u6746\uff0c\u4e5f\u4e3a\u6574\u4e2a\u884c\u4e1a\u5e26\u6765\u4e86\u66f4\u591a\u601d\u8003\u3002<\/p>\n\n\n\n<p>\u5b83\u8ba9\u6211\u4eec\u770b\u5230\uff1a\u5373\u4f7f\u5728\u6a21\u578b\u80fd\u529b\u7a81\u98de\u731b\u8fdb\u7684\u4eca\u5929\uff0c\u771f\u6b63\u9ad8\u8d28\u91cf\u3001\u9ad8\u96be\u5ea6\u7684\u7b97\u6cd5\u6311\u6218\u4f9d\u7136\u80fd\u591f \u201c\u96be\u5012\u201d \u6700\u5148\u8fdb\u7684 AI\u3002\u5c24\u4e3a\u91cd\u8981\u7684\u662f\uff0c\u5e0c\u671b OIBench \u7684\u5f00\u6e90\u548c\u900f\u660e\u80fd\u591f\u4e3a\u793e\u533a\u534f\u4f5c\u548c\u6301\u7eed\u521b\u65b0\u505a\u51fa\u4e00\u4e9b\u8d21\u732e\u3002\u6211\u4eec\u671f\u5f85\u5b83\u80fd\u6210\u4e3a\u8fde\u63a5\u5b66\u672f\u3001\u4ea7\u4e1a\u548c\u5f00\u53d1\u8005\u7684\u6865\u6881\uff0c\u63a8\u52a8\u5927\u6a21\u578b\u5728\u7b97\u6cd5\u667a\u80fd\u9886\u57df\u8fc8\u5411\u65b0\u9ad8\u5ea6\u3002\u672a\u6765\uff0c\u968f\u7740\u6a21\u578b\u80fd\u529b\u548c\u8bc4\u6d4b\u9700\u6c42\u7684\u4e0d\u65ad\u6f14\u8fdb\uff0cOIBench \u4e5f\u4f1a\u6301\u7eed\u8fed\u4ee3\uff0c\u4e0e\u5927\u5bb6\u5171\u540c\u89c1\u8bc1 AI \u63a8\u7406\u7684\u8fdb\u5316\u4e4b\u8def\u3002<\/p>\n\n\n\n<p>\u4e0e\u6b64\u540c\u65f6\uff0c\u6211\u4eec\u4e5f\u89c2\u5bdf\u5230\uff0c\u5bf9\u4e8e\u5927\u591a\u6570\u4eba\u7c7b\u5f00\u53d1\u8005\u6765\u8bf4\uff0c\u5373\u4f7f\u4ed6\u4eec\u63a5\u53d7\u8fc7\u4e13\u4e1a\u7684\u7b97\u6cd5\u8bbe\u8ba1\u8bad\u7ec3\uff0c\u9762\u5bf9\u9ad8\u96be\u5ea6\u7b97\u6cd5\u548c\u590d\u6742\u7cfb\u7edf\u8bbe\u8ba1\uff0c\u540c\u6837\u9700\u8981\u5de5\u5177\u548c\u667a\u80fd\u52a9\u624b\u7684\u8f85\u52a9\u624d\u80fd\u66f4\u4e0a\u4e00\u5c42\u697c\u3002\u5927\u6a21\u578b\u7684\u5f3a\u5927\u63a8\u7406\u548c\u4ee3\u7801\u751f\u6210\u80fd\u529b\uff0c\u6b63\u597d\u80fd\u4e3a\u4eba\u7c7b\u5f00\u53d1\u8005\u63d0\u4f9b\u6709\u529b\u652f\u6301\uff0c\u5e2e\u52a9\u4ed6\u4eec\u63d0\u5347\u7b97\u6cd5\u8bbe\u8ba1\u548c\u4ee3\u7801\u5b9e\u73b0\u7684\u6548\u7387\u3002<strong>OIBench \u4fc3\u4f7f\u6211\u4eec\u6df1\u5165\u601d\u8003\uff1a\u672a\u6765\u7684\u4ee3\u7801\u5f00\u53d1\uff0c\u5df2\u8d85\u8d8a \u201c\u4eba\u201d \u6216 \u201c\u6a21\u578b\u201d \u5355\u6253\u72ec\u6597\u7684\u6a21\u5f0f\uff0c\u8f6c\u53d8\u4e3a\u4eba\u673a\u534f\u540c\u3001\u4f18\u52bf\u4e92\u8865\u7684\u65b0\u8303\u5f0f\u3002<\/strong><\/p>\n\n\n\n<p><strong>5. OI \u8d5b\u9898\u4e4b\u5916 \u2014\u2014 Code Agent \u50ac\u751f\u7684\u8bc4\u6d4b\u8303\u5f0f\u8fc1\u79fb<\/strong><\/p>\n\n\n\n<p>\u5f53\u524d\u5927\u91cf\u6d8c\u73b0\u7684 Code Agent \u7c7b\u6846\u67b6\u4e0e\u4ea7\u54c1\uff0c\u4f7f\u5f97\u4eba\u673a\u534f\u4f5c\u89e3\u51b3\u66f4\u52a0\u590d\u6742\u7684\u5de5\u7a0b\u95ee\u9898\u6210\u4e3a\u53ef\u80fd\uff0c\u8fd9\u9884\u793a\u7740\u5bf9 Code Agent \u5728\u5b9e\u9645\u5de5\u7a0b\u573a\u666f\u4e2d\u4e0e\u4eba\u7c7b\u534f\u4f5c\u80fd\u529b\u7684\u8bc4\u4f30\uff0c\u5c06\u53d8\u5f97\u65e5\u76ca\u5173\u952e\u3002\u7136\u800c\uff0c\u73b0\u6709\u7684 Code Agent \u8bc4\u6d4b\u57fa\u51c6\uff08\u5982 SWE-bench \u7cfb\u5217\uff09\u5b58\u5728\u4e00\u4e2a\u6838\u5fc3\u95ee\u9898\uff1a<strong>\u5b83\u4eec\u5c06\u4eba\u7c7b\u5f00\u53d1\u8005\u5b8c\u5168\u6392\u9664\u5728\u8bc4\u4f30\u6d41\u7a0b\u4e4b\u5916\u3002<\/strong>\u8fd9\u79cd \u201c\u7aef\u5230\u7aef\u201d \u7684\u81ea\u52a8\u5316\u8bc4\u6d4b\uff0c\u867d\u7136\u80fd\u6bd4\u8f83\u5bb9\u6613\u7684\u91cf\u5316\u6a21\u578b\u5728\u5c01\u95ed\u4efb\u52a1\u4e0a\u7684\u8868\u73b0\uff0c\u5374\u65e0\u6cd5\u56de\u7b54\u4e00\u4e2a\u66f4\u5173\u952e\u7684\u95ee\u9898\uff1a\u5728\u771f\u5b9e\u3001\u5f00\u653e\u7684\u5f00\u53d1\u73af\u5883\u4e2d\uff0cCode Agent \u80fd\u5426\u4e0e\u4eba\u7c7b\u9ad8\u6548\u534f\u4f5c\uff1f\u5f53\u524d\u591a\u6570 Code Agent \u6846\u67b6\u5728\u4ea4\u4e92\u8bbe\u8ba1\u4e0a\u5bf9\u4eba\u673a\u4ea4\u4e92\u7684\u5ffd\u89c6\uff0c\u5bfc\u81f4\u5176\u8bc4\u6d4b\u7ed3\u679c\u4e0e\u5b9e\u9645\u5e94\u7528\u4ef7\u503c\u4e4b\u95f4\u5b58\u5728\u660e\u663e\u8131\u8282\u3002<\/p>\n\n\n\n<p>\u7ed3\u5408 OIBench \u5f15\u53d1\u7684\u5173\u4e8e\u4eba\u673a\u534f\u540c\u3001\u4f18\u52bf\u4e92\u8865\u7684\u601d\u8003\uff0cMeituan-M17\u56e2\u961f \u4e5f\u5f00\u59cb\u5173\u6ce8\u4eba\u673a\u534f\u4f5c\u8bc4\u6d4b\u8fd9\u4e00\u65b0\u7684\u8bc4\u6d4b\u8303\u5f0f\u5728 Code Agent \u573a\u666f\u7684\u5e94\u7528\uff0c\u8fdb\u800c\u5f25\u8865\u5f53\u524d\u8303\u5f0f\u5f15\u8d77\u7684\u8bc4\u6d4b\u7ed3\u679c\u4e0e\u5b9e\u9645\u5e94\u7528\u4ef7\u503c\u95f4\u7684\u9e3f\u6c9f\u3002\u57fa\u4e8e\u6b64\uff0c\u6211\u4eec\u4e0e AGI-Eval \u8bc4\u6d4b\u793e\u533a\u5408\u4f5c\uff0c\u8bbe\u8ba1\u5e76\u8ba1\u5212\u4e3e\u529e\u4e00\u9879\u521b\u65b0\u7684\u4eba\u673a\u534f\u4f5c\u7f16\u7a0b\u7ade\u8d5b\u3002<\/p>\n\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u5f53\u524d\uff0c\u5927\u8bed\u8a00\u6a21\u578b\uff08LLMs\uff09\u5728\u7f16\u7a0b\u9886\u57df\u7684\u80fd\u529b\u53d7\u5230\u5e7f\u6cdb\u5173\u6ce8\uff0c\u76f8\u5173\u8bba\u65ad\u5728\u5e02\u573a\u4e2d\u666e\u904d\u5b58\u5728\uff0c\u4f8b\u5982 DeepMind \u7684 [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":1504,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[],"class_list":["post-1501","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-8"],"_links":{"self":[{"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/posts\/1501","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/comments?post=1501"}],"version-history":[{"count":1,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/posts\/1501\/revisions"}],"predecessor-version":[{"id":1508,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/posts\/1501\/revisions\/1508"}],"wp:featuredmedia":[{"embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/media\/1504"}],"wp:attachment":[{"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/media?parent=1501"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/categories?post=1501"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/tags?post=1501"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}