{"id":1234,"date":"2025-07-16T17:24:00","date_gmt":"2025-07-16T09:24:00","guid":{"rendered":"http:\/\/www.applicationsofllm.com\/?p=1234"},"modified":"2025-09-17T17:27:54","modified_gmt":"2025-09-17T09:27:54","slug":"%e6%97%a0%e9%9c%80%e5%be%ae%e8%b0%83%e5%b0%b1%e8%b6%85%e8%b6%8a%e4%b8%b4%e5%ba%8a%e5%8c%bb%e7%94%9f%ef%bc%9f%e5%9f%83%e9%bb%98%e9%87%8c%e5%a4%a7%e5%ad%a6%e7%b3%bb%e7%bb%9f%e8%af%84%e4%bc%b0gpt-5","status":"publish","type":"post","link":"http:\/\/www.applicationsofllm.com\/index.php\/2025\/07\/16\/%e6%97%a0%e9%9c%80%e5%be%ae%e8%b0%83%e5%b0%b1%e8%b6%85%e8%b6%8a%e4%b8%b4%e5%ba%8a%e5%8c%bb%e7%94%9f%ef%bc%9f%e5%9f%83%e9%bb%98%e9%87%8c%e5%a4%a7%e5%ad%a6%e7%b3%bb%e7%bb%9f%e8%af%84%e4%bc%b0gpt-5\/","title":{"rendered":"\u65e0\u9700\u5fae\u8c03\u5c31\u8d85\u8d8a\u4e34\u5e8a\u533b\u751f\uff1f\u57c3\u9ed8\u91cc\u5927\u5b66\u7cfb\u7edf\u8bc4\u4f30GPT-5\u533b\u7597\u51b3\u7b56\u652f\u6301\u80fd\u529b\uff1a\u6f5c\u529b\u5de8\u5927"},"content":{"rendered":"\n

\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff08LLM\uff09\u5df2\u7ecf\u65e5\u76ca\u53d1\u5c55\u6210\u901a\u7528\u7cfb\u7edf\uff0c\u65e0\u9700\u5927\u91cf\u5fae\u8c03\u5373\u53ef\u6267\u884c\u9886\u57df\u7279\u5b9a\u63a8\u7406\u3002<\/p>\n\n\n\n

\u672c\u6708\u521d\uff0cOpenAI \u53d1\u5e03\u4e86\u65b0\u4e00\u4ee3\u65d7\u8230\u6a21\u578b GPT-5\uff0c\u5e76\u8868\u793a\u5176\u80fd\u591f\u81ea\u5df1\u51b3\u5b9a\u4f55\u65f6\u9700\u8981\u6df1\u5165\u601d\u8003\u3002\u4e00\u4e9b\u7814\u7a76\u5f00\u59cb\u63a2\u7d22 GPT-5 \u5728\u79d1\u5b66\u9886\u57df\u7684\u6f5c\u529b\u3002<\/p>\n\n\n\n

\u5728\u533b\u7597\u9886\u57df\uff0c\u51b3\u7b56\u8fc7\u7a0b\u5f80\u5f80\u9700\u8981\u6574\u5408\u5f02\u6784\u4fe1\u606f\u6e90\uff0c\u5305\u62ec\u60a3\u8005\u4e3b\u8bc9\u3001\u7ed3\u6784\u5316\u6570\u636e\u548c\u533b\u5b66\u5f71\u50cf\u7b49\u3002\u6765\u81ea\u57c3\u9ed8\u91cc\u5927\u5b66\u7684\u7814\u7a76\u56e2\u961f\u5c06 GPT-5 \u5b9a\u4f4d\u4e3a\u533b\u7597\u51b3\u7b56\u652f\u6301\u9886\u57df\u7684\u901a\u7528\u591a\u6a21\u6001\u63a8\u7406\u7cfb\u7edf\uff0c\u5728\u7edf\u4e00\u8bc4\u4f30\u6846\u67b6\u4e0b\u7cfb\u7edf\u6027\u5730\u8bc4\u4f30\u4e86\u5176\u5728\u6587\u672c\u95ee\u7b54\u548c\u89c6\u89c9\u95ee\u7b54\u4efb\u52a1\u4e2d\u7684\u96f6\u6837\u672c\u601d\u7ef4\u94fe\u63a8\u7406\u6027\u80fd\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

\u8bba\u6587\u5730\u5740\uff1ahttps:\/\/arxiv.org\/pdf\/2508.08224<\/p>\n\n\n\n

\u8bc4\u4f30\u4e0e\u7ed3\u679c<\/strong><\/p>\n\n\n\n

\u7814\u7a76\u56e2\u961f\u5728 MedQA\u3001MedXpertQA\uff08\u6587\u672c\u4e0e\u591a\u6a21\u6001\uff09\u3001MMLU \u533b\u5b66\u5b50\u96c6\u3001USMLE \u81ea\u6211\u8bc4\u4f30\u4ee5\u53ca VQA-RAD \u7684\u6807\u51c6\u6570\u636e\u96c6\u4e0a\u5bf9 GPT-5\u3001GPT-5-mini\u3001GPT-5-nano \u548c GPT-4o-2024-11-20 \u8fdb\u884c\u4e86\u57fa\u51c6\u6d4b\u8bd5\u3002<\/p>\n\n\n\n

\u6587\u672c\u95ee\u7b54<\/strong><\/p>\n\n\n\n

\u5728 MedQA \u57fa\u51c6\u6d4b\u8bd5\u4e2d\uff0cGPT-5 \u7684\u51c6\u786e\u7387\u8fbe\u5230 95.84%\uff0c\u8f83 GPT-4o \u7edd\u5bf9\u63d0\u5347 4.80%\uff0c\u8868\u660e GPT-5 \u5728\u4e34\u5e8a\u95ee\u9898\u60c5\u5883\u4e2d\u5177\u6709\u66f4\u5f3a\u7684\u4e8b\u5b9e\u53ec\u56de\u4e0e\u8bca\u65ad\u63a8\u7406\u80fd\u529b\u3002<\/p>\n\n\n\n

\u66f4\u663e\u8457\u7684\u63d0\u5347\u51fa\u73b0\u5728 MedXpertQA \u6587\u672c\u5b50\u96c6\uff0cGPT-5 \u7684\u63a8\u7406\u51c6\u786e\u7387\u8f83 GPT-4o \u63d0\u9ad8 26.33%\uff0c\u7406\u89e3\u51c6\u786e\u7387\u63d0\u5347 25.30%\uff0c\u8868\u660e GPT-5 \u5728\u591a\u6b65\u63a8\u7406\u548c\u533b\u5b66\u53d9\u8ff0\u7406\u89e3\u65b9\u9762\u53d6\u5f97\u4e86\u5b9e\u8d28\u6027\u589e\u5f3a\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

USMLE \u81ea\u6211\u8bc4\u4f30<\/strong><\/p>\n\n\n\n

\u5728\u7f8e\u56fd\u533b\u5e08\u6267\u7167\u8003\u8bd5 (USMLE) \u81ea\u6211\u8bc4\u4f30\u4e2d\uff0cGPT-5 \u5728\u4e09\u4e2a\u8003\u8bd5\u9636\u6bb5\u5747\u8d85\u8d8a\u6240\u6709\u57fa\u7ebf\u6a21\u578b\uff0c\u5176\u4e2d\u5728 Step 2\uff08\u4e34\u5e8a\u51b3\u7b56\u4e0e\u7ba1\u7406\uff09\u9636\u6bb5\u4f18\u52bf\u6700\u4e3a\u663e\u8457\uff08+4.17%\uff09\u3002\u8fd9\u4e00\u7ed3\u679c\u4e0e GPT-5 \u6539\u8fdb\u7684\u601d\u7ef4\u94fe\u63a8\u7406\u80fd\u529b\u9ad8\u5ea6\u5951\u5408\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

\u4e09\u9636\u6bb5\u5e73\u5747\u5f97\u5206\u8fbe\u5230 95.22%\uff08\u8f83 GPT-4o \u63d0\u5347 2.88%\uff09\uff0c\u8fdc\u8d85\u4eba\u7c7b\u901a\u8fc7\u8003\u8bd5\u5e38\u89c4\u9608\u503c\uff0c\u8868\u660e GPT-5 \u5df2\u5177\u5907\u5e94\u5bf9\u9ad8\u98ce\u9669\u4e34\u5e8a\u63a8\u7406\u4efb\u52a1\u7684\u6210\u719f\u80fd\u529b\u3002<\/p>\n\n\n\n

\u89c6\u89c9\u95ee\u7b54<\/strong><\/p>\n\n\n\n

\u5728\u591a\u6a21\u6001\u63a8\u7406\u4efb\u52a1\u4e2d\uff0cGPT-5 \u5728 MedXpertQA \u591a\u6a21\u6001\u5b50\u96c6\u4e0a\u5b9e\u73b0\u98de\u8dc3\u6027\u7a81\u7834\uff1a\u63a8\u7406\u548c\u7406\u89e3\u51c6\u786e\u7387\u5206\u522b\u8f83 GPT-4o \u63d0\u5347 29.26% \u548c 26.18%\uff0c\u8868\u660e\u5176\u89c6\u89c9\u4e0e\u6587\u672c\u7ebf\u7d22\u7684\u6574\u5408\u80fd\u529b\u5f97\u5230\u663e\u8457\u589e\u5f3a\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

\u5728 MedXpertQA \u591a\u6a21\u6001\u57fa\u51c6\u7684\u5178\u578b\u6848\u4f8b\u4e2d\uff0cGPT-5 \u5c55\u73b0\u51fa\u4ee5\u4e34\u5e8a\u903b\u8f91\u6574\u5408\u591a\u6a21\u6001\u4fe1\u606f\u7684\u80fd\u529b\uff1aGPT-5 \u6839\u636e CT \u5f71\u50cf\u8868\u73b0\u3001\u5b9e\u9a8c\u5ba4\u6570\u503c\u53ca\u53cd\u590d\u5455\u5410\u540e\u7684\u5173\u952e\u4f53\u5f81\uff0c\u51c6\u786e\u5730\u5c06\u98df\u7ba1\u7a7f\u5b54\u5217\u4e3a\u6700\u53ef\u80fd\u8bca\u65ad\uff1b\u968f\u540e\u63a8\u8350\u6cdb\u5f71\u8461\u80fa\u541e\u54bd\u68c0\u67e5\u4f5c\u4e3a\u4e0b\u4e00\u6b65\u8bca\u7597\u63aa\u65bd\uff0c\u540c\u65f6\u660e\u786e\u6392\u9664\u5176\u4ed6\u9009\u9879\u5e76\u9010\u4e00\u8bba\u8bc1\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

\u8fd9\u4e00\u6848\u4f8b\u4f53\u73b0\u51fa GPT-5 \u80fd\u591f\u6574\u5408\u89c6\u89c9\u8bc1\u636e\u4e0e\u590d\u6742\u53d9\u8ff0\u8bed\u5883\u3001\u4fdd\u6301\u7ed3\u6784\u5316\u8bca\u65ad\u63a8\u7406\u94fe\uff0c\u6700\u7ec8\u505a\u51fa\u7b26\u5408\u4e13\u5bb6\u5171\u8bc6\u7684\u9ad8\u98ce\u9669\u4e34\u5e8a\u51b3\u7b56\u3002<\/p>\n\n\n\n

\u5bf9\u6bd4\u4eba\u7c7b\u4e13\u5bb6<\/strong><\/p>\n\n\n\n

\u5982\u4e0b\u8868\u6240\u793a\uff0c\u7814\u7a76\u56e2\u961f\u5c06\u6a21\u578b\u4e0e\u9884\u6267\u4e1a\u4eba\u7c7b\u4e13\u5bb6\u8fdb\u884c\u4e86\u5bf9\u6bd4\uff0cGPT-4o \u5728\u591a\u6570\u7ef4\u5ea6\u4e0a\u4f4e\u4e8e\u4eba\u7c7b\u4e13\u5bb6\u6c34\u5e73\uff0c\u5728\u6587\u672c\u4e0e\u591a\u6a21\u6001\u573a\u666f\u4e2d\uff0c\u63a8\u7406\u548c\u7406\u89e3\u51c6\u786e\u7387\u8f83\u4eba\u7c7b\u4e13\u5bb6\u4f4e 5.03%-15.90%\uff1b\u800c GPT-5 \u4e0d\u4ec5\u5f25\u5408\u4e86\u8fd9\u4e00\u5dee\u8ddd\uff0c\u8fd8\u4ee5\u663e\u8457\u4f18\u52bf\u8d85\u8d8a\u4e86\u4eba\u7c7b\u4e13\u5bb6\uff1a\u6587\u672c\u63a8\u7406\uff08+15.22%\uff09\u3001\u6587\u672c\u7406\u89e3\uff08+9.40%\uff09\u3001\u591a\u6a21\u6001\u63a8\u7406\uff08+24.23%\uff09\u3001\u591a\u6a21\u6001\u7406\u89e3\uff08+29.40%\uff09\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

\u503c\u5f97\u6ce8\u610f\u7684\u662f\uff0cGPT-5 \u5728\u591a\u6a21\u6001\u573a\u666f\u4e2d\u9886\u5148\u4f18\u52bf\u5c24\u4e3a\u7a81\u51fa\uff0c\u5176\u8868\u73b0\u751a\u81f3\u8d85\u8d8a\u7ecf\u9a8c\u4e30\u5bcc\u7684\u4e34\u5e8a\u533b\u751f\u5728\u9650\u65f6\u6d4b\u8bd5\u6761\u4ef6\u4e0b\u7684\u6c34\u5e73\u3002<\/p>\n\n\n\n

\u603b\u7684\u6765\u8bf4\uff0c\u8fd9\u9879\u7814\u7a76\u51f8\u663e\u51fa GPT-5 \u5728\u771f\u5b9e\u4e16\u754c\u4e34\u5e8a\u51b3\u7b56\u652f\u6301\u4e2d\u6f5c\u529b\u5de8\u5927\u3002\u4e0d\u8fc7\uff0c\u9700\u8981\u5f3a\u8c03\u7684\u662f\uff0c\u73b0\u6709\u57fa\u51c6\u6d4b\u8bd5\u53cd\u6620\u7684\u662f\u7406\u60f3\u5316\u8bc4\u4f30\u73af\u5883\uff0c\u53ef\u80fd\u65e0\u6cd5\u5b8c\u5168\u6355\u6349\u771f\u5b9e\u533b\u7597\u5b9e\u8df5\u7684\u591a\u6837\u6027\u3001\u4e0d\u786e\u5b9a\u6027\u53ca\u4f26\u7406\u8003\u91cf\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff08LLM\uff09\u5df2\u7ecf\u65e5\u76ca\u53d1\u5c55\u6210\u901a\u7528\u7cfb\u7edf\uff0c\u65e0\u9700\u5927\u91cf\u5fae\u8c03\u5373\u53ef\u6267\u884c\u9886\u57df\u7279\u5b9a\u63a8\u7406\u3002 \u672c\u6708\u521d\uff0cOpenAI \u53d1\u5e03\u4e86 […]<\/p>\n","protected":false},"author":1,"featured_media":1235,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[10],"tags":[],"class_list":["post-1234","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-10"],"_links":{"self":[{"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/posts\/1234","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=1234"}],"version-history":[{"count":1,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/posts\/1234\/revisions"}],"predecessor-version":[{"id":1242,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/posts\/1234\/revisions\/1242"}],"wp:featuredmedia":[{"embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/media\/1235"}],"wp:attachment":[{"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/media?parent=1234"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/categories?post=1234"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.applicationsofllm.com\/index.php\/wp-json\/wp\/v2\/tags?post=1234"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}