	{"id":1083983,"date":"2026-02-06T10:28:09","date_gmt":"2026-02-06T10:28:09","guid":{"rendered":"https:\/\/www.artefact.com\/?post_type=blog&#038;p=1083983"},"modified":"2026-02-10T16:18:03","modified_gmt":"2026-02-10T16:18:03","slug":"long-run-ai-agents-part-1-the-problem-nobody-talks-about","status":"publish","type":"blog","link":"https:\/\/www.artefact.com\/zh\/blog\/long-run-ai-agents-part-1-the-problem-nobody-talks-about\/","title":{"rendered":"\u957f\u671f AI \u4ee3\u7406\u4eba\uff0c\u7b2c 1 \u90e8\u5206\uff1a\u65e0\u4eba\u95ee\u6d25\u7684\u95ee\u9898"},"content":{"rendered":"<p><em><a href=\"https:\/\/www.linkedin.com\/in\/victor-coimbra-999a02a0\/\" target=\"_blank\" rel=\"noopener\">\u7ef4\u514b\u6258-\u79d1\u82f1\u5e03\u62c9<\/a> \u4ed6\u56e0\u5728\u4eba\u5de5\u667a\u80fd\u521b\u65b0\u65b9\u9762\u7684\u6770\u51fa\u8d21\u732e\u800c\u8363\u767b\u798f\u5e03\u65af\u5df4\u897f 30 \u5c81\u4ee5\u4e0b\u7cbe\u82f1\u699c\u3002\u4ed6\u5171\u540c\u521b\u529e\u4e86 Artefact \u7684\u62c9\u4e01\u7f8e\u6d32\u4e1a\u52a1\uff0c\u8be5\u4e1a\u52a1\u73b0\u5df2\u6210\u4e3a\u62e5\u6709 200 \u540d\u5458\u5de5\u7684\u5168\u7403\u6280\u672f\u4e2d\u5fc3\u3002\u4ed6\u5728\u6269\u5c55\u4eba\u5de5\u667a\u80fd\u89e3\u51b3\u65b9\u6848\u548c\u5efa\u8bbe\u8de8\u56fd\u9645\u5e02\u573a\u7684\u9ad8\u6027\u80fd\u6280\u672f\u56e2\u961f\u65b9\u9762\u62e5\u6709\u6df1\u539a\u7684\u4e13\u4e1a\u77e5\u8bc6\u3002.<\/em><\/p>\n<p>2025 \u5e74 3 \u6708\uff0c\u4e00\u5bb6\u540d\u4e3a METR \u7684\u7814\u7a76\u673a\u6784\u53d1\u8868\u4e86\u4e00\u9879\u7814\u7a76\u6210\u679c\uff0c\u4f46\u5374\u6ca1\u6709\u5f97\u5230\u5e94\u6709\u7684\u5173\u6ce8\u3002\u4ed6\u4eec\u4e00\u76f4\u5728\u6d4b\u91cf\u4e00\u4e9b\u4e0d\u5408\u65f6\u5b9c\u7684\u4e1c\u897f\uff1aAI \u7cfb\u7edf\u5728\u5b8c\u6210\u4efb\u52a1\u591a\u957f\u65f6\u95f4\u540e\u624d\u4f1a\u51fa\u73b0\u6545\u969c\u3002\u800c\u4e0d\u662f\u5b83\u4eec\u5728\u4e00\u6b21\u4ea4\u4e92\u4e2d\u80fd\u505a\u4ec0\u4e48\u3002METR \u60f3\u77e5\u9053\u7684\u662f\uff0c\u8fd9\u4e9b\u7cfb\u7edf\u80fd\u5728\u591a\u957f\u65f6\u95f4\u5185\u4fdd\u6301\u8fde\u8d2f\u3001\u6709\u7528\u7684\u5de5\u4f5c\u3002.<\/p>\n<p>\u4ed6\u4eec\u91c7\u7528\u4e86\u4e25\u683c\u7684\u65b9\u6cd5\uff1a170 \u9879\u4efb\u52a1\u6d89\u53ca\u8f6f\u4ef6\u3001\u7814\u7a76\u548c\u89e3\u51b3\u95ee\u9898\u3002\u7531\u9886\u57df\u4e13\u5bb6\u8fd0\u884c 236 \u6b21\u540e\u5f97\u51fa\u7684\u4eba\u7c7b\u57fa\u7ebf\u3002\u6839\u636e\u4efb\u52a1\u6301\u7eed\u65f6\u95f4\u5efa\u7acb\u7edf\u8ba1\u6a21\u578b\uff0c\u9884\u6d4b\u6210\u529f\u6982\u7387\u3002.<\/p>\n<p>\u4e3b\u8981\u53d1\u73b0\uff1a\u4efb\u52a1\u5b8c\u6210\u80fd\u529b\u8fde\u7eed\u516d\u5e74\u6bcf\u4e03\u4e2a\u6708\u7ffb\u4e00\u756a\u3002.<\/p>\n<p>2024 \u5e74\u4e2d\u671f\uff0c\u9886\u5148\u7684\u4eba\u5de5\u667a\u80fd\u7cfb\u7edf\u53ef\u4ee5\u53ef\u9760\u5730\u5b8c\u6210\u4eba\u7c7b\u4e13\u5bb6\u9700\u8981 19 \u5206\u949f\u624d\u80fd\u5b8c\u6210\u7684\u4efb\u52a1\u3002\u5230 2025 \u5e74\u521d\uff0c\u8fd9\u4e00\u65f6\u95f4\u5ef6\u957f\u5230\u8fd1\u4e00\u4e2a\u5c0f\u65f6\u3002\u6700\u65b0\u7684\u7cfb\u7edf\u80fd\u8fbe\u5230\u4e24\u4e2a\u591a\u5c0f\u65f6\u3002\u66f2\u7ebf\u6b63\u5728\u52a0\u901f\u3002.<\/p>\n<p>\u4f46\u5728\u8fd9\u4e00\u6807\u9898\u4e4b\u4e0b\uff0c\u9690\u85cf\u7740\u4e00\u4e2a\u66f4\u4e3a\u590d\u6742\u7684\u6545\u4e8b\u3002.<\/p>\n<h2>\u6f14\u793a\u4e0e\u4ea4\u4ed8\u4e4b\u95f4\u7684\u5dee\u8ddd<\/h2>\n<p>2025 \u5e74 8 \u6708\uff0cMETR \u53d1\u5e03\u4e86\u4e00\u4efd\u540e\u7eed\u62a5\u544a\uff0c\u8fd9\u5e94\u8be5\u4f1a\u8ba9\u6240\u6709\u4eba\u505c\u987f\u4e0b\u6765\u3002\u4ed6\u4eec\u5bf9\u4eba\u5de5\u667a\u80fd\u8fdb\u884c\u4e86 18 \u9879\u771f\u5b9e\u4efb\u52a1\u7684\u6d4b\u8bd5\uff0c\u8fd9\u4e9b\u4efb\u52a1\u90fd\u662f\u7ec4\u7ec7\u4e2d\u975e\u5e38\u91cd\u8981\u7684\u5de5\u4f5c\u3002.<\/p>\n<p>\u4e00\u4e9b\u7ed3\u679c\u662f<\/p>\n<ul>\n<li>\u81ea\u52a8\u6d4b\u8bd5\u901a\u8fc7\u7387\uff1a38%.<\/li>\n<li>\u5373\u5f00\u5373\u7528\uff0c\u65e0\u9700\u4eba\u5de5\u6e05\u7406\uff1a0%.<\/li>\n<\/ul>\n<p>\u96f6\u3002\u5728\u6ca1\u6709\u4eba\u5de5\u5e72\u9884\u7684\u60c5\u51b5\u4e0b\uff0c\u4eba\u5de5\u667a\u80fd\u4ea7\u751f\u7684\u5de5\u4f5c\u6ca1\u6709\u4e00\u9879\u662f\u53ef\u4ee5\u4f7f\u7528\u7684\u3002\u5373\u4f7f\u901a\u8fc7\u4e86\u81ea\u52a8\u68c0\u67e5\uff0c\u6bcf\u9879\u4ea7\u51fa\u4e5f\u90fd\u5b58\u5728\u6f0f\u6d1e\uff1a\u6587\u4ef6\u7f3a\u5931\u3001\u9a8c\u8bc1\u4e0d\u5b8c\u6574\u3001\u5b58\u5728\u8d28\u91cf\u95ee\u9898\uff0c\u800c\u8fd9\u4e9b\u95ee\u9898\u5728\u771f\u6b63\u7684\u7ec4\u7ec7\u4e2d\u662f\u7edd\u5bf9\u65e0\u6cd5\u901a\u8fc7\u5ba1\u67e5\u7684\u3002.<\/p>\n<p>\u5e73\u5747\u6e05\u7406\u65f6\u95f4\uff1a26 \u5206\u949f\uff0c\u7ea6\u4e3a\u539f\u4efb\u52a1\u65f6\u95f4\u7684\u4e09\u5206\u4e4b\u4e00\u3002.<\/p>\n<p>\u81ea\u52a8\u5316\u57fa\u51c6\u4e0e\u771f\u5b9e\u4e16\u754c\u53ef\u7528\u6027\u4e4b\u95f4\u7684\u5dee\u8ddd\u662f\u65e0\u4eba\u95ee\u6d25\u7684\u7b2c\u4e00\u4e2a\u95ee\u9898\u3002\u4eba\u5de5\u667a\u80fd\u7cfb\u7edf\u5728\u57fa\u51c6\u6240\u8861\u91cf\u7684\u72ed\u4e49\u65b9\u9762\u8d8a\u6765\u8d8a\u597d\u3002\u800c\u5728 \u201c\u5b8c\u6210\u4efb\u52a1 \u201d\u7684\u5b9e\u9645\u610f\u4e49\u65b9\u9762\uff0c\u5b83\u4eec\u7684\u8fdb\u6b65\u8981\u6162\u5f97\u591a\u3002.<\/p>\n<p>\u5bf9\u4e8e\u4efb\u4f55\u8ba1\u5212\u6295\u8d44\u7684\u4eba\u6765\u8bf4\uff0c\u8fd9\u79cd\u533a\u522b\u90fd\u5f88\u91cd\u8981\u3002\u4e00\u4e2a\u901a\u8fc7\u6d4b\u8bd5\u4f46\u8f93\u51fa\u7ed3\u679c\u65e0\u6cd5\u4f7f\u7528\u7684\u7cfb\u7edf\u5e76\u4e0d\u662f\u4e00\u4e2a\u53ef\u4ee5\u81ea\u4e3b\u5de5\u4f5c\u7684\u7cfb\u7edf\u3002\u5b83\u662f\u4e00\u4e2a\u9700\u8981\u6709\u4eba\u6765\u6e05\u7406\u7684\u7cfb\u7edf\u3002.<\/p>\n<h2>\u751f\u4ea7\u529b\u6096\u8bba<\/h2>\n<p>\u8fd9\u5c31\u662f\u7b2c\u4e8c\u4e2a\u95ee\u9898\u30022025 \u5e74 7 \u6708\uff0cMETR \u5bf9 16 \u540d\u7ecf\u9a8c\u4e30\u5bcc\u7684\u4e13\u4e1a\u4eba\u58eb\u8fdb\u884c\u4e86\u4e00\u9879\u5bf9\u7167\u7814\u7a76\u3002\u95ee\u9898\u662f\uff1a\u4eba\u5de5\u667a\u80fd\u662f\u5426\u771f\u7684\u80fd\u8ba9\u4eba\u53d8\u5f97\u66f4\u5feb\uff1f<\/p>\n<p><strong>\u7814\u7a76\u524d\u7684\u9884\u671f\u7ed3\u679c\uff1a\u63d0\u901f 24%\u3002.<\/strong><br \/>\n<strong>\u5b9e\u9645\u7ed3\u679c\uff1a19% \u51cf\u901f\u3002.<\/strong><\/p>\n<p>\u4f7f\u7528\u4eba\u5de5\u667a\u80fd\u7684\u4eba\u6bd4\u6ca1\u6709\u8f85\u52a9\u7684\u4eba\u5b8c\u6210\u4efb\u52a1\u7684\u65f6\u95f4\u66f4\u957f\u3002\u4ee4\u4eba\u4e0d\u5b89\u7684\u662f\uff1a\u7814\u7a76\u7ed3\u675f\u540e\uff0c\u53c2\u4e0e\u8005\u4ecd\u7136\u8ba4\u4e3a\u4ed6\u4eec\u7684\u901f\u5ea6\u63d0\u9ad8\u4e86 20%\u3002\u4ed6\u4eec\u7684\u8ba4\u77e5\u4e0e\u73b0\u5b9e\u5b8c\u5168\u98a0\u5012\u4e86\u3002.<\/p>\n<p>\u4e94\u4e2a\u56e0\u7d20\u89e3\u91ca\u4e86\u7ecf\u6d4e\u653e\u7f13\u7684\u539f\u56e0\uff1a<\/p>\n<ol>\n<li><strong>\u8c03\u8bd5\u5f00\u9500\u3002.<\/strong> \u8282\u7701\u4e86\u7ea0\u6b63\u9519\u8bef\u7684\u65f6\u95f4\u3002.<\/li>\n<li><strong>\u4e0a\u4e0b\u6587\u5207\u6362\u6210\u672c\u3002.<\/strong> \u5728\u4eba\u5de5\u667a\u80fd\u7684\u5efa\u8bae\u548c\u5b9e\u9645\u5de5\u4f5c\u4e4b\u95f4\u5207\u6362\u4f1a\u4ea7\u751f\u8ba4\u77e5\u8d1f\u8377\u3002.<\/li>\n<li><strong>\u5b66\u4e60\u66f2\u7ebf\u3002.<\/strong> \u5373\u4f7f\u662f\u7ecf\u9a8c\u4e30\u5bcc\u7684\u4e13\u4e1a\u4eba\u58eb\uff0c\u4e5f\u8981\u82b1\u8d39\u65f6\u95f4\u6478\u7d22\u5982\u4f55\u6709\u6548\u4f7f\u7528\u4eba\u5de5\u667a\u80fd\u3002.<\/li>\n<li><strong>\u9690\u85cf\u7684\u8d28\u91cf\u8981\u6c42\u3002.<\/strong> \u4eba\u5de5\u667a\u80fd\u8f93\u51fa\u9700\u8981\u6587\u6863\u3001\u9a8c\u8bc1\u548c\u683c\u5f0f\u5316\uff0c\u800c\u8fd9\u4e9b\u7cfb\u7edf\u90fd\u65e0\u6cd5\u63d0\u4f9b\u3002.<\/li>\n<li><strong>\u73b0\u5b9e\u4e16\u754c\u7684\u6807\u51c6\u3002.<\/strong> \u6210\u719f\u7684\u7ec4\u7ec7\u90fd\u6709\u4eba\u5de5\u667a\u80fd\u751f\u6210\u7684\u5de5\u4f5c\u7cfb\u7edf\u65e0\u6cd5\u8fbe\u5230\u7684\u8d28\u91cf\u6807\u51c6\u3002.<\/li>\n<\/ol>\n<p>\u8fd9\u4e0d\u662f\u4e00\u4e2a\u5173\u4e8e\u4eba\u5de5\u667a\u80fd\u4e00\u65e0\u662f\u5904\u7684\u6545\u4e8b\u3002\u8fd9\u662f\u4e00\u4e2a\u5173\u4e8e\u4f9b\u5e94\u5546\u6f14\u793a\u4e0e\u7ec4\u7ec7\u73b0\u5b9e\u4e4b\u95f4\u5dee\u8ddd\u7684\u6545\u4e8b\u3002\u751f\u4ea7\u7387\u6096\u8bba\u8868\u660e\uff0c\u5bf9\u4e8e\u6210\u719f\u6d41\u7a0b\u4e2d\u7ecf\u9a8c\u4e30\u5bcc\u7684\u4e13\u4e1a\u4eba\u5458\u6765\u8bf4\uff0c\u5f53\u524d\u7684\u4eba\u5de5\u667a\u80fd\u8f85\u52a9\u53ef\u80fd\u4f1a\u4ea7\u751f\u66f4\u591a\u7684\u5de5\u4f5c\uff0c\u800c\u4e0d\u662f\u8282\u7701\u66f4\u591a\u7684\u5de5\u4f5c\u3002.<\/p>\n<h2>\u4eba\u5de5\u667a\u80fd\u7cfb\u7edf\u4e3a\u4f55\u4f1a\u968f\u65f6\u95f4\u9000\u5316<\/h2>\n<p>\u4e3a\u4ec0\u4e48\u4eba\u5de5\u667a\u80fd\u7cfb\u7edf\u96be\u4ee5\u5b8c\u6210\u8f83\u957f\u7684\u4efb\u52a1\uff1f\u7b54\u6848\u662f\u7ed3\u6784\u6027\u7684\uff0c\u4f46\u5176\u5f71\u54cd\u5374\u662f\u6218\u7565\u6027\u7684\u3002.<\/p>\n<p>\u628a\u4eba\u5de5\u667a\u80fd\u60f3\u8c61\u6210\u4e00\u4e2a\u5de5\u4f5c\u8bb0\u5fc6--\u7528\u4e8e\u4fdd\u5b58\u5f53\u524d\u4efb\u52a1\u4fe1\u606f\u7684\u6709\u9650\u5bb9\u91cf\u3002\u968f\u7740\u5bb9\u91cf\u7684\u589e\u52a0\uff0c\u6027\u80fd\u4e5f\u4f1a\u4e0b\u964d\u3002\u7cfb\u7edf\u4f1a\u5fd8\u8bb0\u4e4b\u524d\u7684\u51b3\u5b9a\u3002\u5b83\u4f1a\u81ea\u76f8\u77db\u76fe\u3002\u5b83\u5fd8\u8bb0\u4e86\u81ea\u5df1\u60f3\u8981\u5b8c\u6210\u7684\u4efb\u52a1\u3002.<\/p>\n<p>\u4f7f\u7528\u8fc7\u4eba\u5de5\u667a\u80fd\u52a9\u624b\u8d85\u8fc7 30 \u5206\u949f\u7684\u4eba\u90fd\u6709\u5207\u8eab\u4f53\u4f1a\u3002\u7cfb\u7edf\u5f00\u59cb\u65f6\u5f88\u654f\u9510\u3002\u4e00\u4e2a\u5c0f\u65f6\u540e\uff0c\u5b83\u5c31\u4f1a\u5fd8\u8bb0\u4e8c\u5341\u5206\u949f\u524d\u505a\u51fa\u7684\u51b3\u5b9a\u3002\u5b83\u5f15\u5165\u4e86\u672c\u53ef\u4ee5\u66f4\u65e9\u53d1\u73b0\u7684\u9519\u8bef\u3002\u5b83\u5931\u53bb\u4e86\u4e3b\u7ebf\u3002.<\/p>\n<p>\u7814\u7a76\u5df2\u7ecf\u5bf9\u6b64\u8fdb\u884c\u4e86\u91cf\u5316\u3002\u968f\u7740\u4eba\u5de5\u667a\u80fd\u9700\u8981\u8ffd\u8e2a\u7684\u4fe1\u606f\u91cf\u589e\u52a0\uff0c\u51c6\u786e\u7387\u53ef\u80fd\u4f1a\u4e0b\u964d 20-30 \u4e2a\u767e\u5206\u70b9\u3002\u4e00\u9879\u7814\u7a76\u663e\u793a\uff0c\u5728\u590d\u6742\u7684\u63a8\u7406\u4efb\u52a1\u4e2d\uff0c\u968f\u7740\u4efb\u52a1\u590d\u6742\u5ea6\u7684\u589e\u52a0\uff0c\u51c6\u786e\u7387\u4ece 82% \u4e0b\u964d\u5230 22%\u3002.<\/p>\n<p>\u6570\u5b66\u662f\u65e0\u60c5\u7684\u3002\u5fae\u5c0f\u7684\u9519\u8bef\u4e5f\u4f1a\u9020\u6210\u4e25\u91cd\u540e\u679c\u3002\u5982\u679c\u6bcf\u6b21\u7cfb\u7edf\u5904\u7406\u65b0\u4fe1\u606f\u65f6\u90fd\u6709 1% \u7684\u51e0\u7387\u4e22\u5931\u4e00\u4e2a\u5173\u952e\u7ec6\u8282\uff0c\u90a3\u4e48\u5728 100 \u6b21\u4ea4\u4e92\u4e4b\u540e\uff0c\u4fdd\u7559\u8be5\u7ec6\u8282\u7684\u51e0\u7387\u5c31\u4f1a\u4e0b\u964d\u5230 37%\u3002.<\/p>\n<p>\u8fd9\u4e0d\u662f\u4efb\u4f55\u7279\u5b9a\u4ea7\u54c1\u7684\u9519\u8bef\u3002\u8fd9\u662f\u5f53\u524d\u4eba\u5de5\u667a\u80fd\u7cfb\u7edf\u5de5\u4f5c\u65b9\u5f0f\u7684\u56fa\u6709\u9650\u5236\u3002\u800c\u4e14\uff0c\u8fd9\u4e5f\u4e3a\u5b83\u4eec\u80fd\u62e5\u6709\u7684\u4e1c\u897f\u548c\u53ea\u80fd\u63d0\u4f9b\u534f\u52a9\u7684\u4e1c\u897f\u8bbe\u5b9a\u4e86\u786c\u6027\u4e0a\u9650\u3002.<\/p>\n<h2>\u5e76\u975e\u6240\u6709\u4efb\u52a1\u90fd\u4e00\u6837<\/h2>\n<p>METR \u7684\u7814\u7a76\u63ed\u793a\u4e86\u53e6\u4e00\u4e2a\u590d\u6742\u95ee\u9898\uff1a\u4e0d\u540c\u9886\u57df\u7684\u4eba\u5de5\u667a\u80fd\u80fd\u529b\u5dee\u5f02\u5de8\u5927\u3002.<\/p>\n<p><strong>\u5206\u6790\u548c\u7ed3\u6784\u5316\u4efb\u52a1<\/strong> \u5982 data \u7684\u5206\u6790\u548c\u62a5\u544a\u751f\u6210\u663e\u793a\u51fa\u5f3a\u5927\u7684\u80fd\u529b--\u7cfb\u7edf\u53ef\u4ee5\u5904\u7406\u4eba\u7c7b\u9700\u8981\u4e00\u5230\u4e09\u4e2a\u5c0f\u65f6\u624d\u80fd\u5b8c\u6210\u7684\u5de5\u4f5c\u3002.<\/p>\n<p><strong>\u9700\u8981\u4e0e\u5916\u90e8\u7cfb\u7edf\u4e92\u52a8\u7684\u4efb\u52a1<\/strong>-\u6d4f\u89c8\u3001\u8de8\u5e73\u53f0\u534f\u8c03\u3001\u5904\u7406\u89c6\u89c9\u4fe1\u606f\u7684\u80fd\u529b\u8981\u4f4e 40-100 \u500d\u3002\u5b83\u4eec\u7684\u6210\u719f\u5ea6\u5927\u7ea6\u843d\u540e\u4e24\u5e74\u3002.<\/p>\n<p><strong>\u7269\u7406\u4e16\u754c\u7684\u5e94\u7528<\/strong> \u5982\u81ea\u52a8\u9a7e\u9a76\u6c7d\u8f66\u7684\u6539\u8fdb\u901f\u5ea6\u8981\u6bd4\u6570\u5b57\u4efb\u52a1\u6162\u5f97\u591a\u3002.<\/p>\n<p>\u8fd9\u610f\u5473\u7740\u4ec0\u4e48\uff1f\u201c\u4eba\u5de5\u667a\u80fd\u53ef\u4ee5\u5de5\u4f5c\u6570\u5c0f\u65f6 \u201d\u662f\u9488\u5bf9\u7279\u5b9a\u9886\u57df\u7684\u3002\u4e00\u4e2a\u80fd\u5904\u7406\u4e24\u5c0f\u65f6\u5206\u6790\u4efb\u52a1\u7684\u7cfb\u7edf\uff0c\u5728\u5904\u7406\u4e8c\u5341\u5206\u949f\u534f\u8c03\u4efb\u52a1\u65f6\u53ef\u80fd\u4f1a\u5f88\u5403\u529b\u3002\u80fd\u529b\u66f2\u7ebf\u5e76\u4e0d\u4e00\u81f4\u3002.<\/p>\n<p>\u5bf9\u4e8e\u4f01\u4e1a\u6765\u8bf4\uff0c\u8fd9\u610f\u5473\u7740\u8981\u6839\u636e\u4efb\u52a1\u7279\u70b9\u8c28\u614e\u5339\u914d\u4eba\u5de5\u667a\u80fd\u7684\u90e8\u7f72\u3002\u7092\u4f5c\u5468\u671f\u5c06\u4eba\u5de5\u667a\u80fd\u89c6\u4e3a\u901a\u7528\u89e3\u51b3\u65b9\u6848\u3002\u800c\u73b0\u5b9e\u662f\u975e\u5e38\u5177\u4f53\u7684\u3002.<\/p>\n<h2>\u8d8b\u52bf\u7684\u5b9e\u9645\u610f\u4e49<\/h2>\n<p>METR \u5c06\u4ed6\u4eec\u7684\u53d1\u73b0\u63cf\u8ff0\u4e3a\u6f5c\u5728\u7684 \u201c\u4eba\u7c7b\u5386\u53f2\u4e0a\u6700\u91cd\u8981\u7684\u8d8b\u52bf\u4e4b\u4e00\u201d\u3002\u8fd9\u53ef\u80fd\u6709\u70b9\u8a00\u8fc7\u5176\u5b9e\u3002\u4f46\u8fd9\u4e00\u8d8b\u52bf\u5df2\u7ecf\u6301\u7eed\u4e86\u516d\u5e74\uff0c\u800c\u4e14\u65b9\u5411\u662f\u660e\u786e\u7684\u3002.<\/p>\n<p>\u5f53\u524d\u7684\u4eba\u5de5\u667a\u80fd\u5728\u5b8c\u6210\u4eba\u7c7b\u8017\u65f6\u4e0d\u5230\u56db\u5206\u949f\u7684\u4efb\u52a1\u65f6\uff0c\u51e0\u4e4e\u53ef\u4ee5\u53d6\u5f97\u5b8c\u7f8e\u7684\u6210\u529f\u3002\u800c\u5728\u8017\u65f6\u8d85\u8fc7\u56db\u5c0f\u65f6\u7684\u4efb\u52a1\u4e0a\uff0c\u5b83\u7684\u6210\u529f\u7387\u5219\u4f4e\u4e8e 10%\u3002\u6709\u8da3\u7684\u533a\u57df--\u4e5f\u662f\u5bf9\u7ec4\u7ec7\u51b3\u7b56\u81f3\u5173\u91cd\u8981\u7684\u533a\u57df--\u4ecb\u4e8e\u4e24\u8005\u4e4b\u95f4\u3002.<\/p>\n<p>\u6211\u7684\u771f\u5b9e\u8bc4\u4ef7\u662f\uff1a\u6211\u4eec\u7684\u4eba\u5de5\u667a\u80fd\u53ef\u4ee5\u5728\u7279\u5b9a\u9886\u57df\u7684\u660e\u786e\u4efb\u52a1\u4e0a\u6301\u7eed\u5de5\u4f5c\u4e00\u5230\u4e24\u4e2a\u5c0f\u65f6\u3002\u8fd9\u4e9b\u7cfb\u7edf\u65e0\u6cd5\u53ef\u9760\u5730\u505a\u5230\u8fd9\u4e00\u70b9\u3002\u5b83\u4eec\u65e0\u6cd5\u5728\u6ca1\u6709\u4eba\u7c7b\u76d1\u7763\u7684\u60c5\u51b5\u4e0b\u505a\u5230\u8fd9\u4e00\u70b9\u3002\u5982\u679c\u4e0d\u8fdb\u884c\u6e05\u7406\uff0c\u5b83\u4eec\u4e5f\u65e0\u6cd5\u8fbe\u5230\u771f\u6b63\u7684\u7ec4\u7ec7\u8d28\u91cf\u6807\u51c6\u3002.<\/p>\n<p>\u4f46\u53d1\u5c55\u8f68\u8ff9\u8868\u660e\uff0c\u8fd9\u4e9b\u9650\u5236\u53ef\u80fd\u4e0d\u662f\u6c38\u4e45\u6027\u7684\u3002\u6bcf\u5bb6\u5927\u578b\u4eba\u5de5\u667a\u80fd\u516c\u53f8\u90fd\u4f1a\u5bf9\u6301\u7eed\u8fd0\u884c\u8fdb\u884c\u4f18\u5316\u3002\u6bcf\u4e00\u4e2a\u8ba4\u771f\u7684\u91c7\u7528\u8005\u90fd\u4f1a\u9488\u5bf9\u5f53\u524d\u7684\u5c40\u9650\u6027\u5236\u5b9a\u53d8\u901a\u65b9\u6cd5\u3002\u6bcf\u4e2a\u4f01\u4e1a\u5e73\u53f0\u90fd\u4f1a\u4e3a\u66f4\u957f\u65f6\u95f4\u7684\u4eba\u5de5\u667a\u80fd\u5de5\u4f5c\u589e\u52a0\u57fa\u7840\u8bbe\u65bd\u3002.<\/p>\n<p>\u95ee\u9898\u5e76\u4e0d\u5728\u4e8e\u4eba\u5de5\u667a\u80fd\u6700\u7ec8\u80fd\u5426\u5b9e\u73b0\u5c0f\u65f6\u5de5\u4f5c\u3002\u95ee\u9898\u662f\u4ec0\u4e48\u65f6\u5019\uff0c\u4ee5\u53ca\u76ee\u524d\u8fd9\u4e00\u4ee3\u89e3\u51b3\u65b9\u6848\u80fd\u5426\u5b9e\u73b0\u8fd9\u4e00\u76ee\u6807\uff0c\u8fd8\u662f\u6211\u4eec\u4ecd\u5728\u7b49\u5f85\u5c1a\u672a\u5b9e\u73b0\u7684\u7a81\u7834\u3002.<\/p>\n<p><em>\u5728 <a href=\"https:\/\/www.artefact.com\/zh\/blog\/long-run-ai-agents-part-2-three-approaches-that-actually-work\/\">\u7b2c\u4e8c\u90e8\u5206<\/a>, \u6211\u4eec\u5c06\u7814\u7a76\u5ef6\u957f\u4eba\u5de5\u667a\u80fd\u5de5\u4f5c\u65f6\u95f4\u7684\u4e09\u79cd\u65b9\u6cd5\uff1a\u65b0\u542f\u52a8\u5faa\u73af\u3001\u9009\u62e9\u6027\u8bb0\u5fc6\u548c\u56e2\u961f\u534f\u8c03\u3002.<\/em><\/p>\n<p>&nbsp;<\/p>\n<h3>\u53c2\u8003\u8d44\u6599<\/h3>\n<p><span style=\"text-decoration: underline;\">\u7814\u7a76\u8bba\u6587<\/span><\/p>\n<ul>\n<li>\u5229\u7528\u53ef\u6269\u5c55\u7684\u957f\u671f\u5b58\u50a8\u5668\u6784\u5efa\u53ef\u6295\u5165\u751f\u4ea7\u7684\u4eba\u5de5\u667a\u80fd <a href=\"https:\/\/arxiv.org\/abs\/2504.19413\" target=\"_blank\" rel=\"noopener\">arxiv.org\/abs\/2504.19413<\/a><\/li>\n<li>\u57fa\u4e8e\u591a\u56fe\u7684\u4eba\u5de5\u667a\u80fd\u5185\u5b58\u67b6\u6784 <a href=\"http:\/\/arxiv.org\/abs\/2601.03236\" target=\"_blank\" rel=\"noopener\">arxiv.org\/abs\/2601.03236<\/a><\/li>\n<li>\u8861\u91cf\u751f\u4ea7\u4e2d\u7684\u4eba\u5de5\u667a\u80fd <a href=\"http:\/\/arxiv.org\/abs\/2512.04123\" target=\"_blank\" rel=\"noopener\">arxiv.org\/abs\/2512.04123<\/a><\/li>\n<\/ul>\n<p><span style=\"text-decoration: underline;\">\u884c\u4e1a\u62a5\u544a\u548c\u767d\u76ae\u4e66<\/span><\/p>\n<ul>\n<li>\u4eba\u5de5\u667a\u80fd\u7cfb\u7edf\u7684\u6545\u969c\u6a21\u5f0f - \u5fae\u8f6f<\/li>\n<li>2025 \u5e74\u4eba\u5de5\u667a\u80fd\u4e0e\u4fe1\u4efb\u7684\u7ecf\u9a8c\u6559\u8bad - \u8c37\u6b4c\u4e91<\/li>\n<li>\u4eba\u5de5\u667a\u80fd\u5de5\u7a0b\u73b0\u72b6 - LangChain<\/li>\n<li>\u57fa\u51c6\u4e0e\u771f\u5b9e\u4e16\u754c\u8bc4\u4f30 - METR<\/li>\n<\/ul>\n<p><span style=\"text-decoration: underline;\">\u6280\u672f\u6587\u6863<\/span><\/p>\n<ul>\n<li>\u6211\u4eec\u5982\u4f55\u6784\u5efa\u591a\u4ee3\u7406\u7814\u7a76\u7cfb\u7edf - \u4eba\u7c7b\u5b66<\/li>\n<li>\u6a21\u578b\u4e0a\u4e0b\u6587\u534f\u8bae\u89c4\u8303 - modelcontextprotocol.io<\/li>\n<li>Fresh-Start Cycling Documentation\uff08\u201cRalph Wiggum\u201d\uff09- Geoffrey Huntley (<a href=\"http:\/\/ghuntley.com\/ralph\/\" target=\"_blank\" rel=\"noopener\">ghuntley.com\/ralph\/<\/a>)<\/li>\n<\/ul>","protected":false},"excerpt":{"rendered":"<p>2025 \u5e74 3 \u6708\uff0c\u4e00\u5bb6\u540d\u4e3a METR \u7684\u7814\u7a76\u673a\u6784\u53d1\u8868\u4e86\u4e00\u9879\u7814\u7a76\u6210\u679c\uff0c\u4f46\u5374\u6ca1\u6709\u5f97\u5230\u5e94\u6709\u7684\u5173\u6ce8\u3002\u4ed6\u4eec\u4e00\u76f4\u5728\u6d4b\u91cf\u4e00\u4e9b\u4e0d\u5408\u65f6\u5b9c\u7684\u4e1c\u897f\uff1aAI \u7cfb\u7edf\u5728\u5b8c\u6210\u4efb\u52a1\u591a\u957f\u65f6\u95f4\u540e\u624d\u4f1a\u51fa\u73b0\u6545\u969c\u3002\u800c\u4e0d\u662f\u5b83\u4eec\u5728\u4e00\u6b21\u4ea4\u4e92\u4e2d\u80fd\u505a\u4ec0\u4e48\u3002METR \u60f3\u77e5\u9053\u7684\u662f\uff0c\u8fd9\u4e9b\u7cfb\u7edf\u80fd\u5728\u591a\u957f\u65f6\u95f4\u5185\u4fdd\u6301\u8fde\u8d2f\u3001\u6709\u7528\u7684\u5de5\u4f5c\u3002.<\/p>","protected":false},"featured_media":1086605,"parent":0,"template":"","meta":{"_acf_changed":false,"ep_exclude_from_search":false},"blog-category":[2995],"blog-language":[2991],"class_list":["post-1083983","blog","type-blog","status-publish","has-post-thumbnail","hentry","blog-category-ai-technology","blog-language-en"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.artefact.com\/zh\/wp-json\/wp\/v2\/blog\/1083983","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.artefact.com\/zh\/wp-json\/wp\/v2\/blog"}],"about":[{"href":"https:\/\/www.artefact.com\/zh\/wp-json\/wp\/v2\/types\/blog"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.artefact.com\/zh\/wp-json\/wp\/v2\/media\/1086605"}],"wp:attachment":[{"href":"https:\/\/www.artefact.com\/zh\/wp-json\/wp\/v2\/media?parent=1083983"}],"wp:term":[{"taxonomy":"blog-category","embeddable":true,"href":"https:\/\/www.artefact.com\/zh\/wp-json\/wp\/v2\/blog-category?post=1083983"},{"taxonomy":"blog-language","embeddable":true,"href":"https:\/\/www.artefact.com\/zh\/wp-json\/wp\/v2\/blog-language?post=1083983"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}