{"id":16742,"date":"2026-08-11T23:11:10","date_gmt":"2026-08-11T13:11:10","guid":{"rendered":"https:\/\/geoffreychen.com\/?p=16742"},"modified":"2026-08-11T23:17:55","modified_gmt":"2026-08-11T13:17:55","slug":"prompt-writing-is-not-the-most-important-ai-skill","status":"publish","type":"post","link":"https:\/\/geoffreychen.com\/zh\/prompt-writing-is-not-the-most-important-ai-skill\/","title":{"rendered":"\u4e3a\u4ec0\u4e48\u201c\u4f1a\u5199\u63d0\u793a\u8bcd\u201d\u4e0d\u662f\u6700\u91cd\u8981\u7684AI\u80fd\u529b\uff1f"},"content":{"rendered":"<\/p>\n<p><em>\u300aAI\u8fdb\u5165\u5de5\u4f5c\u4e4b\u540e\u300b\u00b7 \u7b2c\u4e00\u5b63\u201c\u4ece\u56de\u7b54\u95ee\u9898\u5230\u53c2\u4e0e\u5de5\u4f5c\u201d\u00b7 \u7b2c\u4e8c\u7bc7<\/em><\/p>\n<p>\u8bbe\u60f3\u4e24\u4e2a\u4eba\u90fd\u8ba9AI\u8d77\u8349\u4e00\u4efd\u9879\u76ee\u98ce\u9669\u62a5\u544a\u3002<\/p>\n<p>\u7b2c\u4e00\u4e2a\u4eba\u82b1\u4e86\u5f88\u957f\u65f6\u95f4\u6da6\u8272\u63d0\u793a\u8bcd\uff1a\u201c\u4f60\u662f\u4e00\u540d\u5177\u6709\u4e8c\u5341\u5e74\u7ecf\u9a8c\u7684\u9ad8\u7ea7\u98ce\u9669\u987e\u95ee\uff0c\u8bf7\u4ee5\u4e25\u8c28\u3001\u4e13\u4e1a\u3001\u7ed3\u6784\u6e05\u695a\u7684\u65b9\u5f0f\u5206\u6790\u8fd9\u4e2a\u9879\u76ee\uff0c\u5217\u51fa\u4e3b\u8981\u98ce\u9669\u5e76\u63d0\u51fa\u5efa\u8bae\u3002\u201d\u8fd9\u6bb5\u8bdd\u542c\u8d77\u6765\u5f88\u5b8c\u6574\uff0c\u4f46\u4ed6\u6ca1\u6709\u63d0\u4f9b\u9879\u76ee\u9884\u7b97\u3001\u65f6\u95f4\u8868\u3001\u5408\u540c\u3001\u6280\u672f\u4f9d\u8d56\u3001\u5df2\u7ecf\u53d1\u751f\u7684\u95ee\u9898\u548c\u516c\u53f8\u7684\u98ce\u9669\u6807\u51c6\u3002<\/p>\n<p>\u7b2c\u4e8c\u4e2a\u4eba\u7684\u6307\u4ee4\u53ea\u6709\u51e0\u884c\uff0c\u5374\u9644\u4e0a\u4e86\u5f53\u524d\u9879\u76ee\u8ba1\u5212\u3001\u98ce\u9669\u767b\u8bb0\u8868\u3001\u4e0a\u6b21\u4f1a\u8bae\u7eaa\u8981\u548c\u62a5\u544a\u6a21\u677f\u3002\u4ed6\u660e\u786e\u8bf4\u660e\u54ea\u4e9b\u6587\u4ef6\u662f\u6700\u65b0\u7248\u672c\uff0c\u54ea\u4e9b\u6570\u636e\u4ecd\u672a\u786e\u8ba4\uff0c\u8981\u6c42AI\u53ea\u6839\u636e\u7ed9\u5b9a\u6750\u6599\u63d0\u51fa\u98ce\u9669\uff0c\u5e76\u628a\u6bcf\u9879\u5224\u65ad\u94fe\u63a5\u5230\u5177\u4f53\u8bc1\u636e\uff1b\u8bc1\u636e\u4e0d\u8db3\u7684\u5730\u65b9\u5fc5\u987b\u6807\u4e3a\u5f85\u6838\u67e5\u3002<\/p>\n<p>\u901a\u5e38\uff0c\u7b2c\u4e8c\u79cd\u505a\u6cd5\u66f4\u53ef\u80fd\u4ea7\u751f\u53ef\u7528\u7ed3\u679c\u3002\u539f\u56e0\u5e76\u4e0d\u795e\u79d8\uff1a\u7b2c\u4e00\u6bb5\u63d0\u793a\u8bcd\u4e3b\u8981\u89c4\u5b9a\u4e86\u8bed\u6c14\u548c\u89d2\u8272\uff0c\u7b2c\u4e8c\u79cd\u505a\u6cd5\u63d0\u4f9b\u4e86\u5b8c\u6210\u4efb\u52a1\u6240\u9700\u7684\u73b0\u5b9e\u3002<\/p>\n<p>\u63d0\u793a\u8bcd\u5f53\u7136\u91cd\u8981\u3002\u542b\u7cca\u6307\u4ee4\u4f1a\u9020\u6210\u542b\u7cca\u8f93\u51fa\uff0c\u76ee\u6807\u3001\u683c\u5f0f\u548c\u9650\u5236\u4e5f\u9700\u8981\u88ab\u6e05\u695a\u8868\u8fbe\u3002\u4f46\u5f53AI\u8fdb\u5165\u771f\u5b9e\u5de5\u4f5c\u4ee5\u540e\uff0c\u6700\u91cd\u8981\u7684\u80fd\u529b\u5df2\u7ecf\u4e0d\u518d\u662f\u201c\u600e\u6837\u628a\u4e00\u53e5\u8bdd\u5199\u5f97\u66f4\u806a\u660e\u201d\uff0c\u800c\u662f\u600e\u6837\u628a\u4efb\u52a1\u3001\u6750\u6599\u3001\u89c4\u5219\u3001\u5de5\u5177\u548c\u5b8c\u6210\u6807\u51c6\u7ec4\u7ec7\u6210\u4e00\u4e2a\u6a21\u578b\u80fd\u591f\u6b63\u786e\u53c2\u4e0e\u7684\u73af\u5883\u3002<\/p>\n<h2>\u63d0\u793a\u8bcd\u89e3\u51b3\u7684\u662f\u6307\u4ee4\u95ee\u9898<\/h2>\n<p>\u63d0\u793a\u8bcd\u9996\u5148\u56de\u7b54\uff1a\u201c\u73b0\u5728\u5e0c\u671b\u6a21\u578b\u505a\u4ec0\u4e48\uff1f\u201d<\/p>\n<p>\u4e00\u4e2a\u6709\u6548\u63d0\u793a\u901a\u5e38\u9700\u8981\u5305\u542b\u4efb\u52a1\u3001\u5bf9\u8c61\u3001\u5fc5\u8981\u80cc\u666f\u3001\u8f93\u51fa\u5f62\u5f0f\u548c\u9650\u5236\u3002\u4f8b\u5982\uff0c\u201c\u603b\u7ed3\u8fd9\u4efd\u6587\u4ef6\u201d\u8fdc\u4e0d\u5982\u201c\u4e3a\u4e0d\u719f\u6089\u6280\u672f\u7684\u8463\u4e8b\u4f1a\u6210\u5458\uff0c\u7528\u4e94\u70b9\u6982\u62ec\u8fd9\u4efd\u5b89\u5168\u62a5\u544a\uff1b\u4fdd\u7559\u6240\u6709\u98ce\u9669\u7b49\u7ea7\u548c\u65e5\u671f\uff0c\u4e0d\u589e\u52a0\u6587\u4ef6\u4e2d\u6ca1\u6709\u7684\u5224\u65ad\u201d\u660e\u786e\u3002<\/p>\n<p>\u8fd9\u4e9b\u5dee\u5f02\u662f\u771f\u5b9e\u7684\u3002\u597d\u7684\u8868\u8fbe\u80fd\u591f\u51cf\u5c11\u6b67\u4e49\uff0c\u544a\u8bc9\u6a21\u578b\u5173\u6ce8\u4ec0\u4e48\u3001\u5ffd\u7565\u4ec0\u4e48\uff0c\u5e76\u4f7f\u8f93\u51fa\u66f4\u5bb9\u6613\u68c0\u67e5\u3002\u5bf9\u4e8e\u4e00\u6b21\u6027\u7684\u7ffb\u8bd1\u3001\u6539\u5199\u3001\u683c\u5f0f\u8f6c\u6362\u6216\u5934\u8111\u98ce\u66b4\uff0c\u63d0\u793a\u8bcd\u751a\u81f3\u53ef\u80fd\u662f\u5f71\u54cd\u7ed3\u679c\u7684\u4e3b\u8981\u53d8\u91cf\u3002<\/p>\n<p>\u95ee\u9898\u5728\u4e8e\uff0c\u771f\u5b9e\u5de5\u4f5c\u5f88\u5c11\u53ea\u7531\u4e00\u4e2a\u8f93\u5165\u548c\u4e00\u4e2a\u8f93\u51fa\u6784\u6210\u3002<\/p>\n<p>\u4e00\u4efd\u5ba2\u6237\u7b54\u590d\u9700\u8981\u5ba2\u6237\u5386\u53f2\u3001\u5f53\u524d\u5408\u540c\u3001\u9000\u6b3e\u653f\u7b56\u548c\u6388\u6743\u8303\u56f4\uff1b\u4e00\u6b21\u4ee3\u7801\u4fee\u6539\u9700\u8981\u9879\u76ee\u7ed3\u6784\u3001\u4f9d\u8d56\u3001\u6d4b\u8bd5\u3001\u90e8\u7f72\u73af\u5883\u548c\u5df2\u6709\u7ea6\u5b9a\uff1b\u4e00\u7bc7\u7814\u7a76\u6587\u7ae0\u9700\u8981\u53ef\u6838\u67e5\u8d44\u6599\u3001\u9002\u7528\u65e5\u671f\u3001\u8bc1\u636e\u7b49\u7ea7\u548c\u5f15\u7528\u89c4\u5219\u3002\u5373\u4f7f\u63d0\u793a\u8bcd\u5199\u5f97\u975e\u5e38\u4f18\u7f8e\uff0c\u53ea\u8981\u8fd9\u4e9b\u73b0\u5b9e\u6ca1\u6709\u8fdb\u5165\u6a21\u578b\u53ef\u4ee5\u4f7f\u7528\u7684\u73af\u5883\uff0c\u6a21\u578b\u4ecd\u53ea\u80fd\u5728\u7f3a\u5931\u4fe1\u606f\u4e0a\u4f5c\u51fa\u8bed\u8a00\u4e0a\u5408\u7406\u7684\u8865\u5168\u3002<\/p>\n<p>\u628a\u8fd9\u79cd\u5931\u8d25\u5f52\u7ed3\u4e3a\u201c\u63d0\u793a\u8bcd\u5199\u5f97\u4e0d\u591f\u597d\u201d\uff0c\u4f1a\u63a9\u76d6\u771f\u6b63\u7684\u95ee\u9898\uff1a\u7cfb\u7edf\u6ca1\u6709\u83b7\u5f97\u8db3\u591f\u3001\u6b63\u786e\u3001\u53ca\u65f6\u5e76\u53ef\u8ffd\u6eaf\u7684\u4e0a\u4e0b\u6587\u3002<\/p>\n<h2>\u4ece\u63d0\u793a\u8bcd\u8f6c\u5411\u4e0a\u4e0b\u6587\u5de5\u7a0b<\/h2>\n<p>\u8fd1\u5e74\u6765\uff0cAI\u7cfb\u7edf\u8bbe\u8ba1\u8d8a\u6765\u8d8a\u91cd\u89c6\u201c\u4e0a\u4e0b\u6587\u5de5\u7a0b\u201d\u3002\u8fd9\u4e2a\u8bcd\u4e0d\u662f\u63d0\u793a\u8bcd\u7684\u65f6\u9ae6\u66ff\u4ee3\uff0c\u800c\u662f\u628a\u5173\u6ce8\u8303\u56f4\u4ece\u4e00\u53e5\u6307\u4ee4\u6269\u5927\u5230\u6a21\u578b\u5728\u4e00\u6b21\u4efb\u52a1\u4e2d\u5b9e\u9645\u80fd\u591f\u4f7f\u7528\u7684\u5168\u90e8\u4fe1\u606f\u3002<\/p>\n<p>\u4e0a\u4e0b\u6587\u53ef\u80fd\u5305\u62ec\uff1a<\/p>\n<ul>\n<li>\u7cfb\u7edf\u89c4\u5219\u548c\u7528\u6237\u6307\u4ee4\uff1b<\/li>\n<li>\u5f53\u524d\u5bf9\u8bdd\u4e0e\u4efb\u52a1\u72b6\u6001\uff1b<\/li>\n<li>\u6587\u4ef6\u3001\u6570\u636e\u5e93\u8bb0\u5f55\u548c\u68c0\u7d22\u7ed3\u679c\uff1b<\/li>\n<li>\u53ef\u8c03\u7528\u5de5\u5177\u53ca\u5176\u8fd4\u56de\u503c\uff1b<\/li>\n<li>\u793a\u4f8b\u3001\u6a21\u677f\u3001\u4ee3\u7801\u89c4\u8303\u548c\u5b8c\u6210\u6807\u51c6\uff1b<\/li>\n<li>\u8fc7\u53bb\u6b65\u9aa4\u7559\u4e0b\u7684\u6458\u8981\u3001\u8ba1\u5212\u548c\u9519\u8bef\u8bb0\u5f55\u3002<\/li>\n<\/ul>\n<p>Anthropic\u5728\u5173\u4e8e\u4ee3\u7406\u4e0a\u4e0b\u6587\u5de5\u7a0b\u7684\u6280\u672f\u8bf4\u660e\u4e2d\uff0c\u628a\u4e0a\u4e0b\u6587\u89c6\u4e3a\u6709\u9650\u800c\u73cd\u8d35\u7684\u8d44\u6e90\u3002\u7cfb\u7edf\u9700\u8981\u5728\u9700\u8981\u65f6\u68c0\u7d22\u76f8\u5173\u6750\u6599\u3001\u538b\u7f29\u5df2\u7ecf\u5b8c\u6210\u7684\u8fc7\u7a0b\uff0c\u5e76\u907f\u514d\u8ba9\u8fc7\u91cf\u6216\u9648\u65e7\u4fe1\u606f\u6df9\u6ca1\u771f\u6b63\u91cd\u8981\u7684\u5185\u5bb9\u3002<a href=\"https:\/\/www.anthropic.com\/engineering\/effective-context-engineering-for-ai-agents\" target=\"_blank\" rel=\"noopener\">Anthropic\uff1a\u300aEffective context engineering for AI agents\u300b<\/a><\/p>\n<p>\u8fd9\u8bf4\u660e\u201c\u628a\u6240\u6709\u8d44\u6599\u5168\u90e8\u8d34\u8fdb\u53bb\u201d\u4e5f\u4e0d\u662f\u89e3\u51b3\u529e\u6cd5\u3002<\/p>\n<p>\u4e00\u9879\u5173\u4e8e\u957f\u4e0a\u4e0b\u6587\u6a21\u578b\u7684\u7814\u7a76\u53d1\u73b0\uff0c\u5f53\u76f8\u5173\u4fe1\u606f\u4f4d\u4e8e\u5f88\u957f\u8f93\u5165\u7684\u4e2d\u90e8\u65f6\uff0c\u4e00\u4e9b\u6a21\u578b\u5229\u7528\u5b83\u7684\u80fd\u529b\u4f1a\u4e0b\u964d\uff1b\u4fe1\u606f\u653e\u5728\u5f00\u5934\u6216\u7ed3\u5c3e\u65f6\uff0c\u8868\u73b0\u53ef\u80fd\u66f4\u597d\u3002\u8fd9\u4e2a\u73b0\u8c61\u88ab\u6982\u62ec\u4e3a\u201c\u8ff7\u5931\u5728\u4e2d\u95f4\u201d\u3002<a href=\"https:\/\/direct.mit.edu\/tacl\/article\/doi\/10.1162\/tacl_a_00638\/119630\/Lost-in-the-Middle-How-Language-Models-Use-Long\" target=\"_blank\" rel=\"noopener\">TACL\uff1a\u300aLost in the Middle: How Language Models Use Long Contexts\u300b<\/a><\/p>\n<p>\u56e0\u6b64\uff0c\u4e0a\u4e0b\u6587\u5de5\u7a0b\u4e0d\u662f\u4fe1\u606f\u5806\u79ef\uff0c\u800c\u662f\u9009\u62e9\u4e0e\u7f16\u6392\uff1a\u4ec0\u4e48\u5fc5\u987b\u653e\u5728\u5f53\u524d\u89c6\u91ce\u4e2d\uff0c\u4ec0\u4e48\u53ea\u5728\u9700\u8981\u65f6\u68c0\u7d22\uff0c\u4ec0\u4e48\u5e94\u5f53\u88ab\u6458\u8981\uff0c\u4ec0\u4e48\u5fc5\u987b\u4fdd\u6301\u539f\u6587\uff0c\u4ec0\u4e48\u5df2\u7ecf\u8fc7\u671f\uff0c\u4ec0\u4e48\u5177\u6709\u6b63\u5f0f\u6743\u5a01\u3002<\/p>\n<h2>\u6a21\u578b\u9700\u8981\u7684\u4e0d\u53ea\u662f\u80cc\u666f\uff0c\u8fd8\u9700\u8981\u201c\u4e8b\u5b9e\u7b49\u7ea7\u201d<\/h2>\n<p>\u5de5\u4f5c\u6750\u6599\u5e76\u4e0d\u5904\u4e8e\u540c\u4e00\u5730\u4f4d\u3002\u4f1a\u8bae\u4e0a\u7684\u60f3\u6cd5\u3001\u672a\u7ecf\u786e\u8ba4\u7684\u6570\u636e\u3001\u7b7e\u7f72\u540e\u7684\u5408\u540c\u548c\u6b63\u5f0f\u653f\u7b56\u6587\u4ef6\uff0c\u4e0d\u80fd\u56e0\u4e3a\u90fd\u88ab\u653e\u8fdb\u540c\u4e00\u4e2a\u6587\u4ef6\u5939\uff0c\u5c31\u88ab\u6a21\u578b\u89c6\u4e3a\u540c\u7b49\u53ef\u9760\u3002<\/p>\n<p>\u4eba\u7c7b\u5728\u719f\u6089\u7ec4\u7ec7\u65f6\uff0c\u5f80\u5f80\u4f1a\u81ea\u52a8\u7406\u89e3\u8fd9\u4e9b\u5dee\u5f02\u3002\u770b\u5230\u201cdraft\u201d\u5c31\u77e5\u9053\u5c1a\u672a\u6279\u51c6\uff0c\u770b\u5230\u65e7\u65e5\u671f\u4f1a\u5bfb\u627e\u65b0\u7248\u672c\uff0c\u770b\u5230\u540c\u4e8b\u7684\u731c\u6d4b\u4e0d\u4f1a\u5f53\u6210\u6b63\u5f0f\u51b3\u5b9a\u3002\u6a21\u578b\u672a\u5fc5\u62e5\u6709\u8fd9\u79cd\u9690\u542b\u80cc\u666f\u3002\u5982\u679c\u7cfb\u7edf\u53ea\u63d0\u4f9b\u6587\u672c\u800c\u6ca1\u6709\u6765\u6e90\u3001\u65e5\u671f\u3001\u7248\u672c\u548c\u72b6\u6001\uff0cAI\u5f88\u5bb9\u6613\u628a\u4e00\u6761\u9648\u65e7\u8bf4\u660e\u5199\u6210\u73b0\u884c\u89c4\u5219\uff0c\u6216\u628a\u63d0\u8bae\u5199\u6210\u5df2\u7ecf\u51b3\u5b9a\u7684\u4e8b\u9879\u3002<\/p>\n<p>\u6240\u4ee5\uff0c\u9ad8\u8d28\u91cfAI\u5de5\u4f5c\u9700\u8981\u7ed9\u4e0a\u4e0b\u6587\u589e\u52a0\u7ed3\u6784\uff1a<\/p>\n<ul>\n<li>\u6587\u4ef6\u662f\u8c01\u4ea7\u751f\u7684\uff1b<\/li>\n<li>\u9002\u7528\u4e8e\u54ea\u4e2a\u65f6\u671f\u548c\u5bf9\u8c61\uff1b<\/li>\n<li>\u662f\u8349\u7a3f\u3001\u8bc1\u636e\u3001\u63a8\u65ad\u8fd8\u662f\u6b63\u5f0f\u51b3\u5b9a\uff1b<\/li>\n<li>\u662f\u5426\u5df2\u7ecf\u88ab\u540e\u6765\u6587\u4ef6\u66ff\u4ee3\uff1b<\/li>\n<li>\u54ea\u4e9b\u6bb5\u843d\u53ef\u4ee5\u516c\u5f00\uff0c\u54ea\u4e9b\u53d7\u5230\u6743\u9650\u9650\u5236\uff1b<\/li>\n<li>\u51fa\u73b0\u51b2\u7a81\u65f6\u4ee5\u54ea\u4e2a\u6765\u6e90\u4e3a\u51c6\u3002<\/li>\n<\/ul>\n<p>\u8fd9\u5df2\u7ecf\u4e0d\u662f\u5199\u4f5c\u6280\u5de7\uff0c\u800c\u662f\u77e5\u8bc6\u7ba1\u7406\u548c\u5de5\u4f5c\u8bbe\u8ba1\u3002<\/p>\n<p>\u68c0\u7d22\u589e\u5f3a\u751f\u6210\uff08Retrieval-Augmented Generation\uff0cRAG\uff09\u7684\u57fa\u672c\u601d\u60f3\uff0c\u5c31\u662f\u8ba9\u6a21\u578b\u5728\u751f\u6210\u65f6\u8c03\u7528\u5916\u90e8\u77e5\u8bc6\u6765\u6e90\uff0c\u800c\u4e0d\u662f\u5b8c\u5168\u4f9d\u8d56\u8bad\u7ec3\u4e2d\u5f62\u6210\u7684\u53c2\u6570\u5316\u77e5\u8bc6\u3002\u6700\u65e9\u7cfb\u7edf\u5316\u63d0\u51faRAG\u7684\u7814\u7a76\uff0c\u628a\u6a21\u578b\u5185\u90e8\u77e5\u8bc6\u4e0e\u53ef\u66f4\u65b0\u7684\u5916\u90e8\u6587\u6863\u68c0\u7d22\u7ed3\u5408\u8d77\u6765\uff0c\u7528\u4e8e\u77e5\u8bc6\u5bc6\u96c6\u578b\u4efb\u52a1\u3002<a href=\"https:\/\/papers.neurips.cc\/paper\/2020\/hash\/6b493230205f780e1bc26945df7481e5-Abstract.html\" target=\"_blank\" rel=\"noopener\">NeurIPS\uff1a\u300aRetrieval-Augmented Generation for Knowledge-Intensive NLP Tasks\u300b<\/a><\/p>\n<p>\u4f46\u68c0\u7d22\u5e76\u4e0d\u4f1a\u81ea\u52a8\u4ea7\u751f\u771f\u76f8\u3002\u68c0\u7d22\u53ef\u80fd\u627e\u9519\u6587\u4ef6\u3001\u9057\u6f0f\u4f8b\u5916\u3001\u8fd4\u56de\u8fc7\u65f6\u7248\u672c\uff0c\u4e5f\u53ef\u80fd\u53ea\u6293\u5230\u5173\u952e\u8bcd\u76f8\u4f3c\u800c\u5b9e\u9645\u65e0\u5173\u7684\u6bb5\u843d\u3002\u771f\u6b63\u7684\u80fd\u529b\u4ecd\u7136\u5305\u62ec\uff1a\u600e\u6837\u7ec4\u7ec7\u77e5\u8bc6\u5e93\u3001\u600e\u6837\u6807\u8bb0\u6743\u5a01\u6765\u6e90\u3001\u600e\u6837\u8bc4\u4ef7\u68c0\u7d22\u8d28\u91cf\uff0c\u4ee5\u53ca\u600e\u6837\u8ba9\u6700\u7ec8\u5224\u65ad\u56de\u5230\u539f\u59cb\u8bc1\u636e\u3002<\/p>\n<h2>\u4e00\u9879\u4efb\u52a1\u8fd8\u9700\u8981\u5b8c\u6210\u6807\u51c6<\/h2>\n<p>\u63d0\u793a\u8bcd\u6587\u5316\u5bb9\u6613\u628a\u6ce8\u610f\u529b\u653e\u5728\u201c\u8f93\u51fa\u770b\u8d77\u6765\u600e\u6837\u201d\u3002\u771f\u5b9e\u5de5\u4f5c\u66f4\u5173\u5fc3\u201c\u600e\u6837\u8bc1\u660e\u5df2\u7ecf\u5b8c\u6210\u201d\u3002<\/p>\n<p>\u5982\u679c\u8981\u6c42AI\u4fee\u6539\u7a0b\u5e8f\uff0c\u201c\u4ee3\u7801\u5199\u5f97\u7b80\u6d01\u201d\u4e0d\u662f\u5145\u5206\u6807\u51c6\u3002\u5b83\u662f\u5426\u901a\u8fc7\u5355\u5143\u6d4b\u8bd5\u3001\u96c6\u6210\u6d4b\u8bd5\u548c\u5b89\u5168\u68c0\u67e5\uff0c\u624d\u662f\u53ef\u4ee5\u91cd\u590d\u9a8c\u8bc1\u7684\u7ed3\u679c\u3002\u5982\u679c\u8981\u6c42AI\u5206\u6790\u5408\u540c\uff0c\u201c\u8bed\u8a00\u4e13\u4e1a\u201d\u4e5f\u4e0d\u662f\u5b8c\u6210\u6807\u51c6\u3002\u5173\u952e\u6761\u6b3e\u662f\u5426\u5168\u90e8\u8986\u76d6\u3001\u5f15\u7528\u662f\u5426\u5bf9\u5e94\u539f\u6587\u3001\u4f8b\u5916\u662f\u5426\u88ab\u8bc6\u522b\uff0c\u624d\u51b3\u5b9a\u7ed3\u679c\u80fd\u5426\u91c7\u7528\u3002<\/p>\n<p>\u8fd9\u4e5f\u662f\u4e3a\u4ec0\u4e48AI\u4ee3\u7406\u9700\u8981\u8bc4\u4f30\u7cfb\u7edf\uff0c\u800c\u4e0d\u53ea\u662f\u66f4\u957f\u7684\u63d0\u793a\u3002Anthropic\u5728\u4ee3\u7406\u8bc4\u4f30\u5b9e\u8df5\u4e2d\u63d0\u51fa\uff0c\u8bc4\u4f30\u9700\u8981\u6839\u636e\u5b9e\u9645\u4efb\u52a1\u7ed3\u5408\u9759\u6001\u68c0\u67e5\u3001\u7a0b\u5e8f\u5316\u6d4b\u8bd5\u3001\u4eba\u7c7b\u8bc4\u5206\u548c\u6a21\u578b\u8bc4\u5ba1\u7b49\u4e0d\u540c\u65b9\u5f0f\uff1b\u201c\u662f\u5426\u9075\u5b88\u6307\u4ee4\u201d\u201c\u662f\u5426\u7834\u574f\u5df2\u6709\u5185\u5bb9\u201d\u201c\u7ed3\u679c\u662f\u5426\u771f\u6b63\u826f\u597d\u201d\u662f\u4e0d\u540c\u7ef4\u5ea6\u3002<a href=\"https:\/\/www.anthropic.com\/engineering\/demystifying-evals-for-ai-agents\" target=\"_blank\" rel=\"noopener\">Anthropic\uff1a\u300aDemystifying evals for AI agents\u300b<\/a><\/p>\n<p>\u4e00\u4e2a\u4eba\u5982\u679c\u53ea\u4f1a\u5bf9AI\u8bf4\u201c\u8bf7\u66f4\u51c6\u786e\u201d\uff0c\u5374\u4e0d\u80fd\u8bf4\u660e\u54ea\u4e9b\u5b57\u6bb5\u5fc5\u987b\u5b58\u5728\u3001\u54ea\u4e9b\u9519\u8bef\u4e0d\u53ef\u63a5\u53d7\u3001\u4ec0\u4e48\u6d4b\u8bd5\u5fc5\u987b\u901a\u8fc7\uff0c\u4ed6\u4ecd\u7136\u6ca1\u6709\u628a\u5de5\u4f5c\u4ea4\u4ee3\u6e05\u695a\u3002<\/p>\n<p>\u56e0\u6b64\uff0cAI\u65f6\u4ee3\u7684\u4e00\u9879\u91cd\u8981\u80fd\u529b\u662f\u628a\u8d28\u91cf\u8981\u6c42\u53d8\u6210\u53ef\u68c0\u67e5\u7684\u6761\u4ef6\uff1a<\/p>\n<ul>\n<li>\u4e8b\u5b9e\u4e3b\u5f20\u5fc5\u987b\u6307\u5411\u6765\u6e90\uff1b<\/li>\n<li>\u6570\u5b57\u5fc5\u987b\u4fdd\u7559\u5355\u4f4d\u3001\u65e5\u671f\u548c\u5206\u6bcd\uff1b<\/li>\n<li>\u4fee\u6539\u4e0d\u5f97\u8d85\u51fa\u6307\u5b9a\u6587\u4ef6\uff1b<\/li>\n<li>\u6240\u6709\u6d4b\u8bd5\u5fc5\u987b\u901a\u8fc7\uff1b<\/li>\n<li>\u4e0d\u786e\u5b9a\u5185\u5bb9\u5fc5\u987b\u663e\u5f0f\u6807\u8bb0\uff1b<\/li>\n<li>\u9ad8\u98ce\u9669\u64cd\u4f5c\u5fc5\u987b\u505c\u4e0b\u6765\u7b49\u5f85\u6279\u51c6\u3002<\/li>\n<\/ul>\n<p>\u8fd9\u4e9b\u6761\u4ef6\u6bd4\u4e00\u53e5\u201c\u4f60\u662f\u4e16\u754c\u4e00\u6d41\u4e13\u5bb6\u201d\u66f4\u80fd\u6539\u53d8\u5b9e\u9645\u7ed3\u679c\u3002<\/p>\n<h2>\u4e3a\u4ec0\u4e48\u63d0\u793a\u8bcd\u4ecd\u7136\u5982\u6b64\u53d7\u6b22\u8fce<\/h2>\n<p>\u63d0\u793a\u8bcd\u6d41\u884c\uff0c\u6709\u51e0\u4e2a\u53ef\u4ee5\u7406\u89e3\u7684\u539f\u56e0\u3002<\/p>\n<p>\u7b2c\u4e00\uff0c\u5b83\u662f\u7528\u6237\u80fd\u591f\u7acb\u5373\u770b\u5230\u548c\u63a7\u5236\u7684\u90e8\u5206\u3002\u6a21\u578b\u3001\u68c0\u7d22\u7cfb\u7edf\u548c\u6743\u9650\u7ed3\u6784\u9690\u85cf\u5728\u540e\u53f0\uff0c\u800c\u4e00\u53e5\u8bdd\u6539\u53d8\u8f93\u51fa\u7684\u6548\u679c\u975e\u5e38\u76f4\u89c2\u3002<\/p>\n<p>\u7b2c\u4e8c\uff0c\u5b83\u5bb9\u6613\u4f20\u64ad\u3002\u4e00\u4e2a\u201c\u4e07\u80fd\u63d0\u793a\u8bcd\u201d\u53ef\u4ee5\u590d\u5236\u7c98\u8d34\uff0c\u4e00\u5957\u77e5\u8bc6\u6cbb\u7406\u548c\u9a8c\u8bc1\u6d41\u7a0b\u5374\u9700\u8981\u7406\u89e3\u5177\u4f53\u5de5\u4f5c\uff0c\u65e0\u6cd5\u5728\u793e\u4ea4\u5a92\u4f53\u4e0a\u7528\u4e00\u5f20\u56fe\u7247\u89e3\u51b3\u3002<\/p>\n<p>\u7b2c\u4e09\uff0c\u65e9\u671f\u804a\u5929\u5f0fAI\u4e3b\u8981\u4ee5\u5355\u8f6e\u56de\u7b54\u51fa\u73b0\u3002\u90a3\u65f6\u7528\u6237\u4e0e\u6a21\u578b\u7684\u63a5\u89e6\u70b9\u5c31\u662f\u8f93\u5165\u6846\uff0c\u628a\u91cd\u70b9\u653e\u5728\u63d0\u793a\u8bcd\u4e0a\u662f\u81ea\u7136\u7684\u3002\u968f\u7740AI\u5f00\u59cb\u8bfb\u53d6\u9879\u76ee\u6587\u4ef6\u3001\u8c03\u7528\u5de5\u5177\u3001\u957f\u671f\u6267\u884c\u548c\u8de8\u5e94\u7528\u534f\u4f5c\uff0c\u5355\u53e5\u6307\u4ee4\u6240\u80fd\u63a7\u5236\u7684\u6bd4\u4f8b\u6b63\u5728\u4e0b\u964d\u3002<\/p>\n<p>\u7b2c\u56db\uff0c\u63d0\u793a\u8bcd\u628a\u5931\u8d25\u89e3\u91ca\u4e3a\u4e2a\u4eba\u6280\u5de7\u4e0d\u8db3\u3002\u5982\u679c\u7ed3\u679c\u4e0d\u597d\uff0c\u53ea\u8981\u7ee7\u7eed\u4fee\u6539\u63aa\u8f9e\uff0c\u770b\u8d77\u6765\u5c31\u8fd8\u6709\u7b80\u5355\u89e3\u6cd5\u3002\u76f8\u6bd4\u4e4b\u4e0b\uff0c\u627f\u8ba4\u6750\u6599\u6df7\u4e71\u3001\u89c4\u5219\u51b2\u7a81\u3001\u4efb\u52a1\u672c\u8eab\u6ca1\u6709\u5b8c\u6210\u6807\u51c6\uff0c\u610f\u5473\u7740\u7ec4\u7ec7\u9700\u8981\u5904\u7406\u66f4\u56f0\u96be\u7684\u95ee\u9898\u3002<\/p>\n<p>\u63d0\u793a\u8bcd\u4e0d\u4f1a\u6d88\u5931\u3002\u5b83\u4f1a\u50cf\u641c\u7d22\u5173\u952e\u8bcd\u3001\u7a0b\u5e8f\u53c2\u6570\u548c\u5de5\u4f5c\u6307\u4ee4\u4e00\u6837\uff0c\u7ee7\u7eed\u6210\u4e3a\u4eba\u673a\u534f\u4f5c\u7684\u91cd\u8981\u63a5\u53e3\u3002\u4f46\u5b83\u4e0d\u4f1a\u72ec\u81ea\u627f\u62c5\u6574\u4e2a\u7cfb\u7edf\u7684\u53ef\u9760\u6027\u3002<\/p>\n<h2>\u771f\u6b63\u91cd\u8981\u7684\u4e94\u79cd\u80fd\u529b<\/h2>\n<p>\u5982\u679c\u628a\u201c\u4f1a\u7528AI\u201d\u4ece\u4e00\u53e5\u63d0\u793a\u8bcd\u6269\u5c55\u5230\u5b8c\u6574\u5de5\u4f5c\uff0c\u53ef\u4ee5\u770b\u5230\u4e94\u79cd\u66f4\u57fa\u7840\u7684\u80fd\u529b\u3002<\/p>\n<p>\u7b2c\u4e00\u662f<strong>\u5b9a\u4e49\u95ee\u9898<\/strong>\u3002\u77e5\u9053\u8981\u89e3\u51b3\u4ec0\u4e48\u3001\u670d\u52a1\u8c01\u3001\u54ea\u4e9b\u7ed3\u679c\u867d\u7136\u6f02\u4eae\u5374\u504f\u79bb\u76ee\u6807\u3002<\/p>\n<p>\u7b2c\u4e8c\u662f<strong>\u6574\u7406\u4e0a\u4e0b\u6587<\/strong>\u3002\u63d0\u4f9b\u6b63\u786e\u6750\u6599\uff0c\u8bc6\u522b\u7248\u672c\u3001\u6765\u6e90\u3001\u6743\u9650\u548c\u9002\u7528\u8303\u56f4\u3002<\/p>\n<p>\u7b2c\u4e09\u662f<strong>\u8bbe\u8ba1\u4efb\u52a1<\/strong>\u3002\u628a\u6a21\u7cca\u76ee\u6807\u5206\u89e3\u4e3aAI\u53ef\u4ee5\u5b8c\u6210\u3001\u4eba\u53ef\u4ee5\u68c0\u67e5\u7684\u6b65\u9aa4\u3002<\/p>\n<p>\u7b2c\u56db\u662f<strong>\u5efa\u7acb\u9a8c\u8bc1<\/strong>\u3002\u7528\u8bc1\u636e\u3001\u6d4b\u8bd5\u548c\u72ec\u7acb\u68c0\u67e5\u5224\u65ad\u7ed3\u679c\uff0c\u800c\u4e0d\u662f\u7528\u8bed\u6c14\u548c\u6d41\u7545\u5ea6\u5224\u65ad\u3002<\/p>\n<p>\u7b2c\u4e94\u662f<strong>\u63a7\u5236\u91c7\u7528<\/strong>\u3002\u51b3\u5b9a\u8f93\u51fa\u662f\u4f9b\u53c2\u8003\u3001\u8fdb\u5165\u8349\u7a3f\u3001\u81ea\u52a8\u6267\u884c\uff0c\u8fd8\u662f\u5fc5\u987b\u7531\u6709\u8d44\u683c\u7684\u4eba\u6279\u51c6\u3002<\/p>\n<p>\u4e00\u4e2a\u63d0\u793a\u8bcd\u9ad8\u624b\u5982\u679c\u7f3a\u5c11\u8fd9\u4e9b\u80fd\u529b\uff0c\u53ef\u80fd\u66f4\u5feb\u5730\u4ea7\u751f\u4ee4\u4eba\u4fe1\u670d\u7684\u9519\u8bef\u3002\u4e00\u4e2a\u771f\u6b63\u7406\u89e3\u5de5\u4f5c\u7684\u4eba\uff0c\u5373\u4f7f\u6307\u4ee4\u6734\u7d20\uff0c\u4e5f\u80fd\u901a\u8fc7\u826f\u597d\u6750\u6599\u3001\u660e\u786e\u6807\u51c6\u548c\u53cd\u590d\u9a8c\u8bc1\u83b7\u5f97\u7a33\u5b9a\u7ed3\u679c\u3002<\/p>\n<h2>\u7ed3\u8bba\uff1aAI\u80fd\u529b\u9996\u5148\u662f\u4e00\u79cd\u5de5\u4f5c\u7ec4\u7ec7\u80fd\u529b<\/h2>\n<p>\u201c\u4f1a\u5199\u63d0\u793a\u8bcd\u201d\u9002\u5408\u63cf\u8ff0\u4eba\u4e0e\u804a\u5929\u6a21\u578b\u7684\u4e00\u6b21\u4e92\u52a8\uff0c\u5374\u4e0d\u8db3\u4ee5\u63cf\u8ff0AI\u53c2\u4e0e\u771f\u5b9e\u5de5\u4f5c\u7684\u5168\u90e8\u80fd\u529b\u3002<\/p>\n<p>\u5f53\u4efb\u52a1\u7b80\u5355\u3001\u98ce\u9669\u4f4e\u3001\u8f93\u5165\u5b8c\u6574\u65f6\uff0c\u597d\u7684\u63aa\u8f9e\u80fd\u591f\u663e\u8457\u6539\u5584\u8f93\u51fa\u3002\u5f53\u4efb\u52a1\u9700\u8981\u591a\u4efd\u8d44\u6599\u3001\u8fde\u7eed\u6b65\u9aa4\u3001\u5916\u90e8\u5de5\u5177\u548c\u6b63\u5f0f\u8d23\u4efb\u65f6\uff0c\u51b3\u5b9a\u7ed3\u679c\u7684\u5c06\u662f\u66f4\u5927\u7684\u7ed3\u6784\uff1a\u4efb\u52a1\u662f\u5426\u6e05\u695a\u3001\u4e0a\u4e0b\u6587\u662f\u5426\u6b63\u786e\u3001\u6765\u6e90\u662f\u5426\u6709\u7b49\u7ea7\u3001\u6743\u9650\u662f\u5426\u53d7\u63a7\u3001\u7ed3\u679c\u662f\u5426\u53ef\u4ee5\u9a8c\u8bc1\u3002<\/p>\n<p>\u56e0\u6b64\uff0c\u66f4\u51c6\u786e\u7684\u5224\u65ad\u662f\uff1a<\/p>\n<blockquote>\n<p>\u63d0\u793a\u8bcd\u51b3\u5b9a\u6a21\u578b\u6b64\u523b\u542c\u5230\u4ec0\u4e48\uff1b\u5de5\u4f5c\u7cfb\u7edf\u51b3\u5b9a\u5b83\u80fd\u591f\u770b\u89c1\u4ec0\u4e48\u3001\u505a\u4ec0\u4e48\u3001\u600e\u6837\u88ab\u68c0\u67e5\uff0c\u4ee5\u53ca\u7ed3\u679c\u80fd\u5426\u88ab\u6b63\u5f0f\u91c7\u7528\u3002<\/p>\n<\/blockquote>\n<p>\u672a\u6765\u771f\u6b63\u7a00\u7f3a\u7684\u672a\u5fc5\u662f\u80fd\u591f\u80cc\u8bf5\u63d0\u793a\u8bcd\u6a21\u677f\u7684\u4eba\uff0c\u800c\u662f\u80fd\u591f\u7406\u89e3\u4e00\u9879\u5de5\u4f5c\uff0c\u5e76\u628a\u5b83\u91cd\u65b0\u7ec4\u7ec7\u6210\u201c\u4eba\u63d0\u51fa\u76ee\u6807\u2014\u7cfb\u7edf\u63d0\u4f9b\u8bc1\u636e\u2014AI\u751f\u6210\u6216\u6267\u884c\u2014\u9a8c\u8bc1\u673a\u5236\u68c0\u67e5\u2014\u8d23\u4efb\u4eba\u91c7\u7528\u201d\u7684\u4eba\u3002<\/p>\n<p>\u63d0\u793a\u8bcd\u4ecd\u7136\u662f\u5165\u53e3\u3002\u4f46\u5165\u53e3\u4e0d\u662f\u6574\u5ea7\u5efa\u7b51\u3002<\/p>\n<h3>\u4e3b\u8981\u8d44\u6599<\/h3>\n<ul>\n<li><a href=\"https:\/\/www.anthropic.com\/engineering\/effective-context-engineering-for-ai-agents\" target=\"_blank\" rel=\"noopener\">Anthropic\uff1aEffective context engineering for AI agents<\/a><\/li>\n<li><a href=\"https:\/\/www.anthropic.com\/engineering\/demystifying-evals-for-ai-agents\" target=\"_blank\" rel=\"noopener\">Anthropic\uff1aDemystifying evals for AI agents<\/a><\/li>\n<li><a href=\"https:\/\/direct.mit.edu\/tacl\/article\/doi\/10.1162\/tacl_a_00638\/119630\/Lost-in-the-Middle-How-Language-Models-Use-Long\" target=\"_blank\" rel=\"noopener\">Nelson F. Liu et al.\uff1aLost in the Middle: How Language Models Use Long Contexts<\/a><\/li>\n<li><a href=\"https:\/\/papers.neurips.cc\/paper\/2020\/hash\/6b493230205f780e1bc26945df7481e5-Abstract.html\" target=\"_blank\" rel=\"noopener\">Patrick Lewis et al.\uff1aRetrieval-Augmented Generation for Knowledge-Intensive NLP Tasks<\/a><\/li>\n<li><a href=\"https:\/\/www.nist.gov\/publications\/artificial-intelligence-risk-management-framework-generative-artificial-intelligence\" target=\"_blank\" rel=\"noopener\">NIST\uff1aArtificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile<\/a><\/li>\n<\/ul>\n<p><strong>\u7ee7\u7eed\u9605\u8bfb\uff1a<\/strong><a href=\"https:\/\/geoffreychen.com\/zh\/category\/ai-frontiers\/after-ai-enters-the-workflow\/\">\u6d4f\u89c8\u300aAI\u8fdb\u5165\u5de5\u4f5c\u4e4b\u540e\u300b\u7cfb\u5217\u6587\u7ae0<\/a><\/p>\n<p>","protected":false},"excerpt":{"rendered":"","protected":false},"author":17897162,"featured_media":0,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_coblocks_attr":"","_coblocks_dimensions":"","_coblocks_responsive_height":"","_coblocks_accordion_ie_support":"","_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_geoff_meta_description_en":"Argues that prompt skill alone isn\u2019t enough for AI work; context engineering, source authority, and testable completion criteria decide whether outputs are adoptable.","_geoff_meta_description_zh":"\u672c\u6587\u6307\u51fa\u63d0\u793a\u8bcd\u80fd\u529b\u4e0d\u8db3\u4ee5\u51b3\u5b9aAI\u5de5\u4f5c\u8d28\u91cf\uff0c\u5173\u952e\u5728\u4e8e\u4e0a\u4e0b\u6587\u5de5\u7a0b\u3001\u6765\u6e90\u6743\u5a01\u5206\u7ea7\u4e0e\u53ef\u9a8c\u8bc1\u7684\u5b8c\u6210\u6807\u51c6\uff0c\u6700\u7ec8\u51b3\u5b9a\u7ed3\u679c\u80fd\u5426\u88ab\u91c7\u7528\u3002","_wpcom_ai_launchpad_first_post":false,"_jetpack_feature_clip_id":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2},"jetpack_post_was_ever_published":false},"categories":[787296333],"tags":[787296336,787295117,787295118,787296338,787296334,787296342,787296343,787296335],"class_list":["post-16742","post","type-post","status-publish","format-standard","hentry","category-after-ai-enters-the-workflow","tag-after-ai-enters-the-workflow","tag-ai-literacy","tag-ai-literacy-zh","tag-after-ai-enters-the-workflow-zh","tag-from-answering-questions-to-participating-in-work","tag-prompt-engineering","tag-prompt-engineering-zh","tag-from-answering-questions-to-participating-in-work-zh",""],"translatepress_titles":{"en_AU":"Why Prompt Writing Is Not the Most Important AI Skill","zh_CN":"\u4e3a\u4ec0\u4e48\u201c\u4f1a\u5199\u63d0\u793a\u8bcd\u201d\u4e0d\u662f\u6700\u91cd\u8981\u7684AI\u80fd\u529b\uff1f"},"jetpack_publicize_connections":[],"jetpack_likes_enabled":false,"jetpack_sharing_enabled":true,"jetpack_shortlink":"https:\/\/wp.me\/p1ar9H-4m2","jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/posts\/16742","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/users\/17897162"}],"replies":[{"embeddable":true,"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/comments?post=16742"}],"version-history":[{"count":2,"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/posts\/16742\/revisions"}],"predecessor-version":[{"id":16767,"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/posts\/16742\/revisions\/16767"}],"wp:attachment":[{"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/media?parent=16742"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/categories?post=16742"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/tags?post=16742"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}