{"id":1184661,"date":"2026-07-08T20:12:00","date_gmt":"2026-07-09T03:12:00","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-blog-post&p=1184661"},"modified":"2026-08-26T02:51:26","modified_gmt":"2026-08-26T09:51:26","slug":"%e5%a4%a7%e6%a8%a1%e5%9e%8b%e4%bc%9a%e6%89%ae%e6%bc%94%e4%ba%ba%ef%bc%8c%e4%bd%86%e7%9c%9f%e7%9a%84%e6%87%82%e4%ba%ba%e5%90%97%ef%bc%9f%e4%b8%80%e5%a5%97%e6%96%b0%e6%a1%86%e6%9e%b6","status":"publish","type":"msr-blog-post","link":"https:\/\/www.microsoft.com\/en-us\/research\/articles\/%e5%a4%a7%e6%a8%a1%e5%9e%8b%e4%bc%9a%e6%89%ae%e6%bc%94%e4%ba%ba%ef%bc%8c%e4%bd%86%e7%9c%9f%e7%9a%84%e6%87%82%e4%ba%ba%e5%90%97%ef%bc%9f%e4%b8%80%e5%a5%97%e6%96%b0%e6%a1%86%e6%9e%b6\/","title":{"rendered":"\u5927\u6a21\u578b\u4f1a\u201c\u626e\u6f14\u4eba\u201d\uff0c\u4f46\u771f\u7684\u61c2\u4eba\u5417\uff1f\u4e00\u5957\u65b0\u6846\u67b6\uff0c\u7ed9 AI \u7684\u201c\u4eba\u5473\u201d\u505a\u4f53\u68c0"},"content":{"rendered":"\n

\u5f53\u6211\u4eec\u8ba9 AI \u626e\u6f14\u4e00\u4f4d\u7528\u6237\u65f6\uff0c\u5b83\u901a\u5e38\u80fd\u591f\u5feb\u901f\u7ed9\u51fa\u50cf\u6a21\u50cf\u6837\u7684\u56de\u7b54\u3002\u5b83\u53ef\u4ee5\u6a21\u4eff\u4e00\u4e2a\u5f71\u8ff7\u6311\u9009\u7535\u5f71\uff0c\u6a21\u4eff\u4e00\u4f4d\u7f51\u53cb\u8ba8\u8bba\u95ee\u9898\uff0c\u4e5f\u53ef\u4ee5\u6839\u636e\u5386\u53f2\u8db3\u8ff9\u9884\u6d4b\u4e00\u4e2a\u4eba\u4e0b\u4e00\u7ad9\u4f1a\u53bb\u54ea\u91cc\u3002<\/p>\n\n\n\n

\u968f\u7740\u5927\u6a21\u578b\u80fd\u529b\u7684\u5feb\u901f\u63d0\u5347\uff0c\u667a\u80fd NPC\u3001\u7528\u6237\u6a21\u62df\u5668\u3001\u793e\u4f1a\u4eff\u771f\u7b49\u5e94\u7528\u6b63\u5728\u53d8\u5f97\u8d8a\u6765\u8d8a\u771f\u5b9e\u3002\u4f46\u4e00\u4e2a\u66f4\u5173\u952e\u7684\u95ee\u9898\u662f\uff1aAI\u751f\u6210\u7684\u884c\u4e3a\uff0c\u771f\u7684\u50cf\u4eba\u5417\uff1f<\/p>\n\n\n\n

\u5b83\u662f\u5426\u771f\u7684\u7406\u89e3\u4e86\u4e00\u4e2a\u4eba\u7684\u504f\u597d\u3001\u601d\u7ef4\u65b9\u5f0f\u548c\u957f\u671f\u4e60\u60ef\uff1f\u5b83\u80fd\u5426\u5728\u591a\u8f6e\u4ea4\u4e92\u540e\u4f9d\u7136\u4fdd\u6301\u201c\u540c\u4e00\u4e2a\u4eba\u201d\u7684\u4e00\u81f4\u6027\uff1f\u5f53\u6a21\u62df\u6210\u5343\u4e0a\u4e07\u4e2a\u7528\u6237\u65f6\uff0c\u8fd9\u4e9b AI \u53c8\u80fd\u5426\u50cf\u771f\u5b9e\u793e\u4f1a\u4e00\u6837\u5448\u73b0\u4e30\u5bcc\u3001\u591a\u6837\u3001\u751a\u81f3\u5f7c\u6b64\u4e0d\u540c\u7684\u4eba\u7fa4\u5206\u5e03\uff1f<\/p>\n\n\n\n

\u8fd1\u671f\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u7684\u7814\u7a76\u5458\u4eec\u63d0\u51fa\u4e86\u4e00\u5957\u9762\u5411\u5927\u6a21\u578b\u7c7b\u4eba\u884c\u4e3a\u7684\u8ba1\u7b97\u8bc4\u6d4b\u6846\u67b6\uff0c\u5c1d\u8bd5\u56de\u7b54\u8fd9\u4e2a\u95ee\u9898\u3002\u7814\u7a76\u53d1\u73b0\uff1a\u5927\u6a21\u578b\u5df2\u7ecf\u80fd\u591f\u751f\u6210\u4e0d\u5c11\u770b\u4f3c\u5408\u7406\u7684\u4e2a\u4f53\u884c\u4e3a\uff0c\u4f46\u8ddd\u79bb\u771f\u6b63\u7a33\u5b9a\u3001\u53ef\u4fe1\u3001\u4e30\u5bcc\u7684\u4eba\u7c7b\u884c\u4e3a\u6a21\u62df\uff0c\u4ecd\u6709\u5f88\u957f\u7684\u8def\u8981\u8d70\u3002\u76f8\u5173\u8bba\u6587\u5df2\u88abICML 2026\u63a5\u6536<\/strong>\u3002<\/p>\n\n\n\n

\u8bba\u6587\u4fe1\u606f\u5df2\u6574\u7406\u81f3\u6587\u672b\uff0c\u6b22\u8fce\u70b9\u51fb\u76f8\u5173\u94fe\u63a5\uff0c\u4e86\u89e3\u66f4\u591a\u6280\u672f\u8be6\u60c5\u3002<\/em><\/strong><\/p>\n\n\n\n

\u4e00\u4e2a\u4f1a\u804a\u5929\u7684 AI\uff0c\u672a\u5fc5\u4f1a\u201c\u751f\u6d3b\u201d<\/h2>\n\n\n\n

\u4f20\u7edf\u56fe\u7075\u6d4b\u8bd5\u5173\u6ce8\u7684\u662f\uff0c\u4eba\u7c7b\u80fd\u5426\u5728\u77ed\u65f6\u95f4\u5bf9\u8bdd\u4e2d\u5206\u8fa8\u51fa\u5bf9\u65b9\u662f\u4eba\u8fd8\u662f\u673a\u5668\u3002<\/p>\n\n\n\n

\u8fd9\u7c7b\u6d4b\u8bd5\u5f88\u91cd\u8981\uff0c\u4f46\u4e5f\u5b58\u5728\u5929\u7136\u5c40\u9650\u3002\u9996\u5148\uff0c\u77ed\u77ed\u51e0\u5206\u949f\u7684\u804a\u5929\uff0c\u53ea\u80fd\u770b\u5230 AI\u884c\u4e3a\u7684\u4e00\u4e2a\u5207\u9762\u3002\u4e00\u4e2a\u6a21\u578b\u6216\u8bb8\u80fd\u591f\u5728\u4e00\u8f6e\u5bf9\u8bdd\u4e2d\u663e\u5f97\u81ea\u7136\uff0c\u4f46\u672a\u5fc5\u80fd\u5728\u51e0\u5341\u8f6e\u9009\u62e9\u3001\u56de\u7b54\u6216\u884c\u52a8\u4e2d\u6301\u7eed\u4fdd\u6301\u4eba\u683c\u3001\u504f\u597d\u548c\u903b\u8f91\u7684\u4e00\u81f4\u3002<\/p>\n\n\n\n

\u66f4\u91cd\u8981\u7684\u662f\uff0c\u8bb8\u591a\u771f\u5b9e\u4e16\u754c\u4e2d\u7684\u884c\u4e3a\u5e76\u4e0d\u5b58\u5728\u552f\u4e00\u6b63\u786e\u7b54\u6848\u3002\u5373\u4f7f\u662f\u540c\u4e00\u4e2a\u7528\u6237\u5728\u76f8\u4f3c\u60c5\u5883\u4e0b\u91cd\u590d\u8fdb\u884c\u9009\u62e9\u6216\u56de\u7b54\uff0c\u7ed3\u679c\u4e5f\u672a\u5fc5\u5b8c\u5168\u4e00\u81f4\u3002\u4ed6\u53ef\u80fd\u5728\u4e24\u90e8\u90fd\u611f\u5174\u8da3\u7684\u7535\u5f71\u95f4\u505a\u51fa\u4e0d\u540c\u9009\u62e9\uff0c\u4e5f\u53ef\u80fd\u4ee5\u4e0d\u540c\u7684\u63aa\u8f9e\u3001\u89d2\u5ea6\u6216\u8bba\u8bc1\u8def\u5f84\u56de\u7b54\u540c\u4e00\u4e2a\u95ee\u9898\u3002\u6b64\u65f6\uff0c\u8bc4\u4f30\u7684\u5173\u952e\u4e0d\u518d\u662f\u201c\u7b54\u5bf9\u4e86\u5417\u201d\uff0c\u800c\u662f AI \u751f\u6210\u7684\u884c\u4e3a\u662f\u5426\u4ecd\u7136\u7b26\u5408\u8fd9\u4e2a\u4eba\u53ef\u80fd\u5448\u73b0\u7684\u504f\u597d\u4e0e\u884c\u4e3a\u5206\u5e03\u3002<\/p>\n\n\n\n

<\/p>\n\n\n\n

\u7528\u771f\u5b9e\u4e16\u754c\u884c\u4e3a\uff0c\u7ed9 AI \u505a\u4e00\u6b21\u201c\u4eba\u5473\u4f53\u68c0\u201d<\/h2>\n\n\n\n

\u4e3a\u6b64\uff0c\u7814\u7a76\u5458\u4eec\u6ca1\u6709\u4e3a\u6a21\u578b\u8bbe\u8ba1\u4e00\u5957\u4eba\u5de5\u7f16\u5199\u7684\u8003\u9898\uff0c\u800c\u662f\u76f4\u63a5\u4ece\u771f\u5b9e\u4e16\u754c\u7684\u516c\u5f00\u884c\u4e3a\u6570\u636e\u4e2d\u5bfb\u627e\u7b54\u6848\u3002\u8be5\u7814\u7a76\u5de5\u4f5c\u8986\u76d6\u4e86\u4e09\u7c7b\u5178\u578b\u573a\u666f\uff1a<\/p>\n\n\n\n

– \u5728\u7ebf\u8d2d\u7269\uff1a\u7528\u6237\u5728\u7535\u5546\u5e73\u53f0\u4e2d\u7684\u8fde\u7eed\u8d2d\u4e70\u4e0e\u6d4f\u89c8\u504f\u597d\uff1b<\/p>\n\n\n\n

– \u5f00\u653e\u95ee\u7b54\uff1a\u7528\u6237\u5728\u5f00\u653e\u793e\u533a\u4e2d\u7684\u56de\u7b54\u3001\u89c2\u70b9\u4e0e\u8868\u8fbe\u65b9\u5f0f\uff1b<\/p>\n\n\n\n

– \u57ce\u5e02\u51fa\u884c\uff1a\u7528\u6237\u5728\u57fa\u4e8e\u4f4d\u7f6e\u670d\u52a1\u7684\u5e73\u53f0\u4e0a\u7559\u4e0b\u7684\u65f6\u7a7a\u7b7e\u5230\u8f68\u8ff9\u3002<\/p>\n\n\n\n

\u5728\u6bcf\u4e2a\u573a\u666f\u4e2d\uff0c\u7814\u7a76\u5458\u4eec\u90fd\u5148\u7528\u7528\u6237\u5386\u53f2\u884c\u4e3a\u6784\u5efa\u7528\u6237\u753b\u50cf\uff0c\u518d\u8ba9\u5927\u6a21\u578b\u6a21\u62df\u8be5\u7528\u6237\u540e\u7eed\u7684\u591a\u8f6e\u884c\u4e3a\u3002\u968f\u540e\uff0c\u7cfb\u7edf\u5c06AI \u7684\u6a21\u62df\u7ed3\u679c\u4e0e\u771f\u5b9e\u7528\u6237\u884c\u4e3a\u8fdb\u884c\u5bf9\u6bd4\u3002<\/p>\n\n\n\n

\u6574\u4e2a\u6846\u67b6\u7531\u56db\u90e8\u5206\u7ec4\u6210\uff1a\u771f\u5b9e\u884c\u4e3a\u6570\u636e\u3001LLM\u6a21\u62df\u5668\u3001\u7528\u6237\u884c\u4e3a\u7f16\u7801\u5668\u548c\u8bc4\u6d4b\u5668\u3002\u5176\u4e2d\uff0c\u771f\u5b9e\u884c\u4e3a\u6570\u636e\u63d0\u4f9b\u88ab\u6a21\u62df\u7528\u6237\u7684\u5386\u53f2\u8f68\u8ff9\uff0c\u4ee5\u53ca\u7528\u4e8e\u5bf9\u7167\u7684\u771f\u5b9e\u540e\u7eed\u884c\u4e3a\uff1bLLM\u6a21\u62df\u5668\u6839\u636e\u7528\u6237\u753b\u50cf\u548c\u5f53\u524d\u60c5\u5883\uff0c\u9010\u6b65\u751f\u6210\u8be5\u7528\u6237\u53ef\u80fd\u505a\u51fa\u7684\u540e\u7eed\u9009\u62e9\u3001\u56de\u7b54\u6216\u884c\u52a8\uff1b\u7528\u6237\u884c\u4e3a\u7f16\u7801\u5668\u5219\u4ece\u5927\u91cf\u771f\u5b9e\u6570\u636e\u4e2d\u5b66\u4e60\u201c\u4ec0\u4e48\u6837\u7684\u8fde\u7eed\u884c\u4e3a\u5c5e\u4e8e\u540c\u4e00\u4e2a\u4eba\u201d\uff0c\u5c06\u884c\u4e3a\u5e8f\u5217\u6620\u5c04\u4e3a\u80fd\u591f\u8868\u5f81\u504f\u597d\u3001\u98ce\u683c\u4e0e\u884c\u4e3a\u6a21\u5f0f\u7684\u9ad8\u7ef4\u5411\u91cf\uff1b\u6700\u540e\uff0c\u8bc4\u6d4b\u5668\u5728\u8fd9\u4e9b\u8868\u5f81\u4e4b\u4e0a\uff0c\u4ece\u591a\u4e2a\u7ef4\u5ea6\u5224\u65adAI \u884c\u4e3a\u7684\u7c7b\u4eba\u6027\u3002<\/p>\n\n\n\n

\u672c\u6846\u67b6\u7684\u6838\u5fc3\u8bbe\u8ba1\u4e0d\u4f9d\u8d56\u7279\u5b9a\u4efb\u52a1\u5f62\u5f0f\uff0c\u53ea\u8981\u5177\u5907\u8db3\u591f\u7684\u771f\u5b9e\u7528\u6237\u884c\u4e3a\u6570\u636e\uff0c\u4fbf\u53ef\u81ea\u7136\u6269\u5c55\u81f3\u8bc4\u8bba\u64b0\u5199\u3001\u65e5\u8bb0\u521b\u4f5c\u3001\u996e\u98df\u9009\u62e9\u7b49\u66f4\u591a\u5f00\u653e\u5f0f\u884c\u4e3a\u573a\u666f\u3002\u4e0e\u201c\u770b\u4e00\u6761\u56de\u7b54\u50cf\u4e0d\u50cf\u4eba\u201d\u4e0d\u540c\uff0c\u8fd9\u5957\u6846\u67b6\u5e0c\u671b\u4ece\u4e2a\u4f53\u5230\u7fa4\u4f53\u3001\u4ece\u5373\u65f6\u884c\u4e3a\u5230\u957f\u671f\u8f68\u8ff9\uff0c\u5168\u9762\u8bc4\u4f30AI \u7684\u7c7b\u4eba\u7a0b\u5ea6\u3002<\/p>\n\n\n\n

\"diagram\"
\u56fe1\uff1a\u6574\u4f53\u8bc4\u6d4b\u6846\u67b6\u56fe<\/figcaption><\/figure>\n\n\n\n

<\/p>\n\n\n\n

\u5224\u65ad AI \u662f\u5426\u201c\u50cf\u4eba\u201d\uff0c\u4e0d\u80fd\u53ea\u770b\u4e00\u4e2a\u7ef4\u5ea6<\/h2>\n\n\n\n

\u7814\u7a76\u5458\u4eec\u5c06\u201c\u7c7b\u4eba\u6027\u201d\u62c6\u89e3\u4e3a\u4e09\u4e2a\u6838\u5fc3\u539f\u5219\uff1a\u7406\u6027\uff08Rationality\uff09\u3001\u4e00\u81f4\u6027\uff08Consistency\uff09\u548c\u591a\u6837\u6027\uff08Diversity\uff09\u3002<\/p>\n\n\n\n

\u7406\u6027\uff1a\u8fd9\u4e00\u523b\uff0c\u5b83\u50cf\u4e0d\u50cf\u8fd9\u4e2a\u7528\u6237\uff1f<\/strong><\/p>\n\n\n\n

\u7406\u6027\u5e76\u4e0d\u610f\u5473\u7740\u201c\u7edd\u5bf9\u6b63\u786e\u201d\uff0c\u800c\u662f\u6307 AI \u7684\u884c\u4e3a\u662f\u5426\u7b26\u5408\u8be5\u7528\u6237\u7684\u5386\u53f2\u504f\u597d\u4e0e\u884c\u4e3a\u6a21\u5f0f\u3002\u4f8b\u5982\uff0c\u4e00\u4e2a\u7528\u6237\u8fc7\u53bb\u6301\u7eed\u8d2d\u4e70\u6e29\u99a8\u7231\u60c5\u7247\u548c\u5723\u8bde\u9898\u6750\u7684\u7535\u5f71\uff0c\u90a3\u4e48AI \u540e\u7eed\u7a81\u7136\u63a8\u8350\u79d1\u5e7b\u60ca\u609a\u7247\u3001\u6050\u6016\u7247\u6216\u65e0\u5173\u97f3\u4e50\u89c6\u9891\uff0c\u663e\u7136\u53ef\u80fd\u5c31\u504f\u79bb\u4e86\u8fd9\u4e2a\u4eba\u7684\u5174\u8da3\u8f68\u8ff9\u3002<\/p>\n\n\n\n

\u7814\u7a76\u4ece\u4e24\u4e2a\u89d2\u5ea6\u8861\u91cf\u8fd9\u4e00\u70b9\uff1a<\/p>\n\n\n\n

– \u53ef\u533a\u5206\u6027\uff08Distinguishability\uff09\uff1aAI \u751f\u6210\u7684\u884c\u4e3a\u662f\u5426\u5f88\u5bb9\u6613\u88ab\u8bc6\u522b\u4e3a\u673a\u5668\u884c\u4e3a\uff1b<\/p>\n\n\n\n

– \u53ef\u9884\u6d4b\u6027\uff08Predictability\uff09\uff1aAI \u540e\u7eed\u751f\u6210\u7684\u884c\u4e3a\u662f\u5426\u4e0e\u8be5\u7528\u6237\u5df2\u6709\u753b\u50cf\u76f8\u5339\u914d\u3002<\/p>\n\n\n\n

\u6362\u53e5\u8bdd\u8bf4\uff0cAI \u4e0d\u4ec5\u8981\u201c\u8bf4\u5f97\u901a\u201d\uff0c\u8fd8\u8981\u201c\u50cf\u8fd9\u4e2a\u4eba\u4f1a\u505a\u51fa\u7684\u9009\u62e9\u201d\u3002<\/p>\n\n\n\n

\u4e00\u81f4\u6027\uff1a\u524d\u540e\u8fd8\u662f\u4e0d\u662f\u540c\u4e00\u4e2a\u4eba\uff1f<\/strong><\/p>\n\n\n\n

\u4eba\u7c7b\u7684\u884c\u4e3a\u4f1a\u53d8\u5316\uff0c\u4f46\u901a\u5e38\u5b58\u5728\u7a33\u5b9a\u7684\u5e95\u5c42\u6a21\u5f0f\uff1a\u957f\u671f\u5174\u8da3\u3001\u8868\u8fbe\u98ce\u683c\u3001\u77e5\u8bc6\u8fb9\u754c\u3001\u601d\u7ef4\u4e60\u60ef\uff0c\u4ee5\u53ca\u51fa\u884c\u548c\u6d88\u8d39\u4e2d\u7684\u89c4\u5f8b\u3002<\/p>\n\n\n\n

\u5982\u679c\u5b83\u521a\u521a\u5230\u8fbe\u4e86\u4e00\u5904\u9910\u5385\uff0c\u5341\u51e0\u79d2\u540e\u53c8\u201c\u77ac\u79fb\u201d\u5230\u4e8c\u5341\u516c\u91cc\u5916\u7684\u529e\u516c\u5ba4\uff0c\u90a3\u4e48\u8fd9\u79cd\u884c\u4e3a\u5373\u4f7f\u5355\u72ec\u770b\u4f3c\u4e4e\u5408\u7406\uff0c\u653e\u5728\u8fde\u7eed\u8f68\u8ff9\u4e2d\u4e5f\u4f1a\u663e\u5f97\u4e0d\u53ef\u4fe1\u3002\u56e0\u6b64\uff0c\u7814\u7a76\u8fdb\u4e00\u6b65\u8861\u91cf\uff0c\u4e00\u6bb5 AI \u751f\u6210\u884c\u4e3a\u5e8f\u5217\u7684\u524d\u540e\u4e24\u90e8\u5206\uff0c\u662f\u5426\u4ecd\u7136\u50cf\u540c\u4e00\u4e2a\u7528\u6237\u3002<\/p>\n\n\n\n

\u591a\u6837\u6027\uff1a\u4e00\u7fa4\u00a0AI\uff0c\u771f\u7684\u50cf\u4e00\u7fa4\u4eba\u5417\uff1f<\/strong><\/p>\n\n\n\n

\u8fd9\u6216\u8bb8\u662f\u6700\u5bb9\u6613\u88ab\u5ffd\u7565\u3001\u5374\u5bf9\u793e\u4f1a\u6a21\u62df\u6700\u91cd\u8981\u7684\u4e00\u70b9\u3002\u4e00\u4e2a\u6a21\u578b\u53ef\u4ee5\u4e3a\u67d0\u4e2a\u7528\u6237\u751f\u6210\u5408\u7406\u7b54\u6848\uff0c\u5e76\u4e0d\u610f\u5473\u7740\u5b83\u80fd\u6a21\u62df\u4e00\u4e2a\u771f\u5b9e\u4eba\u7fa4\u3002\u82e5\u4e0d\u540cAI \u7528\u6237\u6700\u7ec8\u90fd\u8868\u73b0\u51fa\u76f8\u4f3c\u504f\u597d\u3001\u76f8\u4f3c\u8868\u8fbe\u548c\u76f8\u4f3c\u63a8\u7406\u8def\u5f84\uff0c\u90a3\u4e48\u770b\u4f3c\u4eba\u6570\u5f88\u591a\u7684\u6a21\u62df\u793e\u4f1a\uff0c\u672c\u8d28\u4e0a\u53ef\u80fd\u53ea\u662f\u540c\u4e00\u79cd\u4eba\u683c\u7684\u91cd\u590d\u590d\u5236\u3002<\/p>\n\n\n\n

\u56e0\u6b64\uff0c\u7814\u7a76\u4ece\u7fa4\u4f53\u5c42\u9762\u6bd4\u8f83\u771f\u5b9e\u7528\u6237\u548c AI \u7528\u6237\u7684\u884c\u4e3a\u5206\u5e03\uff1a\u771f\u5b9e\u7528\u6237\u7fa4\u4f53\u4e2d\u7684\u6bcf\u4e00\u79cd\u504f\u597d\u3001\u98ce\u683c\u548c\u884c\u4e3a\u6a21\u5f0f\uff0c\u80fd\u5426\u5728 AI \u6a21\u62df\u7fa4\u4f53\u4e2d\u627e\u5230\u76f8\u8fd1\u7684\u5bf9\u5e94\u8005\uff1f<\/p>\n\n\n\n

<\/p>\n\n\n\n

\u66f4\u5927\u7684\u6a21\u578b\u66f4\u50cf\u4eba\uff0c\u4f46\u4ecd\u672a\u8fbe\u5230\u771f\u5b9e\u4eba\u7c7b\u6c34\u5e73<\/h2>\n\n\n\n

\u672c\u7814\u7a76\u5de5\u4f5c\u8bc4\u4f30\u4e86 Phi\u3001Mistral\u3001Qwen2.5\u3001Qwen3\u3001Llama-3.1\u548c GPT-4o \u7b49 14 \u4e2a\u4e3b\u6d41\u6a21\u578b\u7248\u672c\u3002<\/p>\n\n\n\n

\u603b\u4f53\u4e0a\uff0c\u6a21\u578b\u89c4\u6a21\u8d8a\u5927\uff0c\u7c7b\u4eba\u884c\u4e3a\u6a21\u62df\u80fd\u529b\u8d8a\u5f3a\u3002GPT-4o \u5728\u7efc\u5408\u8868\u73b0\u4e0a\u9886\u5148\uff0cLlama-3.1 \u548c Qwen\u7cfb\u5217\u7d27\u968f\u5176\u540e\u3002\u4f46\u201c\u6700\u597d\u201d\u5e76\u4e0d\u610f\u5473\u7740\u201c\u8db3\u591f\u50cf\u4eba\u201d\u3002\u5728\u5728\u7ebf\u8d2d\u7269\u3001\u5f00\u653e\u95ee\u7b54\u548c\u57ce\u5e02\u51fa\u884c\u4e09\u4e2a\u573a\u666f\u4e2d\uff0c\u8868\u73b0\u6700\u597d\u7684\u6a21\u578b\u4e0e\u771f\u5b9e\u7528\u6237\u76f8\u6bd4\uff0c\u7efc\u5408\u5f97\u5206\u4ecd\u5206\u522b\u5b58\u5728\u7ea610%\u300117% \u548c 10% \u7684\u5dee\u8ddd\u3002\u5176\u4e2d\uff0c\u5f00\u653e\u5f0f\u95ee\u7b54\u662f\u6700\u96be\u7684\u573a\u666f\u3002<\/p>\n\n\n\n

\u8fd9\u662f\u56e0\u4e3a\uff0c\u5728\u5019\u9009\u96c6\u5185\u9009\u5546\u54c1\u6216\u5730\u70b9\uff0c\u6a21\u578b\u53ea\u9700\u8981\u5728\u6709\u9650\u7a7a\u95f4\u4e2d\u505a\u51fa\u9009\u62e9\uff0c\u800c\u5728\u5f00\u653e\u95ee\u7b54\u4e2d\uff0c\u6a21\u578b\u4e0d\u4ec5\u8981\u6a21\u4eff\u7528\u6237\u7684\u8bed\u8a00\u98ce\u683c\uff0c\u8fd8\u9700\u8981\u590d\u73b0\u5176\u77e5\u8bc6\u8303\u56f4\u3001\u4ef7\u503c\u5224\u65ad\u3001\u63a8\u7406\u4e60\u60ef\u548c\u4e16\u754c\u89c2\u3002\u201c\u5199\u5f97\u901a\u987a\u201d\u8fdc\u8fdc\u4e0d\u7b49\u4e8e\u201c\u8fd9\u4e2a\u4eba\u771f\u7684\u4f1a\u8fd9\u4e48\u60f3\u201d\u3002<\/p>\n\n\n\n

\"table\"
\u88681\uff1a\u4e09\u7c7b\u573a\u666f\u4e0b\u7684\u4e3b\u5b9e\u9a8c\u7ed3\u679c\u3002Cons\u3001Dist\u3001Pred\u3001Div \u548c Avg \u5206\u522b\u8868\u793a\u4e00\u81f4\u6027\uff08Consistency\uff09\u3001\u53ef\u533a\u5206\u6027\uff08Distinguishability\uff09\u3001\u53ef\u9884\u6d4b\u6027\uff08Predictability\uff09\u3001\u591a\u6837\u6027\uff08Diversity\uff09\uff0c\u4ee5\u53ca\u56db\u9879\u6307\u6807\u7684\u5e73\u5747\u5f97\u5206\u3002\u8be5\u7814\u7a76\u8bc4\u4f30\u7684\u6240\u6709\u6a21\u578b\u5747\u4e3a\u6307\u4ee4\u5fae\u8c03\u7248\u672c\u3002\u5bf9\u4e8e\u771f\u5b9e\u7528\u6237\uff0c\u7814\u7a76\u5458\u4eec\u76f4\u63a5\u4f7f\u7528\u5176\u771f\u5b9e\u884c\u4e3a\u800c\u975eLLMs \u751f\u6210\u7684\u6a21\u62df\u884c\u4e3a\u6765\u8ba1\u7b97\u5404\u9879\u6307\u6807\u3002<\/figcaption><\/figure>\n\n\n\n

AI \u4f1a\u4e0d\u4f1a\u628a\u5343\u4e07\u4e2a\u7528\u6237\uff0c\u6a21\u62df\u6210\u201c\u540c\u4e00\u79cd\u4eba\u201d\uff1f<\/h2>\n\n\n\n

\u5728\u7fa4\u4f53\u5c42\u9762\u7684\u53ef\u89c6\u5316\u5206\u6790\u4e2d\uff0c\u7814\u7a76\u5458\u4eec\u53d1\u73b0\u4e86\u4e00\u4e2a\u503c\u5f97\u5173\u6ce8\u7684\u73b0\u8c61\uff1a\u5206\u5e03\u574d\u7f29\u3002<\/p>\n\n\n\n

\u8f83\u5c0f\u6a21\u578b\u5728\u6a21\u62df\u4e0d\u540c\u7528\u6237\u65f6\uff0c\u5f80\u5f80\u4f1a\u751f\u6210\u8f83\u4e3a\u96c6\u4e2d\u7684\u884c\u4e3a\u6a21\u5f0f\u3002\u4e0d\u540cAI \u7528\u6237\u770b\u4f3c\u62e5\u6709\u4e0d\u540c\u753b\u50cf\uff0c\u4f46\u6700\u7ec8\u5374\u53ef\u80fd\u7ed9\u51fa\u9ad8\u5ea6\u76f8\u4f3c\u7684\u504f\u597d\u3001\u89c2\u70b9\u548c\u56de\u7b54\u3002\u968f\u7740\u6a21\u578b\u89c4\u6a21\u589e\u5927\uff0c\u8fd9\u79cd\u73b0\u8c61\u4f1a\u6709\u6240\u7f13\u89e3\uff1a\u6a21\u62df\u7528\u6237\u5728\u884c\u4e3a\u7a7a\u95f4\u4e2d\u7684\u5206\u5e03\u53d8\u5f97\u66f4\u5e7f\uff0c\u4e5f\u66f4\u63a5\u8fd1\u771f\u5b9e\u7528\u6237\u7fa4\u4f53\u3002\u4f46\u5373\u4f7f\u662f\u66f4\u5f3a\u7684\u5927\u6a21\u578b\uff0c\u4f9d\u7136\u65e0\u6cd5\u5b8c\u5168\u8986\u76d6\u771f\u5b9e\u4eba\u7fa4\u7684\u4e30\u5bcc\u6027\u3002<\/p>\n\n\n\n

\"chart,
\u56fe2\uff1a\u771f\u5b9e\u7528\u6237\u4e0e\u6a21\u62df\u7528\u6237\u7684\u5d4c\u5165\u5206\u5e03\u53ef\u89c6\u5316<\/figcaption><\/figure>\n\n\n\n

\u800c\u8fd9\u5bf9\u4e8e\u793e\u4f1a\u4eff\u771f\u3001\u7528\u6237\u6a21\u62df\u548c\u4ea7\u54c1\u6d4b\u8bd5\u5c24\u5176\u91cd\u8981\u3002\u5982\u679c AI \u7528\u6237\u7fa4\u4f53\u7f3a\u5c11\u771f\u5b9e\u591a\u6837\u6027\uff0c\u90a3\u4e48\u57fa\u4e8e\u8fd9\u4e9b\u6a21\u62df\u4eba\u7fa4\u5f97\u5230\u7684\u7ed3\u8bba\uff0c\u53ef\u80fd\u4f1a\u9ad8\u4f30\u67d0\u4e9b\u504f\u597d\u3001\u5ffd\u89c6\u5c11\u6570\u7fa4\u4f53\uff0c\u751a\u81f3\u5bfc\u5411\u9519\u8bef\u7684\u4ea7\u54c1\u4e0e\u793e\u4f1a\u5224\u65ad\u3002<\/p>\n\n\n\n

\u4ece\u7b54\u5bf9\u95ee\u9898\uff0c\u8d70\u5411\u62df\u5408\u4eba\u7c7b\u884c\u4e3a\u5206\u5e03<\/h2>\n\n\n\n

\u672c\u7bc7\u8bba\u6587\u7814\u7a76\u7684\u6838\u5fc3\u610f\u4e49\uff0c\u4e0d\u53ea\u662f\u63d0\u51fa\u4e86\u4e00\u5957\u65b0\u7684\u8bc4\u6d4b\u6307\u6807\u3002\u5b83\u5c1d\u8bd5\u628a\u4e00\u4e2a\u957f\u671f\u88ab\u7b80\u5316\u7684\u95ee\u9898\u91cd\u65b0\u5c55\u5f00\uff1a\u8bc4\u4f30AI \u662f\u5426\u50cf\u4eba\uff0c\u4e0d\u80fd\u53ea\u770b\u5b83\u80fd\u5426\u901a\u8fc7\u4e00\u6bb5\u77ed\u5bf9\u8bdd\uff0c\u4e5f\u4e0d\u80fd\u53ea\u770b\u5b83\u80fd\u5426\u7b54\u5bf9\u4e00\u9053\u9898\u3002\u66f4\u91cd\u8981\u7684\u662f\uff0c\u5b83\u80fd\u5426\u5728\u771f\u5b9e\u3001\u5f00\u653e\u3001\u957f\u671f\u7684\u884c\u4e3a\u8fc7\u7a0b\u4e2d\uff0c\u8868\u73b0\u51fa\u7a33\u5b9a\u3001\u5408\u7406\u4e14\u591a\u6837\u7684\u4eba\u7c7b\u7279\u5f81\u3002\u8fd9\u5bf9\u672a\u6765\u7684\u6570\u5b57\u4eba\u3001\u667a\u80fdNPC\u3001\u4e2a\u6027\u5316\u52a9\u624b\u3001\u7528\u6237\u6a21\u62df\u5668\u548c\u793e\u4f1a\u4eff\u771f\u7cfb\u7edf\u800c\u8a00\uff0c\u662f\u4e00\u9879\u57fa\u7840\u80fd\u529b\u3002<\/p>\n\n\n\n

\u5f53\u7136\uff0c\u7c7b\u4eba\u884c\u4e3a\u6a21\u62df\u7684\u76ee\u6807\u5e76\u4e0d\u662f\u8ba9 AI \u5728\u6240\u6709\u573a\u666f\u4e2d\u4f2a\u88c5\u6210\u4eba\uff0c\u800c\u662f\u5e2e\u52a9\u79d1\u7814\u4eba\u5458\u66f4\u6e05\u695a\u5730\u7406\u89e3\uff0c\u5f53 AI \u88ab\u7528\u4e8e\u6a21\u62df\u3001\u9884\u6d4b\u6216\u670d\u52a1\u4eba\u7c7b\u65f6\uff0c\u5b83\u7684\u80fd\u529b\u8fb9\u754c\u7a76\u7adf\u5728\u54ea\u91cc\u3002<\/p>\n\n\n\n

\u4ece\u201c\u50cf\u4eba\u8bf4\u8bdd\u201d\uff0c\u5230\u201c\u50cf\u4eba\u884c\u52a8\u201d\uff0c\u518d\u5230\u201c\u50cf\u4e00\u7fa4\u4e0d\u540c\u7684\u4eba\u5171\u540c\u751f\u6d3b\u201d\u2014\u2014\u5927\u6a21\u578b\u8ddd\u79bb\u771f\u6b63\u53ef\u4fe1\u7684\u4eba\u7c7b\u884c\u4e3a\u6a21\u62df\uff0c\u4ecd\u5728\u8def\u4e0a\u3002<\/p>\n\n\n\n

A Computational Framework for Evaluating Human-likeness in LLMs\u2019 Open-ended Human Behaviors (ICML 2026)<\/strong><\/p>\n\n\n\n

\u8bba\u6587\u94fe\u63a5\uff1ahttps:\/\/icml.cc\/virtual\/2026\/poster\/65092 (opens in new tab)<\/span><\/a><\/p>\n\n\n\n

<\/p>\n","protected":false},"excerpt":{"rendered":"

\u5f53\u6211\u4eec\u8ba9 AI \u626e\u6f14\u4e00\u4f4d\u7528\u6237\u65f6\uff0c\u5b83\u901a\u5e38\u80fd\u591f\u5feb\u901f\u7ed9\u51fa\u50cf\u6a21\u50cf\u6837\u7684\u56de\u7b54\u3002\u5b83\u53ef\u4ee5\u6a21\u4eff\u4e00\u4e2a\u5f71\u8ff7\u6311\u9009\u7535\u5f71\uff0c\u6a21\u4eff\u4e00\u4f4d\u7f51\u53cb\u8ba8\u8bba\u95ee\u9898\uff0c\u4e5f\u53ef\u4ee5\u6839\u636e\u5386\u53f2\u8db3\u8ff9\u9884\u6d4b\u4e00\u4e2a\u4eba\u4e0b\u4e00\u7ad9\u4f1a\u53bb\u54ea\u91cc\u3002 \u968f\u7740\u5927\u6a21\u578b\u80fd\u529b\u7684\u5feb\u901f\u63d0\u5347\uff0c\u667a\u80fd NPC\u3001\u7528\u6237\u6a21\u62df\u5668\u3001\u793e\u4f1a\u4eff\u771f\u7b49\u5e94\u7528\u6b63\u5728\u53d8\u5f97\u8d8a\u6765\u8d8a\u771f\u5b9e\u3002\u4f46\u4e00\u4e2a\u66f4\u5173\u952e\u7684\u95ee\u9898\u662f\uff1aAI\u751f\u6210\u7684\u884c\u4e3a\uff0c\u771f\u7684\u50cf\u4eba\u5417\uff1f \u5b83\u662f\u5426\u771f\u7684\u7406\u89e3\u4e86\u4e00\u4e2a\u4eba\u7684\u504f\u597d\u3001\u601d\u7ef4\u65b9\u5f0f\u548c\u957f\u671f\u4e60\u60ef\uff1f\u5b83\u80fd\u5426\u5728\u591a\u8f6e\u4ea4\u4e92\u540e\u4f9d\u7136\u4fdd\u6301\u201c\u540c\u4e00\u4e2a\u4eba\u201d\u7684\u4e00\u81f4\u6027\uff1f\u5f53\u6a21\u62df\u6210\u5343\u4e0a\u4e07\u4e2a\u7528\u6237\u65f6\uff0c\u8fd9\u4e9b AI \u53c8\u80fd\u5426\u50cf\u771f\u5b9e\u793e\u4f1a\u4e00\u6837\u5448\u73b0\u4e30\u5bcc\u3001\u591a\u6837\u3001\u751a\u81f3\u5f7c\u6b64\u4e0d\u540c\u7684\u4eba\u7fa4\u5206\u5e03\uff1f \u8fd1\u671f\uff0c\u5fae\u8f6f\u4e9a\u6d32\u7814\u7a76\u9662\u7684\u7814\u7a76\u5458\u4eec\u63d0\u51fa\u4e86\u4e00\u5957\u9762\u5411\u5927\u6a21\u578b\u7c7b\u4eba\u884c\u4e3a\u7684\u8ba1\u7b97\u8bc4\u6d4b\u6846\u67b6\uff0c\u5c1d\u8bd5\u56de\u7b54\u8fd9\u4e2a\u95ee\u9898\u3002\u7814\u7a76\u53d1\u73b0\uff1a\u5927\u6a21\u578b\u5df2\u7ecf\u80fd\u591f\u751f\u6210\u4e0d\u5c11\u770b\u4f3c\u5408\u7406\u7684\u4e2a\u4f53\u884c\u4e3a\uff0c\u4f46\u8ddd\u79bb\u771f\u6b63\u7a33\u5b9a\u3001\u53ef\u4fe1\u3001\u4e30\u5bcc\u7684\u4eba\u7c7b\u884c\u4e3a\u6a21\u62df\uff0c\u4ecd\u6709\u5f88\u957f\u7684\u8def\u8981\u8d70\u3002\u76f8\u5173\u8bba\u6587\u5df2\u88abICML 2026\u63a5\u6536\u3002 \u8bba\u6587\u4fe1\u606f\u5df2\u6574\u7406\u81f3\u6587\u672b\uff0c\u6b22\u8fce\u70b9\u51fb\u76f8\u5173\u94fe\u63a5\uff0c\u4e86\u89e3\u66f4\u591a\u6280\u672f\u8be6\u60c5\u3002 \u4f20\u7edf\u56fe\u7075\u6d4b\u8bd5\u5173\u6ce8\u7684\u662f\uff0c\u4eba\u7c7b\u80fd\u5426\u5728\u77ed\u65f6\u95f4\u5bf9\u8bdd\u4e2d\u5206\u8fa8\u51fa\u5bf9\u65b9\u662f\u4eba\u8fd8\u662f\u673a\u5668\u3002 \u8fd9\u7c7b\u6d4b\u8bd5\u5f88\u91cd\u8981\uff0c\u4f46\u4e5f\u5b58\u5728\u5929\u7136\u5c40\u9650\u3002\u9996\u5148\uff0c\u77ed\u77ed\u51e0\u5206\u949f\u7684\u804a\u5929\uff0c\u53ea\u80fd\u770b\u5230 AI\u884c\u4e3a\u7684\u4e00\u4e2a\u5207\u9762\u3002\u4e00\u4e2a\u6a21\u578b\u6216\u8bb8\u80fd\u591f\u5728\u4e00\u8f6e\u5bf9\u8bdd\u4e2d\u663e\u5f97\u81ea\u7136\uff0c\u4f46\u672a\u5fc5\u80fd\u5728\u51e0\u5341\u8f6e\u9009\u62e9\u3001\u56de\u7b54\u6216\u884c\u52a8\u4e2d\u6301\u7eed\u4fdd\u6301\u4eba\u683c\u3001\u504f\u597d\u548c\u903b\u8f91\u7684\u4e00\u81f4\u3002 \u66f4\u91cd\u8981\u7684\u662f\uff0c\u8bb8\u591a\u771f\u5b9e\u4e16\u754c\u4e2d\u7684\u884c\u4e3a\u5e76\u4e0d\u5b58\u5728\u552f\u4e00\u6b63\u786e\u7b54\u6848\u3002\u5373\u4f7f\u662f\u540c\u4e00\u4e2a\u7528\u6237\u5728\u76f8\u4f3c\u60c5\u5883\u4e0b\u91cd\u590d\u8fdb\u884c\u9009\u62e9\u6216\u56de\u7b54\uff0c\u7ed3\u679c\u4e5f\u672a\u5fc5\u5b8c\u5168\u4e00\u81f4\u3002\u4ed6\u53ef\u80fd\u5728\u4e24\u90e8\u90fd\u611f\u5174\u8da3\u7684\u7535\u5f71\u95f4\u505a\u51fa\u4e0d\u540c\u9009\u62e9\uff0c\u4e5f\u53ef\u80fd\u4ee5\u4e0d\u540c\u7684\u63aa\u8f9e\u3001\u89d2\u5ea6\u6216\u8bba\u8bc1\u8def\u5f84\u56de\u7b54\u540c\u4e00\u4e2a\u95ee\u9898\u3002\u6b64\u65f6\uff0c\u8bc4\u4f30\u7684\u5173\u952e\u4e0d\u518d\u662f\u201c\u7b54\u5bf9\u4e86\u5417\u201d\uff0c\u800c\u662f AI \u751f\u6210\u7684\u884c\u4e3a\u662f\u5426\u4ecd\u7136\u7b26\u5408\u8fd9\u4e2a\u4eba\u53ef\u80fd\u5448\u73b0\u7684\u504f\u597d\u4e0e\u884c\u4e3a\u5206\u5e03\u3002 \u4e3a\u6b64\uff0c\u7814\u7a76\u5458\u4eec\u6ca1\u6709\u4e3a\u6a21\u578b\u8bbe\u8ba1\u4e00\u5957\u4eba\u5de5\u7f16\u5199\u7684\u8003\u9898\uff0c\u800c\u662f\u76f4\u63a5\u4ece\u771f\u5b9e\u4e16\u754c\u7684\u516c\u5f00\u884c\u4e3a\u6570\u636e\u4e2d\u5bfb\u627e\u7b54\u6848\u3002\u8be5\u7814\u7a76\u5de5\u4f5c\u8986\u76d6\u4e86\u4e09\u7c7b\u5178\u578b\u573a\u666f\uff1a – \u5728\u7ebf\u8d2d\u7269\uff1a\u7528\u6237\u5728\u7535\u5546\u5e73\u53f0\u4e2d\u7684\u8fde\u7eed\u8d2d\u4e70\u4e0e\u6d4f\u89c8\u504f\u597d\uff1b – \u5f00\u653e\u95ee\u7b54\uff1a\u7528\u6237\u5728\u5f00\u653e\u793e\u533a\u4e2d\u7684\u56de\u7b54\u3001\u89c2\u70b9\u4e0e\u8868\u8fbe\u65b9\u5f0f\uff1b – \u57ce\u5e02\u51fa\u884c\uff1a\u7528\u6237\u5728\u57fa\u4e8e\u4f4d\u7f6e\u670d\u52a1\u7684\u5e73\u53f0\u4e0a\u7559\u4e0b\u7684\u65f6\u7a7a\u7b7e\u5230\u8f68\u8ff9\u3002 \u5728\u6bcf\u4e2a\u573a\u666f\u4e2d\uff0c\u7814\u7a76\u5458\u4eec\u90fd\u5148\u7528\u7528\u6237\u5386\u53f2\u884c\u4e3a\u6784\u5efa\u7528\u6237\u753b\u50cf\uff0c\u518d\u8ba9\u5927\u6a21\u578b\u6a21\u62df\u8be5\u7528\u6237\u540e\u7eed\u7684\u591a\u8f6e\u884c\u4e3a\u3002\u968f\u540e\uff0c\u7cfb\u7edf\u5c06AI \u7684\u6a21\u62df\u7ed3\u679c\u4e0e\u771f\u5b9e\u7528\u6237\u884c\u4e3a\u8fdb\u884c\u5bf9\u6bd4\u3002 \u6574\u4e2a\u6846\u67b6\u7531\u56db\u90e8\u5206\u7ec4\u6210\uff1a\u771f\u5b9e\u884c\u4e3a\u6570\u636e\u3001LLM\u6a21\u62df\u5668\u3001\u7528\u6237\u884c\u4e3a\u7f16\u7801\u5668\u548c\u8bc4\u6d4b\u5668\u3002\u5176\u4e2d\uff0c\u771f\u5b9e\u884c\u4e3a\u6570\u636e\u63d0\u4f9b\u88ab\u6a21\u62df\u7528\u6237\u7684\u5386\u53f2\u8f68\u8ff9\uff0c\u4ee5\u53ca\u7528\u4e8e\u5bf9\u7167\u7684\u771f\u5b9e\u540e\u7eed\u884c\u4e3a\uff1bLLM\u6a21\u62df\u5668\u6839\u636e\u7528\u6237\u753b\u50cf\u548c\u5f53\u524d\u60c5\u5883\uff0c\u9010\u6b65\u751f\u6210\u8be5\u7528\u6237\u53ef\u80fd\u505a\u51fa\u7684\u540e\u7eed\u9009\u62e9\u3001\u56de\u7b54\u6216\u884c\u52a8\uff1b\u7528\u6237\u884c\u4e3a\u7f16\u7801\u5668\u5219\u4ece\u5927\u91cf\u771f\u5b9e\u6570\u636e\u4e2d\u5b66\u4e60\u201c\u4ec0\u4e48\u6837\u7684\u8fde\u7eed\u884c\u4e3a\u5c5e\u4e8e\u540c\u4e00\u4e2a\u4eba\u201d\uff0c\u5c06\u884c\u4e3a\u5e8f\u5217\u6620\u5c04\u4e3a\u80fd\u591f\u8868\u5f81\u504f\u597d\u3001\u98ce\u683c\u4e0e\u884c\u4e3a\u6a21\u5f0f\u7684\u9ad8\u7ef4\u5411\u91cf\uff1b\u6700\u540e\uff0c\u8bc4\u6d4b\u5668\u5728\u8fd9\u4e9b\u8868\u5f81\u4e4b\u4e0a\uff0c\u4ece\u591a\u4e2a\u7ef4\u5ea6\u5224\u65adAI \u884c\u4e3a\u7684\u7c7b\u4eba\u6027\u3002 \u672c\u6846\u67b6\u7684\u6838\u5fc3\u8bbe\u8ba1\u4e0d\u4f9d\u8d56\u7279\u5b9a\u4efb\u52a1\u5f62\u5f0f\uff0c\u53ea\u8981\u5177\u5907\u8db3\u591f\u7684\u771f\u5b9e\u7528\u6237\u884c\u4e3a\u6570\u636e\uff0c\u4fbf\u53ef\u81ea\u7136\u6269\u5c55\u81f3\u8bc4\u8bba\u64b0\u5199\u3001\u65e5\u8bb0\u521b\u4f5c\u3001\u996e\u98df\u9009\u62e9\u7b49\u66f4\u591a\u5f00\u653e\u5f0f\u884c\u4e3a\u573a\u666f\u3002\u4e0e\u201c\u770b\u4e00\u6761\u56de\u7b54\u50cf\u4e0d\u50cf\u4eba\u201d\u4e0d\u540c\uff0c\u8fd9\u5957\u6846\u67b6\u5e0c\u671b\u4ece\u4e2a\u4f53\u5230\u7fa4\u4f53\u3001\u4ece\u5373\u65f6\u884c\u4e3a\u5230\u957f\u671f\u8f68\u8ff9\uff0c\u5168\u9762\u8bc4\u4f30AI \u7684\u7c7b\u4eba\u7a0b\u5ea6\u3002 \u7814\u7a76\u5458\u4eec\u5c06\u201c\u7c7b\u4eba\u6027\u201d\u62c6\u89e3\u4e3a\u4e09\u4e2a\u6838\u5fc3\u539f\u5219\uff1a\u7406\u6027\uff08Rationality\uff09\u3001\u4e00\u81f4\u6027\uff08Consistency\uff09\u548c\u591a\u6837\u6027\uff08Diversity\uff09\u3002 \u7406\u6027\uff1a\u8fd9\u4e00\u523b\uff0c\u5b83\u50cf\u4e0d\u50cf\u8fd9\u4e2a\u7528\u6237\uff1f \u7406\u6027\u5e76\u4e0d\u610f\u5473\u7740\u201c\u7edd\u5bf9\u6b63\u786e\u201d\uff0c\u800c\u662f\u6307 AI \u7684\u884c\u4e3a\u662f\u5426\u7b26\u5408\u8be5\u7528\u6237\u7684\u5386\u53f2\u504f\u597d\u4e0e\u884c\u4e3a\u6a21\u5f0f\u3002\u4f8b\u5982\uff0c\u4e00\u4e2a\u7528\u6237\u8fc7\u53bb\u6301\u7eed\u8d2d\u4e70\u6e29\u99a8\u7231\u60c5\u7247\u548c\u5723\u8bde\u9898\u6750\u7684\u7535\u5f71\uff0c\u90a3\u4e48AI \u540e\u7eed\u7a81\u7136\u63a8\u8350\u79d1\u5e7b\u60ca\u609a\u7247\u3001\u6050\u6016\u7247\u6216\u65e0\u5173\u97f3\u4e50\u89c6\u9891\uff0c\u663e\u7136\u53ef\u80fd\u5c31\u504f\u79bb\u4e86\u8fd9\u4e2a\u4eba\u7684\u5174\u8da3\u8f68\u8ff9\u3002 \u7814\u7a76\u4ece\u4e24\u4e2a\u89d2\u5ea6\u8861\u91cf\u8fd9\u4e00\u70b9\uff1a – \u53ef\u533a\u5206\u6027\uff08Distinguishability\uff09\uff1aAI \u751f\u6210\u7684\u884c\u4e3a\u662f\u5426\u5f88\u5bb9\u6613\u88ab\u8bc6\u522b\u4e3a\u673a\u5668\u884c\u4e3a\uff1b – \u53ef\u9884\u6d4b\u6027\uff08Predictability\uff09\uff1aAI \u540e\u7eed\u751f\u6210\u7684\u884c\u4e3a\u662f\u5426\u4e0e\u8be5\u7528\u6237\u5df2\u6709\u753b\u50cf\u76f8\u5339\u914d\u3002 \u6362\u53e5\u8bdd\u8bf4\uff0cAI \u4e0d\u4ec5\u8981\u201c\u8bf4\u5f97\u901a\u201d\uff0c\u8fd8\u8981\u201c\u50cf\u8fd9\u4e2a\u4eba\u4f1a\u505a\u51fa\u7684\u9009\u62e9\u201d\u3002 \u4e00\u81f4\u6027\uff1a\u524d\u540e\u8fd8\u662f\u4e0d\u662f\u540c\u4e00\u4e2a\u4eba\uff1f \u4eba\u7c7b\u7684\u884c\u4e3a\u4f1a\u53d8\u5316\uff0c\u4f46\u901a\u5e38\u5b58\u5728\u7a33\u5b9a\u7684\u5e95\u5c42\u6a21\u5f0f\uff1a\u957f\u671f\u5174\u8da3\u3001\u8868\u8fbe\u98ce\u683c\u3001\u77e5\u8bc6\u8fb9\u754c\u3001\u601d\u7ef4\u4e60\u60ef\uff0c\u4ee5\u53ca\u51fa\u884c\u548c\u6d88\u8d39\u4e2d\u7684\u89c4\u5f8b\u3002 \u5982\u679c\u5b83\u521a\u521a\u5230\u8fbe\u4e86\u4e00\u5904\u9910\u5385\uff0c\u5341\u51e0\u79d2\u540e\u53c8\u201c\u77ac\u79fb\u201d\u5230\u4e8c\u5341\u516c\u91cc\u5916\u7684\u529e\u516c\u5ba4\uff0c\u90a3\u4e48\u8fd9\u79cd\u884c\u4e3a\u5373\u4f7f\u5355\u72ec\u770b\u4f3c\u4e4e\u5408\u7406\uff0c\u653e\u5728\u8fde\u7eed\u8f68\u8ff9\u4e2d\u4e5f\u4f1a\u663e\u5f97\u4e0d\u53ef\u4fe1\u3002\u56e0\u6b64\uff0c\u7814\u7a76\u8fdb\u4e00\u6b65\u8861\u91cf\uff0c\u4e00\u6bb5 AI \u751f\u6210\u884c\u4e3a\u5e8f\u5217\u7684\u524d\u540e\u4e24\u90e8\u5206\uff0c\u662f\u5426\u4ecd\u7136\u50cf\u540c\u4e00\u4e2a\u7528\u6237\u3002 \u591a\u6837\u6027\uff1a\u4e00\u7fa4\u00a0AI\uff0c\u771f\u7684\u50cf\u4e00\u7fa4\u4eba\u5417\uff1f \u8fd9\u6216\u8bb8\u662f\u6700\u5bb9\u6613\u88ab\u5ffd\u7565\u3001\u5374\u5bf9\u793e\u4f1a\u6a21\u62df\u6700\u91cd\u8981\u7684\u4e00\u70b9\u3002\u4e00\u4e2a\u6a21\u578b\u53ef\u4ee5\u4e3a\u67d0\u4e2a\u7528\u6237\u751f\u6210\u5408\u7406\u7b54\u6848\uff0c\u5e76\u4e0d\u610f\u5473\u7740\u5b83\u80fd\u6a21\u62df\u4e00\u4e2a\u771f\u5b9e\u4eba\u7fa4\u3002\u82e5\u4e0d\u540cAI \u7528\u6237\u6700\u7ec8\u90fd\u8868\u73b0\u51fa\u76f8\u4f3c\u504f\u597d\u3001\u76f8\u4f3c\u8868\u8fbe\u548c\u76f8\u4f3c\u63a8\u7406\u8def\u5f84\uff0c\u90a3\u4e48\u770b\u4f3c\u4eba\u6570\u5f88\u591a\u7684\u6a21\u62df\u793e\u4f1a\uff0c\u672c\u8d28\u4e0a\u53ef\u80fd\u53ea\u662f\u540c\u4e00\u79cd\u4eba\u683c\u7684\u91cd\u590d\u590d\u5236\u3002 \u56e0\u6b64\uff0c\u7814\u7a76\u4ece\u7fa4\u4f53\u5c42\u9762\u6bd4\u8f83\u771f\u5b9e\u7528\u6237\u548c AI \u7528\u6237\u7684\u884c\u4e3a\u5206\u5e03\uff1a\u771f\u5b9e\u7528\u6237\u7fa4\u4f53\u4e2d\u7684\u6bcf\u4e00\u79cd\u504f\u597d\u3001\u98ce\u683c\u548c\u884c\u4e3a\u6a21\u5f0f\uff0c\u80fd\u5426\u5728 AI \u6a21\u62df\u7fa4\u4f53\u4e2d\u627e\u5230\u76f8\u8fd1\u7684\u5bf9\u5e94\u8005\uff1f \u672c\u7814\u7a76\u5de5\u4f5c\u8bc4\u4f30\u4e86 Phi\u3001Mistral\u3001Qwen2.5\u3001Qwen3\u3001Llama-3.1\u548c GPT-4o \u7b49 14 \u4e2a\u4e3b\u6d41\u6a21\u578b\u7248\u672c\u3002 \u603b\u4f53\u4e0a\uff0c\u6a21\u578b\u89c4\u6a21\u8d8a\u5927\uff0c\u7c7b\u4eba\u884c\u4e3a\u6a21\u62df\u80fd\u529b\u8d8a\u5f3a\u3002GPT-4o \u5728\u7efc\u5408\u8868\u73b0\u4e0a\u9886\u5148\uff0cLlama-3.1 \u548c Qwen\u7cfb\u5217\u7d27\u968f\u5176\u540e\u3002\u4f46\u201c\u6700\u597d\u201d\u5e76\u4e0d\u610f\u5473\u7740\u201c\u8db3\u591f\u50cf\u4eba\u201d\u3002\u5728\u5728\u7ebf\u8d2d\u7269\u3001\u5f00\u653e\u95ee\u7b54\u548c\u57ce\u5e02\u51fa\u884c\u4e09\u4e2a\u573a\u666f\u4e2d\uff0c\u8868\u73b0\u6700\u597d\u7684\u6a21\u578b\u4e0e\u771f\u5b9e\u7528\u6237\u76f8\u6bd4\uff0c\u7efc\u5408\u5f97\u5206\u4ecd\u5206\u522b\u5b58\u5728\u7ea610%\u300117% \u548c 10% \u7684\u5dee\u8ddd\u3002\u5176\u4e2d\uff0c\u5f00\u653e\u5f0f\u95ee\u7b54\u662f\u6700\u96be\u7684\u573a\u666f\u3002 \u8fd9\u662f\u56e0\u4e3a\uff0c\u5728\u5019\u9009\u96c6\u5185\u9009\u5546\u54c1\u6216\u5730\u70b9\uff0c\u6a21\u578b\u53ea\u9700\u8981\u5728\u6709\u9650\u7a7a\u95f4\u4e2d\u505a\u51fa\u9009\u62e9\uff0c\u800c\u5728\u5f00\u653e\u95ee\u7b54\u4e2d\uff0c\u6a21\u578b\u4e0d\u4ec5\u8981\u6a21\u4eff\u7528\u6237\u7684\u8bed\u8a00\u98ce\u683c\uff0c\u8fd8\u9700\u8981\u590d\u73b0\u5176\u77e5\u8bc6\u8303\u56f4\u3001\u4ef7\u503c\u5224\u65ad\u3001\u63a8\u7406\u4e60\u60ef\u548c\u4e16\u754c\u89c2\u3002\u201c\u5199\u5f97\u901a\u987a\u201d\u8fdc\u8fdc\u4e0d\u7b49\u4e8e\u201c\u8fd9\u4e2a\u4eba\u771f\u7684\u4f1a\u8fd9\u4e48\u60f3\u201d\u3002 \u5728\u7fa4\u4f53\u5c42\u9762\u7684\u53ef\u89c6\u5316\u5206\u6790\u4e2d\uff0c\u7814\u7a76\u5458\u4eec\u53d1\u73b0\u4e86\u4e00\u4e2a\u503c\u5f97\u5173\u6ce8\u7684\u73b0\u8c61\uff1a\u5206\u5e03\u574d\u7f29\u3002 \u8f83\u5c0f\u6a21\u578b\u5728\u6a21\u62df\u4e0d\u540c\u7528\u6237\u65f6\uff0c\u5f80\u5f80\u4f1a\u751f\u6210\u8f83\u4e3a\u96c6\u4e2d\u7684\u884c\u4e3a\u6a21\u5f0f\u3002\u4e0d\u540cAI \u7528\u6237\u770b\u4f3c\u62e5\u6709\u4e0d\u540c\u753b\u50cf\uff0c\u4f46\u6700\u7ec8\u5374\u53ef\u80fd\u7ed9\u51fa\u9ad8\u5ea6\u76f8\u4f3c\u7684\u504f\u597d\u3001\u89c2\u70b9\u548c\u56de\u7b54\u3002\u968f\u7740\u6a21\u578b\u89c4\u6a21\u589e\u5927\uff0c\u8fd9\u79cd\u73b0\u8c61\u4f1a\u6709\u6240\u7f13\u89e3\uff1a\u6a21\u62df\u7528\u6237\u5728\u884c\u4e3a\u7a7a\u95f4\u4e2d\u7684\u5206\u5e03\u53d8\u5f97\u66f4\u5e7f\uff0c\u4e5f\u66f4\u63a5\u8fd1\u771f\u5b9e\u7528\u6237\u7fa4\u4f53\u3002\u4f46\u5373\u4f7f\u662f\u66f4\u5f3a\u7684\u5927\u6a21\u578b\uff0c\u4f9d\u7136\u65e0\u6cd5\u5b8c\u5168\u8986\u76d6\u771f\u5b9e\u4eba\u7fa4\u7684\u4e30\u5bcc\u6027\u3002 \u800c\u8fd9\u5bf9\u4e8e\u793e\u4f1a\u4eff\u771f\u3001\u7528\u6237\u6a21\u62df\u548c\u4ea7\u54c1\u6d4b\u8bd5\u5c24\u5176\u91cd\u8981\u3002\u5982\u679c AI \u7528\u6237\u7fa4\u4f53\u7f3a\u5c11\u771f\u5b9e\u591a\u6837\u6027\uff0c\u90a3\u4e48\u57fa\u4e8e\u8fd9\u4e9b\u6a21\u62df\u4eba\u7fa4\u5f97\u5230\u7684\u7ed3\u8bba\uff0c\u53ef\u80fd\u4f1a\u9ad8\u4f30\u67d0\u4e9b\u504f\u597d\u3001\u5ffd\u89c6\u5c11\u6570\u7fa4\u4f53\uff0c\u751a\u81f3\u5bfc\u5411\u9519\u8bef\u7684\u4ea7\u54c1\u4e0e\u793e\u4f1a\u5224\u65ad\u3002 \u672c\u7bc7\u8bba\u6587\u7814\u7a76\u7684\u6838\u5fc3\u610f\u4e49\uff0c\u4e0d\u53ea\u662f\u63d0\u51fa\u4e86\u4e00\u5957\u65b0\u7684\u8bc4\u6d4b\u6307\u6807\u3002\u5b83\u5c1d\u8bd5\u628a\u4e00\u4e2a\u957f\u671f\u88ab\u7b80\u5316\u7684\u95ee\u9898\u91cd\u65b0\u5c55\u5f00\uff1a\u8bc4\u4f30AI \u662f\u5426\u50cf\u4eba\uff0c\u4e0d\u80fd\u53ea\u770b\u5b83\u80fd\u5426\u901a\u8fc7\u4e00\u6bb5\u77ed\u5bf9\u8bdd\uff0c\u4e5f\u4e0d\u80fd\u53ea\u770b\u5b83\u80fd\u5426\u7b54\u5bf9\u4e00\u9053\u9898\u3002\u66f4\u91cd\u8981\u7684\u662f\uff0c\u5b83\u80fd\u5426\u5728\u771f\u5b9e\u3001\u5f00\u653e\u3001\u957f\u671f\u7684\u884c\u4e3a\u8fc7\u7a0b\u4e2d\uff0c\u8868\u73b0\u51fa\u7a33\u5b9a\u3001\u5408\u7406\u4e14\u591a\u6837\u7684\u4eba\u7c7b\u7279\u5f81\u3002\u8fd9\u5bf9\u672a\u6765\u7684\u6570\u5b57\u4eba\u3001\u667a\u80fdNPC\u3001\u4e2a\u6027\u5316\u52a9\u624b\u3001\u7528\u6237\u6a21\u62df\u5668\u548c\u793e\u4f1a\u4eff\u771f\u7cfb\u7edf\u800c\u8a00\uff0c\u662f\u4e00\u9879\u57fa\u7840\u80fd\u529b\u3002 \u5f53\u7136\uff0c\u7c7b\u4eba\u884c\u4e3a\u6a21\u62df\u7684\u76ee\u6807\u5e76\u4e0d\u662f\u8ba9 AI \u5728\u6240\u6709\u573a\u666f\u4e2d\u4f2a\u88c5\u6210\u4eba\uff0c\u800c\u662f\u5e2e\u52a9\u79d1\u7814\u4eba\u5458\u66f4\u6e05\u695a\u5730\u7406\u89e3\uff0c\u5f53 AI \u88ab\u7528\u4e8e\u6a21\u62df\u3001\u9884\u6d4b\u6216\u670d\u52a1\u4eba\u7c7b\u65f6\uff0c\u5b83\u7684\u80fd\u529b\u8fb9\u754c\u7a76\u7adf\u5728\u54ea\u91cc\u3002 \u4ece\u201c\u50cf\u4eba\u8bf4\u8bdd\u201d\uff0c\u5230\u201c\u50cf\u4eba\u884c\u52a8\u201d\uff0c\u518d\u5230\u201c\u50cf\u4e00\u7fa4\u4e0d\u540c\u7684\u4eba\u5171\u540c\u751f\u6d3b\u201d\u2014\u2014\u5927\u6a21\u578b\u8ddd\u79bb\u771f\u6b63\u53ef\u4fe1\u7684\u4eba\u7c7b\u884c\u4e3a\u6a21\u62df\uff0c\u4ecd\u5728\u8def\u4e0a\u3002 A Computational Framework for Evaluating Human-likeness in LLMs\u2019 Open-ended Human Behaviors (ICML 2026) \u8bba\u6587\u94fe\u63a5\uff1ahttps:\/\/icml.cc\/virtual\/2026\/poster\/65092 (opens in new […]<\/p>\n","protected":false},"author":44093,"featured_media":1184663,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-content-parent":1012650,"msr_hide_image_in_river":0,"footnotes":""},"research-area":[13556],"msr-locale":[268881],"msr-post-option":[],"class_list":["post-1184661","msr-blog-post","type-msr-blog-post","status-publish","has-post-thumbnail","hentry","msr-research-area-artificial-intelligence","msr-locale-zh_cn"],"msr_assoc_parent":{"id":1012650,"type":"lab"},"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1184661","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-blog-post"}],"author":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/users\/44093"}],"version-history":[{"count":3,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1184661\/revisions"}],"predecessor-version":[{"id":1184763,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-blog-post\/1184661\/revisions\/1184763"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/1184663"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1184661"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1184661"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1184661"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1184661"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}