{"id":16320,"date":"2026-08-10T20:09:12","date_gmt":"2026-08-10T10:09:12","guid":{"rendered":"https:\/\/geoffreychen.com\/do-two-ais-agreeing-make-an-answer-more-reliable\/"},"modified":"2026-08-10T20:09:12","modified_gmt":"2026-08-10T10:09:12","slug":"do-two-ais-agreeing-make-an-answer-more-reliable","status":"publish","type":"post","link":"https:\/\/geoffreychen.com\/zh\/do-two-ais-agreeing-make-an-answer-more-reliable\/","title":{"rendered":"\u4e24\u4e2aAI\u7ed9\u51fa\u76f8\u540c\u7b54\u6848\uff0c\u53ef\u4fe1\u5ea6\u4f1a\u66f4\u9ad8\u5417\uff1f \/ Does Agreement Between Two AI Systems Make an Answer More Reliable?"},"content":{"rendered":"<\/p>\n<p><strong>\u7b80\u77ed\u7b54\u6848<\/strong><\/p>\n<p>\u6709\u65f6\u4f1a\uff0c\u4f46\u8fdc\u6ca1\u6709\u201c\u4e24\u7968\u5bf9\u4e00\u7968\u201d\u90a3\u4e48\u7b80\u5355\u3002\u4e24\u4e2a\u771f\u6b63\u72ec\u7acb\u3001\u80fd\u529b\u4e92\u8865\u5e76\u4e14\u4f9d\u636e\u4e0d\u540c\u8bc1\u636e\u8def\u5f84\u7684\u7cfb\u7edf\u8fbe\u6210\u4e00\u81f4\uff0c\u53ef\u4ee5\u589e\u52a0\u4fe1\u5fc3\uff1b\u5982\u679c\u5b83\u4eec\u5171\u4eab\u8bad\u7ec3\u8d44\u6599\u3001\u6a21\u578b\u67b6\u6784\u3001\u641c\u7d22\u7ed3\u679c\u3001\u63d0\u793a\u6216\u540c\u4e00\u4e2a\u9519\u8bef\u524d\u63d0\uff0c\u4e00\u81f4\u53ea\u4ee3\u8868\u76f8\u5173\u6027\uff0c\u4e0d\u4ee3\u8868\u771f\u76f8\u3002AI\u4e4b\u95f4\u7684\u5206\u6b67\u662f\u6709\u7528\u7684\u8b66\u62a5\uff0c\u800c\u4e00\u81f4\u6700\u591a\u662f\u4e00\u4e2a\u4fe1\u53f7\uff0c\u4e0d\u80fd\u66ff\u4ee3\u539f\u59cb\u8bc1\u636e\u3002<\/p>\n<p><strong>\u591a\u6570\u8868\u51b3\u4e3a\u4ec0\u4e48\u5728\u4eba\u7c7b\u76f4\u89c9\u4e2d\u6709\u6548<\/strong><\/p>\n<p>\u5982\u679c\u51e0\u4f4d\u5224\u65ad\u8005\u5404\u81ea\u6709\u4e00\u5b9a\u51c6\u786e\u7387\uff0c\u800c\u4e14\u9519\u8bef\u76f8\u4e92\u72ec\u7acb\uff0c\u591a\u6570\u8868\u51b3\u53ef\u4ee5\u964d\u4f4e\u5355\u4e2a\u5224\u65ad\u8005\u5076\u7136\u51fa\u9519\u7684\u5f71\u54cd\u3002\u5173\u952e\u6761\u4ef6\u4e0d\u662f\u201c\u4eba\u6570\u591a\u201d\uff0c\u800c\u662f\u201c\u72ec\u7acb\u4e14\u6709\u80fd\u529b\u201d\u3002\u4e94\u4e2a\u4eba\u90fd\u6284\u540c\u4e00\u4efd\u9519\u8bef\u7b54\u6848\uff0c\u4e0d\u4f1a\u83b7\u5f97\u4e94\u4efd\u8bc1\u636e\uff1b\u4e94\u4e2a\u6e29\u5ea6\u8ba1\u4f7f\u7528\u540c\u4e00\u4e2a\u574f\u4f20\u611f\u5668\uff0c\u4e5f\u4e0d\u4f1a\u56e0\u8bfb\u6570\u4e00\u81f4\u800c\u66f4\u51c6\u786e\u3002<\/p>\n<p>\u5c06\u8fd9\u4e2a\u76f4\u89c9\u76f4\u63a5\u642c\u5230\u8bed\u8a00\u6a21\u578b\u4f1a\u9047\u5230\u95ee\u9898\u3002\u4e0d\u540c\u4ea7\u54c1\u53ef\u80fd\u4f7f\u7528\u76f8\u4f3c\u516c\u5f00\u8bed\u6599\u3001\u76f8\u540c\u57fa\u51c6\u3001\u8fd1\u4f3c\u5fae\u8c03\u65b9\u6cd5\u548c\u5171\u540c\u7684\u641c\u7d22\u7f51\u9875\u3002\u4e00\u4e2a\u6a21\u578b\u7684\u591a\u4e2a\u56de\u7b54\u66f4\u660e\u663e\u5171\u4eab\u53c2\u6570\u4e0e\u76f2\u70b9\u3002\u8868\u9762\u540d\u79f0\u4e0d\u540c\uff0c\u4e0d\u4fdd\u8bc1\u9519\u8bef\u6765\u6e90\u72ec\u7acb\u3002<\/p>\n<p>2025\u5e74\u4e00\u9879\u5bf9350\u591a\u4e2a\u8bed\u8a00\u6a21\u578b\u7684\u7814\u7a76\u53d1\u73b0\uff0c\u6a21\u578b\u9519\u8bef\u5b58\u5728\u663e\u8457\u76f8\u5173\u6027\uff1b\u5728\u5176\u4e2d\u4e00\u4e2a\u6392\u884c\u699c\u6570\u636e\u96c6\u4e0a\uff0c\u5f53\u4e24\u4e2a\u6a21\u578b\u90fd\u7b54\u9519\u65f6\uff0c\u670960%\u7684\u60c5\u51b5\u5b83\u4eec\u540c\u610f\u540c\u4e00\u4e2a\u9519\u8bef\u7b54\u6848\u3002\u7814\u7a76\u8fd8\u53d1\u73b0\uff0c\u5927\u578b\u3001\u66f4\u51c6\u786e\u7684\u6a21\u578b\u4e4b\u95f4\u4e5f\u53ef\u80fd\u51fa\u73b0\u9ad8\u5ea6\u76f8\u5173\u9519\u8bef\uff0c\u5e76\u8ba8\u8bba\u4e86\u8fd9\u79cd\u201c\u7b97\u6cd5\u5355\u4e00\u5316\u201d\u5728\u62db\u8058\u7b49\u573a\u666f\u4e2d\u7684\u540e\u679c\u3002<a href=\"https:\/\/openreview.net\/forum?id=kzYq2hfyHB\" target=\"_blank\" rel=\"noopener\">Kim et al.: Correlated Errors in Large Language Models<\/a><\/p>\n<p><strong>\u540c\u4e00\u4e2a\u6a21\u578b\u95ee\u4e24\u6b21\uff0c\u4e0d\u662f\u4e24\u4f4d\u72ec\u7acb\u4e13\u5bb6<\/strong><\/p>\n<p>\u751f\u6210\u5177\u6709\u968f\u673a\u6027\uff0c\u56e0\u6b64\u540c\u4e00\u4e2a\u63d0\u793a\u91cd\u590d\u8fd0\u884c\u53ef\u80fd\u5f97\u5230\u4e0d\u540c\u8868\u8ff0\u6216\u4e0d\u540c\u7b54\u6848\u3002\u591a\u6b21\u91c7\u6837\u540e\u9009\u591a\u6570\u7b54\u6848\uff0c\u79f0\u4e3a\u4e00\u79cd\u81ea\u6d3d\u7b56\u7565\uff0c\u5728\u67d0\u4e9b\u6570\u5b66\u6216\u7ed3\u6784\u6e05\u695a\u7684\u4efb\u52a1\u4e0a\u53ef\u80fd\u6539\u5584\u8868\u73b0\u3002\u4f46\u8fd9\u4e9b\u6837\u672c\u4ecd\u6765\u81ea\u540c\u4e00\u6a21\u578b\u3001\u540c\u4e00\u77e5\u8bc6\u4e0e\u540c\u4e00\u63d0\u793a\u6846\u67b6\u3002<\/p>\n<p>\u5982\u679c\u6a21\u578b\u7a33\u5b9a\u5730\u4f7f\u7528\u4e86\u9519\u8bef\u516c\u5f0f\u3001\u6df7\u6dc6\u4e86\u540c\u4e00\u4eba\u7269\uff0c\u6216\u63d0\u793a\u5305\u542b\u9519\u8bef\u524d\u63d0\uff0c\u5341\u6b21\u56de\u7b54\u53ef\u80fd\u5341\u6b21\u4e00\u81f4\u3002\u6b64\u65f6\u81ea\u6d3d\u6d4b\u5230\u7684\u662f\u6a21\u578b\u5185\u90e8\u7a33\u5b9a\uff0c\u4e0d\u662f\u771f\u5b9e\u4e16\u754c\u6b63\u786e\u3002\u53cd\u8fc7\u6765\uff0c\u7b54\u6848\u8868\u8ff0\u4e0d\u540c\u4e5f\u672a\u5fc5\u4ee3\u8868\u4e8b\u5b9e\u5206\u6b67\uff0c\u53ef\u80fd\u53ea\u662f\u63aa\u8f9e\u548c\u8ba1\u7b97\u8def\u5f84\u4e0d\u540c\u3002<\/p>\n<p>\u6240\u4ee5\u5148\u628a\u7b54\u6848\u6807\u51c6\u5316\u4e3a\u53ef\u6bd4\u8f83\u4e3b\u5f20\u3002\u4e24\u4e2a\u6a21\u578b\u5206\u522b\u8bf4\u201c\u6211\u5efa\u8baeA\u201d\u548c\u201c\u5728\u8fd9\u4e9b\u6761\u4ef6\u4e0bA\u6210\u672c\u6700\u4f4e\u201d\uff0c\u53ef\u80fd\u7ed3\u8bba\u76f8\u540c\u4f46\u7406\u7531\u4e0d\u540c\uff1b\u82e5\u7406\u7531\u4e2d\u4e00\u4e2a\u4f9d\u8d56\u9519\u8bef\u4ef7\u683c\uff0c\u4e00\u81f4\u5c31\u5f88\u8106\u5f31\u3002<\/p>\n<p><strong>\u4e0d\u540c\u4f9b\u5e94\u5546\u4e5f\u53ef\u80fd\u5171\u4eab\u4fe1\u606f\u751f\u6001<\/strong><\/p>\n<p>\u6a21\u578b\u53ef\u80fd\u4ece\u4e92\u8054\u7f51\u4e0a\u5b66\u4e60\u76f8\u540c\u7684\u6d41\u884c\u8bef\u89e3\uff0c\u53c8\u5728\u56de\u7b54\u65f6\u68c0\u7d22\u5230\u6392\u540d\u9760\u524d\u7684\u540c\u4e00\u6279\u9875\u9762\u3002\u591a\u4e2a\u7f51\u9875\u91cd\u590d\u4e00\u6761\u672a\u7ecf\u8bc1\u5b9e\u7684\u8bf4\u6cd5\uff0c\u4f1a\u4f7f\u4e0d\u540c\u7cfb\u7edf\u72ec\u7acb\u641c\u7d22\u540e\u4ecd\u7136\u6c47\u5408\u3002\u5b83\u4eec\u7684\u8ba1\u7b97\u8fc7\u7a0b\u5206\u5f00\uff0c\u8bc1\u636e\u7956\u5148\u5374\u76f8\u540c\u3002<\/p>\n<p>\u8fd9\u5728\u6700\u65b0\u65b0\u95fb\u3001\u4eba\u7269\u4f20\u95fb\u3001\u4ea7\u54c1\u6bd4\u8f83\u548c\u5386\u53f2\u8f76\u4e8b\u4e2d\u5e38\u89c1\u3002\u8981\u5224\u65ad\u72ec\u7acb\u6027\uff0c\u4e0d\u53ea\u662f\u6362\u4e00\u4e2a\u804a\u5929\u7a97\u53e3\u6216\u54c1\u724c\uff0c\u800c\u8981\u95ee\uff1a\u5b83\u662f\u5426\u4f7f\u7528\u4e0d\u540c\u7684\u4e00\u624b\u6765\u6e90\uff1b\u662f\u5426\u6709\u4e0d\u540c\u4e13\u4e1a\u65b9\u6cd5\uff1b\u662f\u5426\u80fd\u5c55\u793a\u8bc1\u636e\uff1b\u662f\u5426\u9488\u5bf9\u4e0d\u540c\u5931\u8d25\u6a21\u5f0f\u8bbe\u8ba1\u3002<\/p>\n<p>\u4f8b\u5982\u4e00\u4e2a\u6a21\u578b\u603b\u7ed3\u6cd5\u89c4\uff0c\u53e6\u4e00\u4e2a\u6a21\u578b\u4e5f\u603b\u7ed3\u540c\u4e00\u7bc7\u4e8c\u624b\u535a\u5ba2\uff0c\u5e76\u4e0d\u662f\u72ec\u7acb\u590d\u6838\u3002\u4e00\u4e2a\u6a21\u578b\u603b\u7ed3\u6cd5\u89c4\uff0c\u968f\u540e\u7531\u4eba\u6253\u5f00\u5b98\u65b9\u6587\u672c\u6838\u5bf9\u5173\u952e\u6761\u6b3e\uff0c\u624d\u662f\u65b0\u7684\u8bc1\u636e\u8def\u5f84\u3002<\/p>\n<p><strong>\u5206\u6b67\u5f80\u5f80\u6bd4\u4e00\u81f4\u66f4\u6709\u4fe1\u606f\u91cf<\/strong><\/p>\n<p>\u4e24\u4e2a\u7cfb\u7edf\u5728\u4e00\u4e2a\u660e\u786e\u4e8b\u5b9e\u4e0a\u7684\u5206\u6b67\uff0c\u8bf4\u660e\u81f3\u5c11\u4e00\u4e2a\u7b54\u6848\u9519\u8bef\u3001\u95ee\u9898\u542b\u7cca\u3001\u8d44\u6599\u7248\u672c\u4e0d\u540c\uff0c\u6216\u7cfb\u7edf\u4e0d\u786e\u5b9a\u3002\u4e0d\u8981\u7acb\u5373\u8ba9\u7b2c\u4e09\u4e2a\u6a21\u578b\u6295\u7968\uff0c\u800c\u8981\u5b9a\u4f4d\u5206\u6b67\uff1a\u662f\u65e5\u671f\u3001\u5b9a\u4e49\u3001\u5730\u533a\u3001\u8ba1\u7b97\u8fd8\u662f\u4ef7\u503c\u6807\u51c6\uff1f\u7136\u540e\u5bfb\u627e\u80fd\u591f\u88c1\u51b3\u8be5\u70b9\u7684\u539f\u59cb\u6765\u6e90\u3002<\/p>\n<p>\u5206\u6b67\u4e5f\u53ef\u4ee5\u7528\u6765\u5206\u914d\u4eba\u5de5\u6ce8\u610f\u529b\u3002\u7814\u7a76\u8868\u660e\uff0c\u591a\u6a21\u578b\u7684\u4e0d\u4e00\u81f4\u5728\u67d0\u4e9b\u4efb\u52a1\u4e2d\u80fd\u591f\u4f5c\u4e3a\u4e0d\u786e\u5b9a\u6027\u6216\u9700\u8981\u4eba\u5de5\u6807\u6ce8\u7684\u4fe1\u53f7\uff1b\u4f46\u5177\u4f53\u6548\u679c\u4f9d\u8d56\u4efb\u52a1\u548c\u7ec4\u5408\uff0c\u4e0d\u80fd\u63a8\u5e7f\u6210\u901a\u7528\u4fdd\u8bc1\u3002<a href=\"https:\/\/aclanthology.org\/2023.gem-1.5\/\" target=\"_blank\" rel=\"noopener\">Du et al.: Ensemble Disagreement Scores in Large Language Models<\/a><\/p>\n<p>\u4e00\u81f4\u5219\u4e0d\u5e94\u81ea\u52a8\u901a\u8fc7\u3002\u5b83\u53ef\u80fd\u63d0\u9ad8\u4f18\u5148\u7ea7\u5224\u65ad\uff1a\u4f4e\u98ce\u9669\u4e8b\u5b9e\u4e14\u8bc1\u636e\u8def\u5f84\u771f\u6b63\u4e0d\u540c\uff0c\u53ef\u51cf\u5c11\u989d\u5916\u68c0\u67e5\uff1b\u9ad8\u98ce\u9669\u51b3\u5b9a\u5373\u4f7f\u4e00\u81f4\uff0c\u4e5f\u5e94\u56de\u5230\u539f\u59cb\u8bb0\u5f55\u3002\u5206\u6b67\u662f\u201c\u5fc5\u987b\u67e5\u201d\uff0c\u4e00\u81f4\u53ea\u662f\u201c\u53ef\u80fd\u8f83\u7a33\u5b9a\u201d\u3002<\/p>\n<p><strong>\u5148\u6bd4\u8f83\u8bc1\u636e\uff0c\u518d\u6bd4\u8f83\u7ed3\u8bba<\/strong><\/p>\n<p>\u8ba9\u4e24\u4e2aAI\u53ea\u62a5\u6700\u7ec8\u7b54\u6848\uff0c\u65e0\u6cd5\u77e5\u9053\u5b83\u4eec\u662f\u5426\u72ec\u7acb\u3002\u66f4\u6709\u6548\u7684\u65b9\u5f0f\u662f\u8981\u6c42\u5404\u81ea\u5217\u51fa\u627f\u91cd\u4e8b\u5b9e\u3001\u6765\u6e90\u3001\u5047\u8bbe\u3001\u672a\u77e5\u4e0e\u4f1a\u6539\u53d8\u7b54\u6848\u7684\u6761\u4ef6\u3002\u7136\u540e\u6bd4\u8f83\uff1a\u5b83\u4eec\u662f\u5426\u5f15\u7528\u540c\u4e00\u6765\u6e90\uff1b\u662f\u5426\u4f7f\u7528\u4e0d\u540c\u6570\u636e\uff1b\u662f\u5426\u5728\u540c\u4e00\u672a\u8bc1\u5b9e\u524d\u63d0\u4e0a\u5efa\u7acb\u7ed3\u8bba\uff1b\u662f\u5426\u5bf9\u98ce\u9669\u548c\u4ef7\u503c\u4f7f\u7528\u76f8\u540c\u9ed8\u8ba4\u6807\u51c6\u3002<\/p>\n<p>\u82e5\u7ed3\u8bba\u4e00\u81f4\u800c\u8bc1\u636e\u4e0d\u540c\uff0c\u53ef\u4ee5\u56de\u67e5\u4e24\u6761\u8def\u5f84\u662f\u5426\u771f\u6b63\u6536\u655b\u3002\u82e5\u7ed3\u8bba\u4e00\u81f4\u4f46\u8bc1\u636e\u51b2\u7a81\uff0c\u8bf4\u660e\u81f3\u5c11\u4e00\u4e2a\u63a8\u7406\u6709\u95ee\u9898\u3002\u82e5\u7ed3\u8bba\u4e0d\u540c\u4f46\u4e8b\u5b9e\u76f8\u540c\uff0c\u5dee\u5f02\u53ef\u80fd\u6765\u81ea\u4ef7\u503c\u6743\u91cd\uff0c\u5e94\u8be5\u7531\u4eba\u660e\u786e\u9009\u62e9\uff0c\u800c\u4e0d\u662f\u9760\u6a21\u578b\u6295\u7968\u3002<\/p>\n<p><strong>\u4ec0\u4e48\u65f6\u5019\u591a\u6a21\u578b\u65b9\u6cd5\u6bd4\u8f83\u6709\u7528<\/strong><\/p>\n<p>\u5728\u6709\u660e\u786e\u7b54\u6848\u3001\u53ef\u5927\u91cf\u6d4b\u8bd5\u4e14\u9519\u8bef\u6210\u672c\u53d7\u63a7\u7684\u4efb\u52a1\u4e2d\uff0c\u591a\u6a21\u578b\u6216\u591a\u6b21\u91c7\u6837\u53ef\u4ee5\u6210\u4e3a\u5de5\u7a0b\u7b56\u7565\uff0c\u4f8b\u5982\u5206\u7c7b\u3001\u5b57\u6bb5\u63d0\u53d6\u3001\u4ee3\u7801\u5019\u9009\u6216\u6570\u5b66\u9898\u3002\u524d\u63d0\u662f\u5148\u7528\u5e26\u6b63\u786e\u7b54\u6848\u7684\u4ee3\u8868\u6027\u6570\u636e\u6d4b\u91cf\u6bcf\u4e2a\u6a21\u578b\u3001\u7ec4\u5408\u89c4\u5219\u548c\u5931\u8d25\u76f8\u5173\u6027\uff0c\u800c\u4e0d\u662f\u770b\u5230\u4e00\u6b21\u4e00\u81f4\u5c31\u5ba3\u5e03\u53ef\u9760\u3002<\/p>\n<p>\u53ef\u4ee5\u628a\u5206\u6b67\u8def\u7531\u7ed9\u4eba\uff0c\u628a\u4e00\u81f4\u7ed3\u679c\u518d\u901a\u8fc7\u786e\u5b9a\u6027\u89c4\u5219\u6216\u62bd\u6837\u5ba1\u8ba1\uff1b\u4e5f\u53ef\u4ee5\u7ed9\u6a21\u578b\u4e0d\u540c\u89d2\u8272\uff0c\u4e00\u4e2a\u751f\u6210\u3001\u4e00\u4e2a\u5bfb\u627e\u53cd\u4f8b\u3001\u4e00\u4e2a\u53ea\u6838\u5bf9\u6765\u6e90\u3002\u4f46\u89d2\u8272\u63d0\u793a\u4ecd\u4e0d\u80fd\u521b\u9020\u771f\u6b63\u72ec\u7acb\u77e5\u8bc6\uff0c\u6700\u7ec8\u8981\u770b\u6d4b\u8bd5\u6570\u636e\u3002<\/p>\n<p>\u5bf9\u4e8e\u5f00\u653e\u5f0f\u5efa\u8bae\u3001\u4f26\u7406\u9009\u62e9\u3001\u6cd5\u5f8b\u6216\u533b\u7597\u5224\u65ad\uff0c\u6a21\u578b\u6570\u91cf\u4e0d\u80fd\u66ff\u4ee3\u9002\u7528\u80cc\u666f\u4e0e\u4e13\u4e1a\u8d23\u4efb\u3002\u4e09\u79cd\u7cfb\u7edf\u90fd\u53ef\u80fd\u7ed9\u51fa\u6d41\u884c\u4f46\u4e0d\u9002\u5408\u4e2a\u6848\u7684\u5efa\u8bae\u3002<\/p>\n<p><strong>\u600e\u6837\u8bbe\u8ba1\u66f4\u6709\u4ef7\u503c\u7684\u7b2c\u4e8c\u610f\u89c1<\/strong><\/p>\n<p>\u4e0d\u8981\u628a\u7b2c\u4e00\u4e2a\u7b54\u6848\u5c55\u793a\u7ed9\u7b2c\u4e8c\u4e2a\u7cfb\u7edf\uff0c\u5426\u5219\u5b83\u53ef\u80fd\u88ab\u951a\u5b9a\u3002\u5206\u522b\u7ed9\u51fa\u539f\u59cb\u95ee\u9898\uff0c\u5e76\u6545\u610f\u6539\u53d8\u4efb\u52a1\uff1a\u7b2c\u4e00\u4e2a\u5f62\u6210\u6700\u4f73\u89e3\u91ca\uff1b\u7b2c\u4e8c\u4e2a\u5bfb\u627e\u80fd\u591f\u63a8\u7ffb\u7ed3\u8bba\u7684\u8bc1\u636e\uff1b\u7b2c\u4e09\u4e2a\u6838\u5bf9\u6307\u5b9a\u4e00\u624b\u6765\u6e90\u3002\u628a\u751f\u6210\u3001\u53cd\u9a73\u548c\u9a8c\u8bc1\u5206\u5f00\uff0c\u6bd4\u8ba9\u4e09\u4e2a\u6a21\u578b\u90fd\u201c\u56de\u7b54\u95ee\u9898\u201d\u66f4\u80fd\u66b4\u9732\u7f3a\u53e3\u3002<\/p>\n<p>\u7b2c\u4e8c\u610f\u89c1\u8fd8\u5e94\u4f7f\u7528\u4e0d\u540c\u8bc1\u636e\u7c7b\u578b\u3002\u8bed\u8a00\u6a21\u578b\u7ed9\u51fa\u89e3\u91ca\u540e\uff0c\u7528\u6570\u636e\u5e93\u786e\u8ba4\u8bb0\u5f55\u3001\u7528\u8ba1\u7b97\u5668\u91cd\u7b97\u6570\u5b57\u3001\u7528\u6d4b\u8bd5\u8fd0\u884c\u4ee3\u7801\u3001\u7531\u4e13\u4e1a\u4eba\u58eb\u5224\u65ad\u8fb9\u754c\u3002\u771f\u6b63\u7684\u591a\u6837\u6027\u6765\u81ea\u65b9\u6cd5\uff0c\u4e0d\u53ea\u662f\u6a21\u578b\u540d\u5355\u3002<\/p>\n<p>\u8bb0\u5f55\u6bcf\u4e2a\u7cfb\u7edf\u7684\u7248\u672c\u4e0e\u8f93\u5165\u4e5f\u5f88\u91cd\u8981\u3002\u82e5\u6a21\u578b\u540e\u6765\u66f4\u65b0\uff0c\u8fc7\u53bb\u7684\u201c\u72ec\u7acb\u7ec4\u5408\u201d\u53ef\u80fd\u6539\u53d8\uff1b\u82e5\u6240\u6709\u6a21\u578b\u901a\u8fc7\u540c\u4e00\u4e2a\u641c\u7d22\u4f9b\u5e94\u5546\uff0c\u5b83\u4eec\u7684\u68c0\u7d22\u591a\u6837\u6027\u4e5f\u53ef\u80fd\u4f4e\u4e8e\u9884\u671f\u3002<\/p>\n<p><strong>\u4e00\u4e2a\u7b80\u5355\u7684\u8bc1\u636e\u77e9\u9635<\/strong><\/p>\n<p>\u4e3a\u91cd\u8981\u95ee\u9898\u5efa\u7acb\u56db\u5217\uff1a\u4e3b\u5f20\u3001AI A\u4f9d\u636e\u3001AI B\u4f9d\u636e\u3001\u72ec\u7acb\u88c1\u51b3\u8bc1\u636e\u3002\u524d\u4e24\u5217\u7528\u4e8e\u66b4\u9732\u76f8\u540c\u6765\u6e90\u548c\u5047\u8bbe\uff0c\u6700\u540e\u4e00\u5217\u5fc5\u987b\u6765\u81ea\u80fd\u591f\u76f4\u63a5\u89e3\u51b3\u95ee\u9898\u7684\u5916\u90e8\u8bb0\u5f55\u3002\u82e5\u6ca1\u6709\u88c1\u51b3\u8bc1\u636e\uff0c\u5c06\u7ed3\u8bba\u6807\u8bb0\u4e3a\u672a\u89e3\u51b3\uff0c\u800c\u4e0d\u662f\u91c7\u7528\u591a\u6570\u7968\u3002<\/p>\n<p>\u8fd8\u53ef\u4ee5\u52a0\u4e00\u5217\u201c\u9519\u8bef\u540e\u679c\u201d\u3002\u5bf9\u4e8e\u4f4e\u540e\u679c\u9879\u76ee\uff0c\u4e24\u6761\u5408\u7406\u72ec\u7acb\u8def\u5f84\u53ef\u80fd\u8db3\u591f\uff1b\u5bf9\u4e8e\u9ad8\u540e\u679c\u9879\u76ee\uff0c\u5373\u4f7f\u4e09\u8005\u4e00\u81f4\uff0c\u4e5f\u9700\u8981\u6b63\u5f0f\u5ba1\u6279\u6216\u4e13\u4e1a\u590d\u6838\u3002\u8bc1\u636e\u95e8\u69db\u7531\u540e\u679c\u51b3\u5b9a\uff0c\u4e0d\u7531\u6a21\u578b\u6570\u91cf\u51b3\u5b9a\u3002<\/p>\n<p><strong>\u6211\u7684\u5224\u65ad\uff1a\u628a\u6a21\u578b\u4e00\u81f4\u5f53\u4f5c\u7ebf\u7d22\uff0c\u4e0d\u5f53\u4f5c\u8bc1\u636e<\/strong><\/p>\n<p>\u6211\u4f1a\u66f4\u91cd\u89c6\u5206\u6b67\uff0c\u56e0\u4e3a\u5b83\u660e\u786e\u544a\u8bc9\u6211\u54ea\u91cc\u9700\u8981\u67e5\uff1b\u5bf9\u4e8e\u4e00\u81f4\uff0c\u6211\u9996\u5148\u5bfb\u627e\u5171\u540c\u7956\u5148\u3002\u5982\u679c\u4e24\u4e2a\u56de\u7b54\u4f7f\u7528\u540c\u4e00\u4e2a\u7f51\u9875\u3001\u540c\u4e00\u79cd\u9ed8\u8ba4\u5047\u8bbe\u6216\u540c\u4e00\u5957\u8bad\u7ec3\u4e60\u60ef\uff0c\u5b83\u4eec\u5b9e\u9645\u4e0a\u53ea\u6709\u4e00\u7968\u3002<\/p>\n<p>\u6700\u5f3a\u7684\u7ec4\u5408\u901a\u5e38\u4e0d\u662f\u201cAI\u52a0AI\u201d\uff0c\u800c\u662f\u751f\u6210\u5f0fAI\u52a0\u4e0a\u4e0d\u540c\u6027\u8d28\u7684\u9a8c\u8bc1\uff1a\u5b98\u65b9\u8bb0\u5f55\u3001\u53ef\u91cd\u590d\u8ba1\u7b97\u3001\u6d4b\u8bd5\u3001\u539f\u59cb\u6587\u4ef6\u548c\u5177\u5907\u8d23\u4efb\u7684\u4e13\u4e1a\u5224\u65ad\u3002\u8fd9\u4e9b\u65b9\u6cd5\u4f1a\u5728\u6a21\u578b\u5171\u540c\u76f2\u70b9\u5904\u63d0\u4f9b\u963b\u529b\u3002<\/p>\n<p><strong>\u68c0\u67e5\u6e05\u5355<\/strong><\/p>\n<ul>\n<li>\u4e24\u4e2a\u7b54\u6848\u6765\u81ea\u4e0d\u540c\u91c7\u6837\uff0c\u8fd8\u662f\u4e0d\u540c\u6a21\u578b\u4e0e\u4e0d\u540c\u8bc1\u636e\u8def\u5f84\uff1f<\/li>\n<li>\u7b2c\u4e8c\u4e2a\u7cfb\u7edf\u662f\u5426\u5728\u770b\u5230\u7b2c\u4e00\u4e2a\u7b54\u6848\u524d\u72ec\u7acb\u5224\u65ad\uff1f<\/li>\n<li>\u5b83\u4eec\u662f\u5426\u5f15\u7528\u540c\u4e00\u7f51\u9875\u3001\u6570\u636e\u5e93\u6216\u4e8c\u624b\u8bf4\u6cd5\uff1f<\/li>\n<li>\u7ed3\u8bba\u4e00\u81f4\u65f6\uff0c\u7406\u7531\u548c\u627f\u91cd\u4e8b\u5b9e\u662f\u5426\u4e5f\u4e00\u81f4\u4e14\u53ef\u6838\u67e5\uff1f<\/li>\n<li>\u7ed3\u8bba\u5206\u6b67\u662f\u4e8b\u5b9e\u3001\u7248\u672c\u3001\u5b9a\u4e49\uff0c\u8fd8\u662f\u4ef7\u503c\u6807\u51c6\u4e0d\u540c\uff1f<\/li>\n<li>\u662f\u5426\u6709\u539f\u59cb\u8bb0\u5f55\u3001\u8ba1\u7b97\u6216\u6d4b\u8bd5\u53ef\u4ee5\u88c1\u51b3\uff1f<\/li>\n<li>\u7ec4\u5408\u65b9\u6cd5\u662f\u5426\u5728\u4ee3\u8868\u6027\u771f\u5b9e\u6848\u4f8b\u4e0a\u6d4b\u8bd5\u8fc7\uff1f<\/li>\n<li>\u9519\u8bef\u540e\u679c\u662f\u5426\u5141\u8bb8\u4ec5\u51ed\u6a21\u578b\u4e00\u81f4\u5c31\u91c7\u53d6\u884c\u52a8\uff1f<\/li>\n<\/ul>\n<p><strong>\u7ed3\u8bba<\/strong><\/p>\n<p>\u4e24\u4e2aAI\u540c\u610f\u53ef\u4ee5\u63d0\u4f9b\u6709\u9650\u7684\u7a33\u5b9a\u6027\u4fe1\u53f7\uff0c\u4f46\u53ea\u6709\u5728\u80fd\u529b\u8db3\u591f\u3001\u9519\u8bef\u8fd1\u4f3c\u72ec\u7acb\u3001\u8bc1\u636e\u8def\u5f84\u4e0d\u540c\u7684\u6761\u4ef6\u4e0b\uff0c\u624d\u53ef\u80fd\u5b9e\u8d28\u63d0\u9ad8\u53ef\u4fe1\u5ea6\u3002\u73b0\u5b9e\u6a21\u578b\u5e38\u5171\u4eab\u8d44\u6599\u3001\u65b9\u6cd5\u548c\u8bef\u89e3\uff0c\u56e0\u800c\u80fd\u591f\u4e00\u81f4\u5730\u7b54\u9519\u3002\u7528\u5206\u6b67\u89e6\u53d1\u8c03\u67e5\uff0c\u7528\u8bc1\u636e\u88c1\u51b3\u95ee\u9898\uff1b\u4e0d\u8981\u628a\u591a\u4e2a\u76f8\u4f3c\u7cfb\u7edf\u7684\u5171\u8bc6\u8bef\u8ba4\u4e3a\u72ec\u7acb\u4e8b\u5b9e\u3002<\/p>\n<p><strong>\u76f8\u5173\u95ee\u9898<\/strong><\/p>\n<ul>\n<li><a href=\"https:\/\/geoffreychen.com\/zh\/how-to-verify-ai-without-redoing-the-work\/\">\u600e\u6837\u5feb\u901f\u9a8c\u8bc1AI\u7b54\u6848\uff0c\u800c\u4e0d\u662f\u628a\u5de5\u4f5c\u91cd\u65b0\u505a\u4e00\u904d\uff1f<\/a><\/li>\n<li><a href=\"https:\/\/geoffreychen.com\/zh\/why-human-review-still-misses-ai-errors\/\">\u4eba\u5de5\u5ba1\u6838AI\u8f93\u51fa\uff0c\u4e3a\u4ec0\u4e48\u4ecd\u7136\u53ef\u80fd\u6f0f\u6389\u9519\u8bef\uff1f<\/a><\/li>\n<\/ul>\n<p><strong>\u7ee7\u7eed\u9605\u8bfb\u672c\u7cfb\u5217\uff1a<\/strong> <a href=\"https:\/\/geoffreychen.com\/zh\/ai-judgment\/\">\u300aAI\u53ef\u4ee5\u76f8\u4fe1\u5230\u4ec0\u4e48\u7a0b\u5ea6\uff1f\u300b\u5168\u90e8\u6587\u7ae0<\/a><\/p>\n<p>","protected":false},"excerpt":{"rendered":"","protected":false},"author":17897162,"featured_media":0,"comment_status":"","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":"Agreement between two AI systems can raise confidence only when they\u2019re truly independent; shared data or premises make it correlation, not truth.","_geoff_meta_description_zh":"\u4e24\u4e2aAI\u4e00\u81f4\u53ef\u5728\u4e00\u5b9a\u6761\u4ef6\u4e0b\u63d0\u5347\u4fe1\u5fc3\uff0c\u4f46\u82e5\u5171\u4eab\u8bad\u7ec3\u3001\u67b6\u6784\u3001\u68c0\u7d22\u6216\u9519\u8bef\u524d\u63d0\uff0c\u4e00\u81f4\u53ea\u8bf4\u660e\u76f8\u5173\u6027\u800c\u975e\u771f\u76f8\u3002","_wpcom_ai_launchpad_first_post":false,"_jetpack_hide_featured_image":false,"_jetpack_featured_image_duplicate_dismissed":0,"_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":[787295760],"tags":[787295762,787295787,787295763,787295786,787295104,787295785,787295103,787295784],"class_list":["post-16320","post","type-post","status-publish","format-standard","hentry","category-ai-judgment","tag-ai-judgment","tag-correlated-ai-errors-zh","tag-ai-judgment-zh","tag-correlated-ai-errors","tag-responsible-ai-zh","tag-model-agreement","tag-responsible-ai","tag-model-agreement-zh",""],"translatepress_titles":{"en_AU":"\u4e24\u4e2aAI\u7ed9\u51fa\u76f8\u540c\u7b54\u6848\uff0c\u53ef\u4fe1\u5ea6\u4f1a\u66f4\u9ad8\u5417\uff1f \/ Does Agreement Between Two AI Systems Make an Answer More Reliable?"},"jetpack_publicize_connections":[],"jetpack_likes_enabled":false,"jetpack_sharing_enabled":true,"jetpack_shortlink":"https:\/\/wp.me\/p1ar9H-4fe","jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/posts\/16320","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=16320"}],"version-history":[{"count":0,"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/posts\/16320\/revisions"}],"wp:attachment":[{"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/media?parent=16320"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/categories?post=16320"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/geoffreychen.com\/zh\/wp-json\/wp\/v2\/tags?post=16320"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}