{"id":34908,"date":"2024-12-15T05:58:57","date_gmt":"2024-12-14T21:58:57","guid":{"rendered":"https:\/\/17aitech.com\/?p=34908"},"modified":"2024-12-15T05:58:57","modified_gmt":"2024-12-14T21:58:57","slug":"%e9%a6%96%e5%b1%8a%e5%a4%a7%e6%a8%a1%e5%9e%8b%e9%a1%b6%e4%bc%9acolm-%e9%ab%98%e5%88%86%e8%ae%ba%e6%96%87%ef%bc%9a%e5%81%8f%e5%a5%bd%e6%90%9c%e7%b4%a2%e7%ae%97%e6%b3%95pairs%ef%bc%8c%e8%ae%a9%e5%a4%a7","status":"publish","type":"post","link":"https:\/\/17aitech.com\/?p=34908","title":{"rendered":"\u9996\u5c4a\u5927\u6a21\u578b\u9876\u4f1aCOLM \u9ad8\u5206\u8bba\u6587\uff1a\u504f\u597d\u641c\u7d22\u7b97\u6cd5PairS\uff0c\u8ba9\u5927\u6a21\u578b\u8fdb\u884c\u6587\u672c\u8bc4\u4f30\u66f4\u9ad8\u6548"},"content":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-08-04\" target=\"_blank\">\u9996\u5c4a\u5927\u6a21\u578b\u9876\u4f1aCOLM \u9ad8\u5206\u8bba\u6587\uff1a\u504f\u597d\u641c\u7d22\u7b97\u6cd5PairS\uff0c\u8ba9\u5927\u6a21\u578b\u8fdb\u884c\u6587\u672c\u8bc4\u4f30\u66f4\u9ad8\u6548<\/a><\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-2f91a5574163de717c5146ac0a709263.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-2f91a5574163de717c5146ac0a709263.png\"><\/a><\/section>\n<blockquote data-author-name=\"\" data-content-utf8-length=\"166\" data-source-title=\"\" data-type=\"2\" data-url=\"\">\n<section>\n<section>AIxiv\u4e13\u680f\u662f<mark data-type=\"institutions\" data-id=\"9e96434f-5827-41e3-9b3c-36d39bc0d446\">\u673a\u5668\u4e4b\u5fc3<\/mark>\u53d1\u5e03\u5b66\u672f\u3001\u6280\u672f\u5185\u5bb9\u7684\u680f\u76ee\u3002\u8fc7\u53bb\u6570\u5e74\uff0c<mark data-type=\"institutions\" data-id=\"9e96434f-5827-41e3-9b3c-36d39bc0d446\">\u673a\u5668\u4e4b\u5fc3<\/mark>AIxiv\u4e13\u680f\u63a5\u6536\u62a5\u9053\u4e862000\u591a\u7bc7\u5185\u5bb9\uff0c\u8986\u76d6\u5168\u7403\u5404\u5927\u9ad8\u6821\u4e0e\u4f01\u4e1a\u7684\u9876\u7ea7\u5b9e\u9a8c\u5ba4\uff0c\u6709\u6548\u4fc3\u8fdb\u4e86\u5b66\u672f\u4ea4\u6d41\u4e0e\u4f20\u64ad\u3002\u5982\u679c\u60a8\u6709\u4f18\u79c0\u7684\u5de5\u4f5c\u60f3\u8981\u5206\u4eab\uff0c\u6b22\u8fce\u6295\u7a3f\u6216\u8005\u8054\u7cfb\u62a5\u9053\u3002\u6295\u7a3f\u90ae\u7bb1\uff1aliyazhou@jiqizhixin.com\uff1bzhaoyunfeng@jiqizhixin.com<\/section>\n<\/section>\n<\/blockquote>\n<p><strong>\u6587\u7ae0\u4f5c\u8005\u7686\u4e3a\u6765\u81ea\u5251\u6865\u5927\u5b66\u8bed\u8a00\u6280\u672f\u5b9e\u9a8c\u5ba4\uff0c\u4e00\u4f5c\u4e3a\u4e09\u5e74\u7ea7\u535a\u58eb\u751f\u5218\u80e4\u5b8f\uff0c\u5bfc\u5e08\u4e3a\u6559\u6388 Nigel Collier \u548c Ehsan Shareghi\u3002\u4ed6\u7684\u7814\u7a76\u5174\u8da3\u662f\u5927\u6a21\u578b\u548c\u6587\u672c\u8bc4\u4f30\uff0c\u6570\u636e\u751f\u6210\u7b49\u3002\u5171\u540c\u4e00\u4f5c\u4e3a\u4e8c\u5e74\u7ea7\u535a\u58eb\u751f\u5468\u6db5\uff0c\u5bfc\u5e08\u4e3a\u6559\u6388 Anna Korhonen \u548c Ivan Vuli\u0107\uff0c\u4ed6\u7684\u7814\u7a76\u5174\u8da3\u662f\u9ad8\u6548\u5927\u6a21\u578b\u3002<\/strong><\/p>\n<p>\u5927\u6a21\u578b\u5c55\u73b0\u51fa\u4e86\u5353\u8d8a\u7684\u6307\u4ee4\u8ddf\u4ece\u548c\u4efb\u52a1\u6cdb\u5316\u7684\u80fd\u529b\uff0c\u8fd9\u79cd\u72ec\u7279\u7684\u80fd\u529b\u6e90\u81ea LLMs \u5728\u8bad\u7ec3\u4e2d\u4f7f\u7528\u4e86\u6307\u4ee4\u8ddf\u968f\u6570\u636e\u4ee5\u53ca\u4eba\u7c7b\u53cd\u9988<mark data-type=\"tech_methods\" data-id=\"ee1a8f69-3170-4ddf-b2b6-47d91c844425\">\u5f3a\u5316\u5b66\u4e60<\/mark>\uff08RLHF\uff09\u3002\u5728 RLHF \u8bad\u7ec3\u8303\u5f0f\u4e2d\uff0c\u5956\u52b1\u6a21\u578b\u6839\u636e\u6392\u540d\u6bd4\u8f83\u6570\u636e\u4e0e\u4eba\u7c7b\u504f\u597d\u5bf9\u9f50\u3002\u8fd9\u589e\u5f3a\u4e86 LLMs \u4e0e\u4eba\u7c7b\u4ef7\u503c\u89c2\u7684\u5bf9\u9f50\uff0c\u4ece\u800c\u751f\u6210\u66f4\u597d\u5730\u5e2e\u52a9\u4eba\u7c7b\u5e76\u9075\u5b88\u4eba\u7c7b\u4ef7\u503c\u89c2\u7684\u56de\u5e94\u3002<\/p>\n<p>\u8fd1\u65e5\uff0c\u7b2c\u4e00\u5c4a\u5927\u6a21\u578b\u9876\u4f1a COLM \u521a\u521a\u516c\u5e03\u63a5\u6536\u7ed3\u679c\uff0c\u5176\u4e2d\u4e00\u9879\u9ad8\u5206\u5de5\u4f5c\u5206\u6790\u4e86 LLM \u4f5c\u4e3a\u6587\u672c\u8bc4\u4f30\u5668\u65f6\u96be\u4ee5\u907f\u514d\u548c\u7ea0\u6b63\u7684\u5206\u6570\u504f\u89c1\u95ee\u9898\uff0c\u5e76\u63d0\u51fa\u4e86\u5c06\u8bc4\u4f30\u95ee\u9898\u8f6c\u6362\u6210\u504f\u597d\u6392\u5e8f\u95ee\u9898\uff0c\u4ece\u800c\u8bbe\u8ba1\u4e86 PairS \u7b97\u6cd5\uff0c\u4e00\u4e2a\u53ef\u4ee5\u4ece\u6210\u5bf9\u504f\u597d\uff08pairwise preference\uff09\u4e2d\u641c\u7d22\u548c\u6392\u5e8f\u7684\u7b97\u6cd5\u3002\u901a\u8fc7\u5229\u7528\u4e0d\u786e\u5b9a\u6027\u548c LLM \u4f20\u9012\u6027\uff08transitivity\uff09\u7684\u5047\u8bbe\uff0cPairS \u53ef\u4ee5\u7ed9\u51fa\u9ad8\u6548\uff0c\u51c6\u786e\u7684\u504f\u597d\u6392\u5e8f\uff0c\u5e76\u5728\u591a\u4e2a\u6d4b\u8bd5\u96c6\u4e0a\u5c55\u73b0\u51fa\u548c\u4eba\u7c7b\u5224\u65ad\u66f4\u9ad8\u7684\u4e00\u81f4\u6027\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-38122948e907c13b161c84534646aed3.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-38122948e907c13b161c84534646aed3.png\"><\/a><\/p>\n<ul>\n<li>\n<p>\u8bba\u6587\u94fe\u63a5: https:\/\/arxiv.org\/abs\/2403.16950<\/p>\n<\/li>\n<li>\n<p>\u8bba\u6587\u6807\u9898\uff1aAligning with Human Judgement: The Role of Pairwise Preference in Large Language Model Evaluators<\/p>\n<\/li>\n<li>\n<p>Github \u5730\u5740: https:\/\/github.com\/cambridgeltl\/PairS<\/p>\n<\/li>\n<\/ul>\n<p><strong>\u7528\u5927\u6a21\u578b\u8bc4\u4f30\u6709\u4ec0\u4e48\u95ee\u9898\uff1f<\/strong><\/p>\n<p>\u6700\u8fd1\u5927\u91cf\u7684\u5de5\u4f5c\u5c55\u793a\u4e86 LLMs \u5728\u8bc4\u4f30\u6587\u672c\u8d28\u91cf\u4e0a\u7684\u51fa\u8272\u8868\u73b0\uff0c\u5f62\u6210\u4e86\u4e00\u79cd\u65e0\u9700\u53c2\u8003\u7684\u751f\u6210\u4efb\u52a1\u8bc4\u4f30\u65b0\u8303\u5f0f\uff0c\u907f\u514d\u4e86\u6602\u8d35\u7684\u4eba\u7c7b\u6807\u6ce8\u6210\u672c\u3002\u7136\u800c\uff0cLLM \u8bc4\u4f30\u5668\uff08evaluator\uff09\u5bf9\u63d0\u793a\uff08prompt\uff09\u8bbe\u8ba1\u9ad8\u5ea6\u654f\u611f\uff0c\u751a\u81f3\u4f1a\u53d7\u5230\u591a\u79cd\u504f\u89c1\u7684\u5f71\u54cd\uff0c\u5305\u62ec\u4f4d\u7f6e\u504f\u89c1\u3001\u5197\u957f\u504f\u89c1\u548c\u4e0a\u4e0b\u6587\u504f\u89c1\u3002\u8fd9\u4e9b\u504f\u89c1\u963b\u788d\u4e86 LLM \u8bc4\u4f30\u5668\u7684\u516c\u5e73\u548c\u53ef\u4fe1\uff0c\u5bfc\u81f4\u4e0e\u4eba\u7c7b\u5224\u65ad\u7684\u4e0d\u4e00\u81f4\u548c\u4e0d\u5bf9\u9f50\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-6fc53c6de45d13d802ebf87566cbfc17.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-6fc53c6de45d13d802ebf87566cbfc17.png\"><\/a><\/p>\n<p>\u4e3a\u4e86\u51cf\u5c11 LLMs \u7684\u504f\u89c1\u9884\u6d4b\uff0c\u4e4b\u524d\u7684\u5de5\u4f5c\u5f00\u53d1\u4e86\u6821\u51c6\u6280\u672f\uff08calibration\uff09\u4ee5\u51cf\u5c11 LLM \u9884\u6d4b\u4e2d\u7684\u504f\u89c1\u3002\u6211\u4eec\u5148\u5bf9\u6821\u51c6\u6280\u672f\u5728\u5bf9\u9f50\u5355\u70b9\uff08pointwise\uff09 LLM \u8bc4\u4f30\u5668\u7684\u6709\u6548\u6027\u8fdb\u884c\u4e86\u7cfb\u7edf\u5206\u6790\u3002\u5982\u4e0a\u56fe 2 \u6240\u793a\uff0c\u5373\u4f7f\u63d0\u4f9b\u4e86\u76d1\u7763\u6570\u636e\uff0c\u73b0\u6709\u7684\u6821\u51c6\u65b9\u6cd5\u4ecd\u7136\u4e0d\u80fd\u5f88\u597d\u7684\u5bf9\u9f50 LLM \u8bc4\u4f30\u5668\u3002<\/p>\n<p>\u5982\u516c\u5f0f 1 \u6240\u793a\uff0c\u6211\u4eec\u8ba4\u4e3a\u8bc4\u4f30\u4e0d\u5bf9\u9f50\u7684\u4e3b\u8981\u539f\u56e0\u5e76\u975e LLM \u8bc4\u4f30\u5206\u6570\u5206\u5e03\u7684\u5148\u9a8c\u5177\u6709\u504f\u89c1\uff08biased priors over evaluation score distribution\uff09\uff0c\u800c\u662f\u8bc4\u4f30\u6807\u51c6\uff08evaluation standard\uff09\u7684\u9519\u4f4d\uff0c\u5373 LLM \u8bc4\u4f30\u5668\u7684\u4f3c\u7136\uff08likelihood\uff09\u3002\u6211\u4eec\u8ba4\u4e3a\u505a\u6210\u5bf9\uff08pairwise\uff09\u8bc4\u4f30\u65f6\uff0cLLM \u8bc4\u4f30\u5668\u4f1a\u4e0e\u4eba\u7c7b\u6709\u66f4\u4e00\u81f4\u7684\u8bc4\u4ef7\u6807\u51c6\uff0c\u56e0\u6b64\uff0c\u6211\u4eec\u63a2\u7d22\u4e86\u4e00\u79cd\u65b0\u7684 LLM \u8bc4\u4f30\u8303\u5f0f\uff0c\u4ee5\u4fc3\u8fdb\u66f4\u5bf9\u9f50\u7684\u5224\u65ad\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-37c1caa4b7aa6b05c9635361eabe396d.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-37c1caa4b7aa6b05c9635361eabe396d.png\"><\/a><\/p>\n<p><strong>RLHF \u5e26\u6765\u7684\u542f\u53d1<\/strong><\/p>\n<p>\u5982\u4e0b\u56fe 1 \u6240\u793a\uff0c\u53d7\u5230 RLHF \u4e2d\u901a\u8fc7\u504f\u597d\u6570\u636e\u5bf9\u5956\u52b1\u6a21\u578b\u8fdb\u884c\u5bf9\u9f50\u7684\u542f\u53d1\uff0c\u6211\u4eec\u8ba4\u4e3a LLM \u8bc4\u4f30\u5668\u53ef\u4ee5\u901a\u8fc7\u751f\u6210\u504f\u597d\u6392\u5e8f\uff08preference ranking\uff09\u6765\u5f97\u5230\u66f4\u548c\u4eba\u7c7b\u5bf9\u9f50\u7684\u9884\u6d4b\u3002\u6700\u8fd1\u5df2\u6709\u4e00\u4e9b\u5de5\u4f5c\u5f00\u59cb\u901a\u8fc7\u8ba9 LLM \u8fdb\u884c\u6210\u5bf9\u6bd4\u8f83\uff08pairwise comparison\uff09\u6765\u5f97\u5230\u504f\u597d\u6392\u5e8f\u3002\u7136\u800c\uff0c\u8bc4\u4f30\u504f\u597d\u6392\u5e8f\u7684\u590d\u6742\u6027\u548c\u53ef\u6269\u5c55\u6027\u5728\u5f88\u5927\u7a0b\u5ea6\u4e0a\u88ab\u5ffd\u89c6\u4e86\u3002\u5b83\u4eec\u5ffd\u7565\u4e86\u4f20\u9012\u6027\u5047\u8bbe\uff08transitivity assumption\uff09\uff0c\u4f7f\u5f97\u6bd4\u8f83\u6b21\u6570\u7684\u590d\u6742\u5ea6\u4e3a O (N^2)\uff0c\u8ba9\u8bc4\u4f30\u8fc7\u7a0b\u53d8\u5f97\u6602\u8d35\u800c\u4e0d\u53ef\u884c\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-4b8d61027ab2caded9e63942454ea647.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-4b8d61027ab2caded9e63942454ea647.png\"><\/a><\/p>\n<p><strong>PairS\uff1a\u9ad8\u6548\u504f\u597d\u641c\u7d22\u7b97\u6cd5<\/strong><\/p>\n<p>\u5728\u672c\u5de5\u4f5c\u4e2d\uff0c\u6211\u4eec\u63d0\u51fa\u4e86\u4e24\u79cd\u6210\u5bf9\u504f\u597d\u641c\u7d22\u7b97\u6cd5\uff08PairS-greedy \u548c PairS-beam\uff09\u3002PairS-greedy \u662f\u57fa\u4e8e\u5b8c\u5168\u7684\u4f20\u9012\u6027\u5047\u8bbe\u548c\u5408\u5e76\u6392\u5e8f\uff08merge sort\uff09\u7684\u7b97\u6cd5\uff0c\u53ea\u9700\u8981\u901a\u8fc7 O (NlogN) \u7684\u590d\u6742\u5ea6\u5c31\u53ef\u4ee5\u5f97\u5230\u5168\u5c40\u7684\u504f\u597d\u6392\u5e8f\u3002\u4f20\u9012\u6027\u5047\u8bbe\u662f\u6307\uff0c\u6bd4\u5982\u5bf9\u4e8e 3 \u4e2a\u5019\u9009\u9879\uff0cLLM \u603b\u662f\u6709\u5982\u679c A\u227bB \u4ee5\u53ca B\u227bC\uff0c\u5219 A\u227bC\u3002\u5728\u8fd9\u4e2a\u5047\u8bbe\u4e0b\u6211\u4eec\u53ef\u4ee5\u76f4\u63a5\u7528\u4f20\u7edf\u7684<mark data-type=\"tech_methods\" data-id=\"05cfb700-bc4f-48a8-ba2a-78169045e42e\">\u6392\u5e8f\u7b97\u6cd5<\/mark>\u4ece\u6210\u5bf9\u504f\u597d\u4e2d\u83b7\u5f97\u504f\u597d\u6392\u5e8f\u3002<\/p>\n<p>\u4f46\u662f LLM \u5e76\u4e0d\u5177\u6709\u5b8c\u7f8e\u7684\u4f20\u9012\u6027\uff0c\u6240\u4ee5\u6211\u4eec\u53c8\u8bbe\u8ba1\u4e86 PairS-beam \u7b97\u6cd5\u3002\u5728\u66f4\u5bbd\u677e\u4f20\u9012\u6027\u5047\u8bbe\u4e0b\uff0c\u6211\u4eec\u63a8\u5bfc\u5e76\u5316\u7b80\u4e86\u504f\u597d\u6392\u5e8f\u7684<mark data-type=\"concepts\" data-id=\"c45d003e-8a8f-44f6-b820-9894b1d478d4\">\u4f3c\u7136\u51fd\u6570<\/mark>\uff08likelihood function\uff09\u3002PairS-beam \u5728\u5408\u5e76<mark data-type=\"tech_methods\" data-id=\"05cfb700-bc4f-48a8-ba2a-78169045e42e\">\u6392\u5e8f\u7b97\u6cd5<\/mark>\u7684\u6bcf\u4e00\u6b21\u7684\u5408\u5e76\u64cd\u4f5c\uff08merge operation\uff09\u4e2d\u6309\u4f3c\u7136\u503c\u505a\u96c6\u675f\u641c\u7d22\uff0c\u5e76\u901a\u8fc7\u504f\u597d\u7684\u4e0d\u786e\u5b9a\u6027\uff08uncertainty\uff09\u6765\u51cf\u679d\u6210\u5bf9\u6bd4\u8f83\u7684\u7a7a\u95f4\u7684\u641c\u7d22\u65b9\u6cd5\u3002PairS-beam \u53ef\u4ee5\u8c03\u6574\u5bf9\u6bd4\u590d\u6742\u5ea6\u548c\u6392\u5e8f\u8d28\u91cf\uff0c \u9ad8\u6548\u7684\u7ed9\u51fa\u504f\u597d\u6392\u5e8f\u7684<mark data-type=\"tech_methods\" data-id=\"378d0ed1-b022-4de1-b3de-58b6f4722361\">\u6700\u5927\u4f3c\u7136\u4f30\u8ba1<\/mark>\uff08MLE\uff09\u3002\u5728\u4e0b\u56fe 3 \u4e2d\u6211\u4eec\u5c55\u793a\u4e86\u4e00\u4e2a PairS-beam \u5982\u4f55\u505a\u5408\u5e76\u64cd\u4f5c\u7684\u4f8b\u5b50\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-dca20ce5c7124dc67975d2c322664d2b.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-dca20ce5c7124dc67975d2c322664d2b.png\"><\/a><\/p>\n<p><strong>\u5b9e\u9a8c\u7ed3\u679c<\/strong><\/p>\n<p>\u6211\u4eec\u5728\u591a\u4e2a\u5177\u6709\u4ee3\u8868\u6027\u7684\u6570\u636e\u96c6\u4e0a\u8fdb\u884c\u4e86\u6d4b\u8bd5\uff0c\u5305\u62ec\u95ed\u5408\u5f0f\u751f\u6210\u7684\u7f29\u5199\u4efb\u52a1NewsRoom \u548c SummEval\uff0c\u548c\u5f00\u653e\u5f0f\u7684\u6545\u4e8b\u751f\u6210\u4efb\u52a1HANNA\uff0c\u5e76\u5bf9\u6bd4\u4e86\u591a\u4e2a LLM \u5355\u70b9\u8bc4\u4f30\u7684\u57fa\u7ebf\u65b9\u6cd5\uff0c\u5305\u62ec\u65e0\u76d1\u7763\u7684 direct scoring, G-Eval, GPTScore \u548c\u6709\u76d1\u7763\u8bad\u7ec3\u8fc7\u7684 UniEval \u4ee5\u53ca BARTScore\u3002\u5982\u4e0b\u8868 1 \u6240\u793a\uff0cPairS \u5728\u6bcf\u4e2a\u4efb\u52a1\u4e0a\u548c\u4ed6\u4eec\u76f8\u6bd4\u90fd\u6709\u7740\u548c\u4eba\u7c7b\u8bc4\u5206\u66f4\u9ad8\u7684\u4e00\u81f4\u6027\u3002GPT-4-turbo \u66f4\u662f\u80fd\u8fbe\u5230 SOTA \u7684\u6548\u679c\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-da966ddf31123c14a487bbbf8401ca8d.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-da966ddf31123c14a487bbbf8401ca8d.png\"><\/a><\/p>\n<p>\u5728\u6587\u7ae0\u4e2d\uff0c\u6211\u4eec\u8fd8\u5bf9\u6bd4\u4e86\u4e24\u79cd\u504f\u597d\u6392\u5e8f\u7684\u57fa\u7ebf\u65b9\u6cd5\uff0cwin rate \u548c ELO rating\u3002PairS \u53ef\u4ee5\u4ec5\u7528\u7ea6 30% \u7684\u5bf9\u6bd4\u6b21\u6570\u5c31\u80fd\u8fbe\u5230\u4ed6\u4eec\u540c\u6837\u8d28\u91cf\u7684\u504f\u597d\u6392\u5e8f\u3002\u8bba\u6587\u8fd8\u63d0\u4f9b\u4e86\u66f4\u591a\u5173\u4e8e\u5982\u4f55\u4f7f\u7528\u6210\u5bf9\u504f\u597d\u6765<mark data-type=\"tech_tasks\" data-id=\"e1abaa20-0000-4ff4-ad42-f97ca7fca4b1\">\u91cf\u5316<\/mark>\u8ba1\u7b97 LLM \u8bc4\u4f30\u5668\u7684\u4f20\u9012\u6027\uff0c\u4ee5\u53ca\u6210\u5bf9\u8bc4\u4f30\u5668\u5982\u4f55\u5728\u6821\u51c6\u4e2d\u53d7\u76ca\u7684\u89c1\u89e3\u3002<\/p>\n<p>\u66f4\u591a\u7814\u7a76\u7ec6\u8282\uff0c\u53ef\u53c2\u8003\u539f\u8bba\u6587\u3002<\/p>\n<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-08-04\" target=\"_blank\">\u9996\u5c4a\u5927\u6a21\u578b\u9876\u4f1aCOLM \u9ad8\u5206\u8bba\u6587\uff1a\u504f\u597d\u641c\u7d22\u7b97\u6cd5PairS\uff0c\u8ba9\u5927\u6a21\u578b\u8fdb\u884c\u6587\u672c\u8bc4\u4f30\u66f4\u9ad8\u6548<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:\u9996\u5c4a\u5927\u6a21\u578b\u9876\u4f1aCOLM [&hellip;]<\/p>\n","protected":false},"author":3,"featured_media":0,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[27],"tags":[68],"class_list":["post-34908","post","type-post","status-publish","format-standard","hentry","category-news","tag-68"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.9 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>\u9996\u5c4a\u5927\u6a21\u578b\u9876\u4f1aCOLM \u9ad8\u5206\u8bba\u6587\uff1a\u504f\u597d\u641c\u7d22\u7b97\u6cd5PairS\uff0c\u8ba9\u5927\u6a21\u578b\u8fdb\u884c\u6587\u672c\u8bc4\u4f30\u66f4\u9ad8\u6548 - \u4e00\u8d77AI\u6280\u672f<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/17aitech.com\/?p=34908\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34908#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34908\"},\"author\":{\"name\":\"AI\u5c0f\u52a9\u624b\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\"},\"headline\":\"\u9996\u5c4a\u5927\u6a21\u578b\u9876\u4f1aCOLM \u9ad8\u5206\u8bba\u6587\uff1a\u504f\u597d\u641c\u7d22\u7b97\u6cd5PairS\uff0c\u8ba9\u5927\u6a21\u578b\u8fdb\u884c\u6587\u672c\u8bc4\u4f30\u66f4\u9ad8\u6548\",\"datePublished\":\"2024-12-14T21:58:57+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34908\"},\"wordCount\":141,\"commentCount\":0,\"image\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34908#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/08\\\/frc-2f91a5574163de717c5146ac0a709263.png\",\"keywords\":[\"\u5927\u6a21\u578b\"],\"articleSection\":[\"\u884c\u4e1a\u8d44\u8baf\"],\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/17aitech.com\\\/?p=34908#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34908\",\"url\":\"https:\\\/\\\/17aitech.com\\\/?p=34908\",\"name\":\"\u9996\u5c4a\u5927\u6a21\u578b\u9876\u4f1aCOLM \u9ad8\u5206\u8bba\u6587\uff1a\u504f\u597d\u641c\u7d22\u7b97\u6cd5PairS\uff0c\u8ba9\u5927\u6a21\u578b\u8fdb\u884c\u6587\u672c\u8bc4\u4f30\u66f4\u9ad8\u6548 - \u4e00\u8d77AI\u6280\u672f\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34908#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34908#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/08\\\/frc-2f91a5574163de717c5146ac0a709263.png\",\"datePublished\":\"2024-12-14T21:58:57+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\"},\"breadcrumb\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34908#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/17aitech.com\\\/?p=34908\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34908#primaryimage\",\"url\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/08\\\/frc-2f91a5574163de717c5146ac0a709263.png\",\"contentUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/08\\\/frc-2f91a5574163de717c5146ac0a709263.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34908#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\\\/\\\/17aitech.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u9996\u5c4a\u5927\u6a21\u578b\u9876\u4f1aCOLM \u9ad8\u5206\u8bba\u6587\uff1a\u504f\u597d\u641c\u7d22\u7b97\u6cd5PairS\uff0c\u8ba9\u5927\u6a21\u578b\u8fdb\u884c\u6587\u672c\u8bc4\u4f30\u66f4\u9ad8\u6548\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/#website\",\"url\":\"https:\\\/\\\/17aitech.com\\\/\",\"name\":\"\u4e00\u8d77AI\u6280\u672f\",\"description\":\"\u8ba9AI\u77e5\u8bc6\u89e6\u624b\u53ef\u53ca\",\"alternateName\":\"\u4e00\u8d77AI\u6280\u672f\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/17aitech.com\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\",\"name\":\"AI\u5c0f\u52a9\u624b\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/04\\\/robot_3.png\",\"url\":\"\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/04\\\/robot_3.png\",\"contentUrl\":\"\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/04\\\/robot_3.png\",\"caption\":\"AI\u5c0f\u52a9\u624b\"},\"description\":\"\u8fd9\u4e2a\u4eba\u5f88\u61d2\uff0c\u4ec0\u4e48\u90fd\u6ca1\u6709\u7559\u4e0b\uff5e\",\"url\":\"https:\\\/\\\/17aitech.com\\\/?page_id=33738&user=3\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u9996\u5c4a\u5927\u6a21\u578b\u9876\u4f1aCOLM \u9ad8\u5206\u8bba\u6587\uff1a\u504f\u597d\u641c\u7d22\u7b97\u6cd5PairS\uff0c\u8ba9\u5927\u6a21\u578b\u8fdb\u884c\u6587\u672c\u8bc4\u4f30\u66f4\u9ad8\u6548 - \u4e00\u8d77AI\u6280\u672f","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/17aitech.com\/?p=34908","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/17aitech.com\/?p=34908#article","isPartOf":{"@id":"https:\/\/17aitech.com\/?p=34908"},"author":{"name":"AI\u5c0f\u52a9\u624b","@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"headline":"\u9996\u5c4a\u5927\u6a21\u578b\u9876\u4f1aCOLM \u9ad8\u5206\u8bba\u6587\uff1a\u504f\u597d\u641c\u7d22\u7b97\u6cd5PairS\uff0c\u8ba9\u5927\u6a21\u578b\u8fdb\u884c\u6587\u672c\u8bc4\u4f30\u66f4\u9ad8\u6548","datePublished":"2024-12-14T21:58:57+00:00","mainEntityOfPage":{"@id":"https:\/\/17aitech.com\/?p=34908"},"wordCount":141,"commentCount":0,"image":{"@id":"https:\/\/17aitech.com\/?p=34908#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-2f91a5574163de717c5146ac0a709263.png","keywords":["\u5927\u6a21\u578b"],"articleSection":["\u884c\u4e1a\u8d44\u8baf"],"inLanguage":"zh-Hans","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/17aitech.com\/?p=34908#respond"]}]},{"@type":"WebPage","@id":"https:\/\/17aitech.com\/?p=34908","url":"https:\/\/17aitech.com\/?p=34908","name":"\u9996\u5c4a\u5927\u6a21\u578b\u9876\u4f1aCOLM \u9ad8\u5206\u8bba\u6587\uff1a\u504f\u597d\u641c\u7d22\u7b97\u6cd5PairS\uff0c\u8ba9\u5927\u6a21\u578b\u8fdb\u884c\u6587\u672c\u8bc4\u4f30\u66f4\u9ad8\u6548 - \u4e00\u8d77AI\u6280\u672f","isPartOf":{"@id":"https:\/\/17aitech.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/17aitech.com\/?p=34908#primaryimage"},"image":{"@id":"https:\/\/17aitech.com\/?p=34908#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-2f91a5574163de717c5146ac0a709263.png","datePublished":"2024-12-14T21:58:57+00:00","author":{"@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"breadcrumb":{"@id":"https:\/\/17aitech.com\/?p=34908#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/17aitech.com\/?p=34908"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/?p=34908#primaryimage","url":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-2f91a5574163de717c5146ac0a709263.png","contentUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/08\/frc-2f91a5574163de717c5146ac0a709263.png"},{"@type":"BreadcrumbList","@id":"https:\/\/17aitech.com\/?p=34908#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/17aitech.com\/"},{"@type":"ListItem","position":2,"name":"\u9996\u5c4a\u5927\u6a21\u578b\u9876\u4f1aCOLM \u9ad8\u5206\u8bba\u6587\uff1a\u504f\u597d\u641c\u7d22\u7b97\u6cd5PairS\uff0c\u8ba9\u5927\u6a21\u578b\u8fdb\u884c\u6587\u672c\u8bc4\u4f30\u66f4\u9ad8\u6548"}]},{"@type":"WebSite","@id":"https:\/\/17aitech.com\/#website","url":"https:\/\/17aitech.com\/","name":"\u4e00\u8d77AI\u6280\u672f","description":"\u8ba9AI\u77e5\u8bc6\u89e6\u624b\u53ef\u53ca","alternateName":"\u4e00\u8d77AI\u6280\u672f","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/17aitech.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d","name":"AI\u5c0f\u52a9\u624b","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","url":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","contentUrl":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","caption":"AI\u5c0f\u52a9\u624b"},"description":"\u8fd9\u4e2a\u4eba\u5f88\u61d2\uff0c\u4ec0\u4e48\u90fd\u6ca1\u6709\u7559\u4e0b\uff5e","url":"https:\/\/17aitech.com\/?page_id=33738&user=3"}]}},"_links":{"self":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/34908","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=34908"}],"version-history":[{"count":0,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/34908\/revisions"}],"wp:attachment":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=34908"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=34908"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=34908"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}