{"id":120,"date":"2026-07-12T02:22:13","date_gmt":"2026-07-12T02:22:13","guid":{"rendered":"https:\/\/blog.top-api.cc\/?p=120"},"modified":"2026-07-12T02:22:13","modified_gmt":"2026-07-12T02:22:13","slug":"llm-as-judge-bias-calibration-review-20260712","status":"publish","type":"post","link":"https:\/\/blog.top-api.cc\/?p=120","title":{"rendered":"LLM-as-a-Judge \u600e\u4e48\u907f\u514d\u81ea\u55e8\uff1a\u4f4d\u7f6e\u504f\u5dee\u3001\u8bc4\u5206\u6f02\u79fb\u548c\u4eba\u5de5\u6821\u51c6\u8981\u4e00\u8d77\u6d4b"},"content":{"rendered":"<p>\u8ba9\u4e00\u4e2a\u6a21\u578b\u7ed9\u56de\u7b54\u6253\u5206\uff0c\u80fd\u628a\u8bc4\u6d4b\u901f\u5ea6\u4ece\u4eba\u5de5\u5c0f\u65f6\u7f29\u77ed\u5230\u5206\u949f\uff0c\u4f46\u5b83\u4e0d\u662f\u5ba2\u89c2\u88c1\u5224\u3002\u8bc4\u5ba1\u6a21\u578b\u53ef\u80fd\u504f\u7231\u66f4\u957f\u3001\u66f4\u50cf\u6a21\u677f\u7684\u56de\u7b54\uff0c\u53ef\u80fd\u603b\u662f\u9009\u62e9\u7b2c\u4e00\u4e2a\u5019\u9009\uff0c\u4e5f\u53ef\u80fd\u5bf9\u540c\u5bb6\u65cf\u6a21\u578b\u66f4\u5bbd\u5bb9\u3002\u6ca1\u6709\u6821\u51c6\u7684\u8bc4\u5206\u8868\uff0c\u5f88\u5bb9\u6613\u8ba9\u56e2\u961f\u4f18\u5316\u5230\u201c\u8ba8\u597d\u8bc4\u5ba1\u201d\u800c\u4e0d\u662f\u89e3\u51b3\u7528\u6237\u95ee\u9898\u3002<\/p>\n<h2>\u5148\u505a\u987a\u5e8f\u548c\u957f\u5ea6\u76f2\u6d4b<\/h2>\n<p>\u628a A\/B \u56de\u7b54\u968f\u673a\u4ea4\u6362\u4f4d\u7f6e\uff0c\u89c2\u5bdf\u80dc\u7387\u662f\u5426\u53d8\u5316\uff1b\u628a\u540c\u4e00\u7b54\u6848\u538b\u7f29\u6216\u6269\u5199\uff0c\u89c2\u5bdf\u5206\u6570\u662f\u5426\u88ab\u957f\u5ea6\u5f71\u54cd\u3002\u5bf9\u4e8b\u5b9e\u95ee\u7b54\u3001\u5de5\u5177\u4efb\u52a1\u3001\u5199\u4f5c\u548c\u5b89\u5168\u7b56\u7565\u5206\u522b\u5efa\u6837\u672c\uff0c\u56e0\u4e3a\u540c\u4e00\u79cd\u8bc4\u5ba1\u63d0\u793a\u8bcd\u4e0d\u9002\u5408\u6240\u6709\u4efb\u52a1\u3002\u8bc4\u5206\u7ed3\u679c\u5e94\u4fdd\u7559\u8bc4\u5ba1\u7248\u672c\u3001\u63d0\u793a\u8bcd\u7248\u672c\u3001\u6e29\u5ea6\u548c\u539f\u59cb\u7406\u7531\u6458\u8981\u3002<\/p>\n<p>\u4eba\u5de5\u6807\u6ce8\u4e0d\u662f\u4e3a\u4e86\u66ff\u4ee3\u5168\u90e8\u81ea\u52a8\u8bc4\u6d4b\uff0c\u800c\u662f\u4e3a\u4e86\u5efa\u7acb\u951a\u70b9\u3002\u5b9a\u671f\u62bd\u53d6\u6709\u4e89\u8bae\u3001\u4f4e\u7f6e\u4fe1\u548c\u9ad8\u5f71\u54cd\u6837\u672c\uff0c\u7531\u81f3\u5c11\u4e24\u4f4d\u5ba1\u9605\u8005\u72ec\u7acb\u6807\u6ce8\uff0c\u518d\u6bd4\u8f83\u4eba\u673a\u4e00\u81f4\u6027\u548c\u4eba\u5de5\u4e4b\u95f4\u7684\u4e00\u81f4\u6027\u3002\u82e5\u8bc4\u5ba1\u6a21\u578b\u4e0e\u4eba\u5de5\u6301\u7eed\u504f\u79bb\uff0c\u5e94\u4fee\u8ba2\u4efb\u52a1\u5b9a\u4e49\u6216\u6362\u8bc4\u5ba1\u65b9\u6cd5\uff0c\u800c\u4e0d\u662f\u5f3a\u884c\u76f8\u4fe1\u81ea\u52a8\u5206\u6570\u3002<\/p>\n<h2>\u9632\u6b62\u8bc4\u5206\u968f\u65f6\u95f4\u6f02\u79fb<\/h2>\n<p>\u8bc4\u5ba1\u6a21\u578b\u3001\u7cfb\u7edf\u63d0\u793a\u8bcd\u3001\u5de5\u5177\u548c\u6570\u636e\u96c6\u53d8\u5316\u540e\uff0c\u5386\u53f2\u5206\u6570\u53ef\u80fd\u4e0d\u518d\u53ef\u6bd4\u3002\u8981\u4fdd\u7559\u56fa\u5b9a\u57fa\u51c6\u96c6\uff0c\u6309\u7248\u672c\u91cd\u8dd1\uff1b\u5bf9\u5206\u6570\u7a81\u7136\u53d8\u5316\u7684\u4efb\u52a1\u68c0\u67e5\u662f\u5019\u9009\u6a21\u578b\u53d8\u4e86\u3001\u8bc4\u5ba1\u6a21\u578b\u53d8\u4e86\uff0c\u8fd8\u662f\u6570\u636e\u5206\u5e03\u53d8\u4e86\u3002\u8bc4\u5206\u7406\u7531\u4e5f\u53ea\u80fd\u4f5c\u4e3a\u8f85\u52a9\u8bc1\u636e\uff0c\u4e0d\u80fd\u81ea\u52a8\u5f53\u4f5c\u4e8b\u5b9e\u3002<\/p>\n<p>DeepEval\u3001PromptBench \u7b49\u6846\u67b6\u63d0\u4f9b\u8bc4\u6d4b\u7f16\u6392\u80fd\u529b\uff0c\u4f46\u6307\u6807\u5b9a\u4e49\u548c\u4eba\u5de5\u6821\u51c6\u4ecd\u7531\u56e2\u961f\u8d1f\u8d23\u3002\u5de5\u5177\u53ef\u4ee5\u8ba9\u5b9e\u9a8c\u53ef\u91cd\u590d\uff0c\u5374\u4e0d\u4f1a\u81ea\u52a8\u6d88\u9664\u504f\u89c1\u3002<\/p>\n<p>\u5982\u679c\u9700\u8981\u6bd4\u8f83\u591a\u4e2a\u5019\u9009\u6a21\u578b\uff0c\u53ef\u5728 <a href=\"https:\/\/top-api.cc\">top-api.cc<\/a> \u56fa\u5b9a\u8bf7\u6c42\u3001\u6a21\u578b\u53c2\u6570\u548c\u6837\u672c\u7248\u672c\uff0c\u518d\u628a\u5019\u9009\u8f93\u51fa\u4ea4\u7ed9\u72ec\u7acb\u7684\u8bc4\u5ba1\u6d41\u7a0b\u3002\u4e0d\u8981\u8ba9\u540c\u4e00\u6a21\u578b\u65e2\u751f\u6210\u7b54\u6848\u53c8\u5355\u72ec\u51b3\u5b9a\u81ea\u5df1\u662f\u5426\u6700\u597d\u3002<\/p>\n<h2>\u6700\u5c0f\u8bc4\u6d4b\u77e9\u9635<\/h2>\n<ul>\n<li>A\/B \u4f4d\u7f6e\u4ea4\u6362\u540e\u7684\u80dc\u7387\uff1b<\/li>\n<li>\u4e0d\u540c\u957f\u5ea6\u548c\u6587\u98ce\u7684\u8bc4\u5206\u53d8\u5316\uff1b<\/li>\n<li>\u81ea\u52a8\u8bc4\u5206\u4e0e\u4eba\u5de5\u8bc4\u5206\u7684\u4e00\u81f4\u6027\uff1b<\/li>\n<li>\u8bc4\u5ba1\u6a21\u578b\u548c\u63d0\u793a\u8bcd\u7248\u672c\u6f02\u79fb\uff1b<\/li>\n<li>\u9ad8\u98ce\u9669\u4efb\u52a1\u7684\u4eba\u5de5\u590d\u6838\u8986\u76d6\u7387\uff1b<\/li>\n<li>\u4f4e\u7f6e\u4fe1\u3001\u5e73\u5206\u548c\u51b2\u7a81\u6837\u672c\u7684\u5904\u7406\u65b9\u5f0f\u3002<\/li>\n<\/ul>\n<p>\u53c2\u8003\u8d44\u6599\uff1aDeepEval\u3001PromptBench \u7684\u516c\u5f00\u6587\u6863\u4e0e LLM-as-a-Judge \u7814\u7a76\u3002\u5206\u6570\u662f\u51b3\u7b56\u8f93\u5165\uff0c\u4e0d\u5e94\u66ff\u4ee3\u660e\u786e\u7684\u4e1a\u52a1\u9a8c\u6536\u6807\u51c6\u3002<\/p>\n<p>\u5728 <a href=\"https:\/\/top-api.cc\">top-api.cc<\/a> \u4fdd\u5b58\u5019\u9009\u8c03\u7528\u5143\u6570\u636e\u540e\uff0c\u53ef\u4ee5\u5728\u8bc4\u5ba1\u7ed3\u679c\u5f02\u5e38\u65f6\u5feb\u901f\u56de\u653e\u76f8\u540c\u6837\u672c\uff0c\u800c\u4e0d\u4f1a\u628a\u53c2\u6570\u53d8\u5316\u6df7\u5165\u7ed3\u8bba\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u7528\u6a21\u578b\u7ed9\u6a21\u578b\u6253\u5206\u5f88\u9ad8\u6548\uff0c\u4f46\u4e5f\u5bb9\u6613\u53d7\u5230\u7b54\u6848\u987a\u5e8f\u3001\u6587\u98ce\u957f\u5ea6\u3001\u6a21\u578b\u504f\u597d\u548c\u8bc4\u5ba1\u63d0\u793a\u8bcd\u53d8\u5316\u5f71\u54cd\u3002LLM-as-a-Judge \u5fc5\u987b\u548c\u4eba\u5de5\u6837\u672c\u3001\u76f2\u6d4b\u548c\u6f02\u79fb\u76d1\u63a7\u4e00\u8d77\u4f7f\u7528\u3002<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-120","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/blog.top-api.cc\/index.php?rest_route=\/wp\/v2\/posts\/120","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/blog.top-api.cc\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blog.top-api.cc\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blog.top-api.cc\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/blog.top-api.cc\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=120"}],"version-history":[{"count":0,"href":"https:\/\/blog.top-api.cc\/index.php?rest_route=\/wp\/v2\/posts\/120\/revisions"}],"wp:attachment":[{"href":"https:\/\/blog.top-api.cc\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=120"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blog.top-api.cc\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=120"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blog.top-api.cc\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=120"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}