{"id":34738,"date":"2024-12-12T01:57:53","date_gmt":"2024-12-11T17:57:53","guid":{"rendered":"https:\/\/17aitech.com\/?p=34738"},"modified":"2024-12-12T01:57:53","modified_gmt":"2024-12-11T17:57:53","slug":"rlhf%e4%b8%8d%e5%a4%9f%e7%94%a8%e4%ba%86%ef%bc%8copenai%e8%ae%be%e8%ae%a1%e5%87%ba%e4%ba%86%e6%96%b0%e7%9a%84%e5%a5%96%e5%8a%b1%e6%9c%ba%e5%88%b6","status":"publish","type":"post","link":"https:\/\/17aitech.com\/?p=34738","title":{"rendered":"RLHF\u4e0d\u591f\u7528\u4e86\uff0cOpenAI\u8bbe\u8ba1\u51fa\u4e86\u65b0\u7684\u5956\u52b1\u673a\u5236"},"content":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-07-25-3\" target=\"_blank\">RLHF\u4e0d\u591f\u7528\u4e86\uff0cOpenAI\u8bbe\u8ba1\u51fa\u4e86\u65b0\u7684\u5956\u52b1\u673a\u5236<\/a><\/p>\n<blockquote data-author-name=\"\" data-content-utf8-length=\"22\" data-source-title=\"\" data-type=\"2\" data-url=\"\">\n<section>\n<p>OpenAI \u7684\u65b0\u5956\u52b1\u673a\u5236\uff0c\u8ba9\u5927\u6a21\u578b\u66f4\u542c\u8bdd\u4e86\u3002<\/p>\n<\/section>\n<\/blockquote>\n<p>\u81ea\u5927\u6a21\u578b\u5174\u8d77\u4ee5\u6765\uff0c\u4f7f\u7528<mark data-type=\"tech_methods\" data-id=\"ee1a8f69-3170-4ddf-b2b6-47d91c844425\">\u5f3a\u5316\u5b66\u4e60<\/mark>\u4ece\u4eba\u7c7b\u53cd\u9988\uff08RLHF\uff09\u4e2d\u5fae\u8c03<mark data-type=\"tech_tasks\" data-id=\"bf35ef94-d956-4033-a533-0c0828308c36\">\u8bed\u8a00\u6a21\u578b<\/mark>\u4e00\u76f4\u662f\u786e\u4fdd AI \u51c6\u786e\u9075\u5faa\u6307\u4ee4\u7684\u9996\u9009\u65b9\u6cd5\u3002<\/p>\n<p>\u4e3a\u4e86\u786e\u4fdd AI \u7cfb\u7edf\u5b89\u5168\u8fd0\u884c\u5e76\u4e0e\u4eba\u7c7b\u4ef7\u503c\u89c2\u4fdd\u6301\u4e00\u81f4\uff0c\u6211\u4eec\u9700\u8981\u5b9a\u4e49\u671f\u671b\u884c\u4e3a\u5e76\u6536\u96c6\u4eba\u7c7b\u53cd\u9988\u6765\u8bad\u7ec3\u300c\u5956\u52b1\u6a21\u578b\u300d\u3002\u8fd9\u79cd\u6a21\u578b\u901a\u8fc7\u53d1\u51fa\u671f\u671b\u7684\u52a8\u4f5c\u6765\u6307\u5bfc AI\u3002\u4f46\u662f\uff0c\u6536\u96c6\u8fd9\u4e9b\u5e38\u89c4\u548c\u91cd\u590d\u4efb\u52a1\u7684\u4eba\u7c7b\u53cd\u9988\u901a\u5e38\u6548\u7387\u4e0d\u9ad8\u3002\u6b64\u5916\uff0c\u5982\u679c\u5b89\u5168\u653f\u7b56\u53d1\u751f\u53d8\u5316\uff0c\u5df2\u7ecf\u6536\u96c6\u7684\u53cd\u9988\u53ef\u80fd\u4f1a\u8fc7\u65f6\uff0c\u9700\u8981\u65b0\u7684\u6570\u636e\u3002<\/p>\n<p>\u6211\u4eec\u80fd\u5426\u6784\u5efa\u4e00\u79cd\u65b0\u7684\u673a\u5236\u6765\u5b8c\u6210\u8fd9\u4e9b\u4efb\u52a1\uff1f\u8fd1\u65e5\uff0cOpenAI \u516c\u5e03\u4e86\u4e00\u79cd\u6559\u5bfc AI \u6a21\u578b\u9075\u5b88\u5b89\u5168\u653f\u7b56\u7684\u65b0\u65b9\u6cd5\uff0c\u79f0\u4e3a\u57fa\u4e8e\u89c4\u5219\u7684\u5956\u52b1\uff08Rule-Based Rewards\uff0cRBR\uff09\u3002\u00a0<\/p>\n<p>\u76f8\u5173\u8bba\u6587\u5df2\u7ecf\u653e\u51fa\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-3833915be2182c19caea6e98194f8efb.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-3833915be2182c19caea6e98194f8efb.png\"><\/a><\/p>\n<ul>\n<li>\n<p>\u8bba\u6587\u6807\u9898\uff1aRule Based Rewards for Language Model Safety<\/p>\n<\/li>\n<li>\n<p>\u8bba\u6587\u5730\u5740\uff1ahttps:\/\/cdn.openai.com\/rule-based-rewards-for-language-model-safety.pdf<\/p>\n<\/li>\n<li>\n<p>\u4ee3\u7801\u94fe\u63a5\uff1ahttps:\/\/github.com\/openai\/safety-rbr-code-and-data<\/p>\n<\/li>\n<\/ul>\n<p>\u8bba\u6587\u4f5c\u8005\u4e4b\u4e00\u3001OpenAI \u5b89\u5168\u7cfb\u7edf\u8d1f\u8d23\u4eba Lilian Weng \u8868\u793a\uff0c\u300cRBR \u53ef\u4ee5\u81ea\u52a8\u6267\u884c\u4e00\u4e9b\u6a21\u578b\u5fae\u8c03\u3002\u4f20\u7edf\u4e0a\uff0c \u6211\u4eec\u4f9d\u8d56\u4e8e\u6765\u81ea\u4eba\u7c7b\u53cd\u9988\u7684<mark data-type=\"tech_methods\" data-id=\"ee1a8f69-3170-4ddf-b2b6-47d91c844425\">\u5f3a\u5316\u5b66\u4e60<\/mark>\u4f5c\u4e3a\u9ed8\u8ba4\u7684\u5bf9\u9f50\u8bad\u7ec3\u65b9\u6cd5\u6765\u8bad\u7ec3\u6a21\u578b\uff0c\u8fd9\u786e\u5b9e\u6709\u6548\u3002\u7136\u800c\u5728\u5b9e\u8df5\u4e2d\uff0c\u6211\u4eec\u9762\u4e34\u7684\u6311\u6218\u662f\uff0c\u6211\u4eec\u82b1\u4e86\u5f88\u591a\u65f6\u95f4\u8ba8\u8bba\u653f\u7b56\u7684\u7ec6\u8282\uff0c\u800c\u5230\u6700\u540e\uff0c\u653f\u7b56\u53ef\u80fd\u5df2\u7ecf\u53d1\u751f\u4e86\u53d8\u5316\u3002\u300d<\/p>\n<p>RBR \u6839\u636e\u4e00\u7ec4\u5b89\u5168\u89c4\u5219\u63d0\u4f9b RL \u4fe1\u53f7\uff0c\u4f7f\u5176\u66f4\u5bb9\u6613\u9002\u5e94\u4e0d\u65ad\u53d8\u5316\u7684\u5b89\u5168\u653f\u7b56\uff0c\u800c\u65e0\u9700\u4e25\u91cd\u4f9d\u8d56\u4eba\u7c7b\u6570\u636e\u3002\u6b64\u5916\uff0c\u501f\u52a9 RBR\uff0c\u7814\u7a76\u8005\u80fd\u591f\u4ee5\u66f4\u7edf\u4e00\u7684\u89c6\u89d2\u770b\u5f85\u5b89\u5168\u6027\u548c\u6a21\u578b\u80fd\u529b\uff0c\u56e0\u4e3a\u66f4\u5f3a\u5927\u7684\u5206\u7ea7\u6a21\u578b\u53ef\u4ee5\u63d0\u4f9b\u66f4\u9ad8\u8d28\u91cf\u7684 RL \u4fe1\u53f7\u3002<\/p>\n<p>OpenAI \u8868\u793a\u81ea GPT-4 \u53d1\u5e03\u4ee5\u6765\uff0c\u4ed6\u4eec\u4e00\u76f4\u5c06 RBR \u7528\u4f5c\u5b89\u5168\u5806\u6808\u7684\u4e00\u90e8\u5206\uff0c\u5305\u62ec GPT-4o mini\uff0c\u5e76\u8ba1\u5212\u5728\u672a\u6765\u7684\u6a21\u578b\u4e2d\u5b9e\u65bd\u5b83\u3002<\/p>\n<p><strong>\u4e3a\u4ec0\u4e48\u8981\u63d0\u51fa RBR\uff1f \u00a0<\/strong><\/p>\n<p>\u968f\u7740\u5927\u578b<mark data-type=\"tech_tasks\" data-id=\"bf35ef94-d956-4033-a533-0c0828308c36\">\u8bed\u8a00\u6a21\u578b<\/mark>\uff08LLM\uff09\u529f\u80fd\u7684\u589e\u5f3a\u548c\u666e\u53ca\uff0c\u786e\u4fdd\u5176\u5b89\u5168\u6027\u548c\u5bf9\u9f50\u53d8\u5f97\u8d8a\u6765\u8d8a\u91cd\u8981\u3002\u6700\u8fd1\u7684\u8bb8\u591a\u5de5\u4f5c\u90fd\u96c6\u4e2d\u5728\u4f7f\u7528\u4eba\u7c7b\u504f\u597d\u6570\u636e\u6765\u8c03\u6574\u6a21\u578b\u4e0a\uff0c\u4f8b\u5982\u57fa\u4e8e\u4eba\u7c7b\u53cd\u9988\u7684<mark data-type=\"tech_methods\" data-id=\"ee1a8f69-3170-4ddf-b2b6-47d91c844425\">\u5f3a\u5316\u5b66\u4e60<\/mark>\uff08RLHF\uff09\u3002<\/p>\n<p>\u7136\u800c\uff0c\u4ec5\u4f7f\u7528\u4eba\u7c7b\u53cd\u9988\u6765\u5b9e\u73b0\u76ee\u6807\u5b89\u5168\u89c4\u8303\u8fd8\u9762\u4e34\u8bb8\u591a\u6311\u6218\u3002\u4e3a\u6a21\u578b\u5b89\u5168\u6027\u6536\u96c6\u548c\u7ef4\u62a4\u4eba\u7c7b\u6570\u636e\u901a\u5e38\u65e2\u8d39\u94b1\u53c8\u8d39\u65f6\uff0c\u800c\u4e14\u968f\u7740\u6a21\u578b\u80fd\u529b\u7684\u63d0\u9ad8\u6216\u7528\u6237\u884c\u4e3a\u7684\u6539\u53d8\uff0c\u5b89\u5168\u51c6\u5219\u4e5f\u4f1a\u53d1\u751f\u53d8\u5316\uff0c\u8fd9\u4e9b\u6570\u636e\u53ef\u80fd\u4f1a\u8fc7\u65f6\u3002\u5373\u4f7f\u8981\u6c42\u76f8\u5bf9\u7a33\u5b9a\uff0c\u4e5f\u5f88\u96be\u5411\u6ce8\u91ca\u8005\u4f20\u8fbe\u3002\u5b89\u5168\u65b9\u9762\u7684\u60c5\u51b5\u5c24\u5176\u5982\u6b64\uff0c\u56e0\u4e3a\u6240\u9700\u7684\u6a21\u578b\u54cd\u5e94\u975e\u5e38\u590d\u6742\uff0c\u9700\u8981\u5bf9\u662f\u5426\u54cd\u5e94\u4ee5\u53ca\u5982\u4f55\u54cd\u5e94\u8bf7\u6c42\u505a\u51fa\u7ec6\u5fae\u5dee\u522b\u3002\u5982\u679c\u8bf4\u660e\u4e0d\u591f\u660e\u786e\uff0c\u6ce8\u91ca\u8005\u53ef\u80fd\u4e0d\u5f97\u4e0d\u4f9d\u8d56\u4e2a\u4eba\u504f\u89c1\uff0c\u4ece\u800c\u5bfc\u81f4\u8d85\u51fa\u9884\u671f\u7684\u6a21\u578b\u884c\u4e3a\uff0c\u5982\u53d8\u5f97\u8fc7\u4e8e\u8c28\u614e\uff0c\u6216\u4ee5\u4e0d\u7406\u60f3\u7684\u98ce\u683c\uff08\u5982\u8bc4\u5224\uff09\u505a\u51fa\u54cd\u5e94\u3002<\/p>\n<p>\u4f8b\u5982\uff0c\u5728 OpenAI \u7684\u4e00\u6b21\u5b9e\u9a8c\u4e2d\uff0c\u4e00\u4e9b\u6ce8\u91ca\u8005\u5728\u5bf9\u7528\u6237\u6709\u5173\u81ea\u6b8b\u8bf7\u6c42\u7684\u53ef\u80fd\u56de\u590d\u8fdb\u884c\u6392\u5e8f\u65f6\uff0c\u504f\u5411\u4e8e\u5c06\u7528\u6237\u8f6c\u5230\u7f8e\u56fd\u81ea\u6740\u70ed\u7ebf\uff0c\u800c\u8fd9\u5bf9\u7f8e\u56fd\u4ee5\u5916\u7684\u7528\u6237\u6ca1\u6709\u5e2e\u52a9\u3002\u8981\u89e3\u51b3\u8fd9\u4e9b\u95ee\u9898\uff0c\u5f80\u5f80\u9700\u8981\u91cd\u65b0\u6807\u6ce8\u6216\u6536\u96c6\u65b0\u6570\u636e\uff0c\u8fd9\u65e2\u6602\u8d35\u53c8\u8017\u65f6\u3002<\/p>\n<p>\u4e3a\u4e86\u89e3\u51b3\u8fd9\u4e9b\u95ee\u9898\uff0c\u4f7f\u7528 AI \u53cd\u9988\u7684\u65b9\u6cd5\u6700\u8fd1\u8d8a\u6765\u8d8a\u53d7\u6b22\u8fce\uff0c\u5176\u4e2d\u6700\u7a81\u51fa\u7684\u662f\u5baa\u6cd5 AI\uff08Constitutional AI\uff09\u3002\u8fd9\u4e9b\u65b9\u6cd5\u5229\u7528 AI \u53cd\u9988\u5408\u6210\u8bad\u7ec3\u6570\u636e\uff0c\u4e0e\u4eba\u7c7b\u6570\u636e\u76f8\u7ed3\u5408\uff0c\u7528\u4e8e\u76d1\u7763\u5fae\u8c03\uff08SFT\uff09\u548c\u5956\u52b1\u6a21\u578b\uff08RM\uff09\u8bad\u7ec3\u6b65\u9aa4\u3002\u4e0d\u8fc7\uff0c\u5728\u5baa\u6cd5 AI \u548c\u5176\u4ed6\u65b9\u6cd5\u4e2d\uff0c\u300c\u5baa\u6cd5\u300d\u6d89\u53ca\u300c\u9009\u62e9\u5371\u5bb3\u8f83\u5c0f\u7684\u54cd\u5e94\u300d\u7b49\u4e00\u822c\u6027\u6307\u5bfc\u539f\u5219\uff0cAI \u6a21\u578b\u6709\u5f88\u5927\u7684\u81ea\u7531\u88c1\u91cf\u6743\u6765\u51b3\u5b9a\u4ec0\u4e48\u662f\u6709\u5bb3\u7684\u3002\u5728\u73b0\u5b9e\u4e16\u754c\u7684\u90e8\u7f72\u4e2d\uff0c\u6211\u4eec\u9700\u8981\u6267\u884c\u66f4\u8be6\u7ec6\u7684\u653f\u7b56\uff0c\u89c4\u5b9a\u5e94\u8be5\u62d2\u7edd\u54ea\u4e9b\u63d0\u793a\uff0c\u4ee5\u53ca\u62d2\u7edd\u7684\u65b9\u5f0f\u662f\u4ec0\u4e48\u3002<\/p>\n<p>\u56e0\u6b64\uff0c\u5728\u8fd9\u7bc7\u8bba\u6587\u4e2d\uff0cOpenAI \u7684\u7814\u7a76\u8005\u63d0\u51fa\u4e86\u4e00\u79cd\u65b0\u7684 AI \u53cd\u9988\u65b9\u6cd5 \u2014\u2014RBR\uff0c\u5b83\u5141\u8bb8\u4eba\u7c7b\u8be6\u7ec6\u8bf4\u660e\u6240\u9700\u7684\u6a21\u578b\u54cd\u5e94\uff0c\u7c7b\u4f3c\u4e8e\u7ed9\u4eba\u7c7b\u6ce8\u91ca\u8005\u7684\u6307\u793a\u3002<\/p>\n<p><strong>RBR \u7684\u5de5\u4f5c\u539f\u7406\u662f\u600e\u6837\u7684\uff1f<\/strong><\/p>\n<p>\u5b9e\u65bd RBR \u7684\u65b9\u6cd5\u5305\u62ec\u5b9a\u4e49\u4e00\u7ec4\u547d\u9898 \u2014\u2014 \u5173\u4e8e\u6a21\u578b\u54cd\u5e94\u4e2d\u671f\u671b\u6216\u4e0d\u671f\u671b\u65b9\u9762\u7684\u7b80\u5355\u9648\u8ff0\uff0c\u4f8b\u5982\u300c\u5e26\u6709\u8bc4\u5224\u6027\u300d\u3001\u300c\u5305\u542b\u4e0d\u5141\u8bb8\u7684\u5185\u5bb9\u300d\u3001\u300c\u63d0\u53ca\u5b89\u5168\u653f\u7b56\u300d\u3001\u300c\u514d\u8d23\u58f0\u660e\u300d\u7b49\u3002\u7136\u540e\uff0c\u8fd9\u4e9b\u547d\u9898\u88ab\u7528\u6765\u5f62\u6210\u89c4\u5219\uff0c\u8fd9\u4e9b\u89c4\u5219\u88ab\u7cbe\u5fc3\u8bbe\u8ba1\u4ee5\u6355\u6349\u5728\u5404\u79cd\u573a\u666f\u4e2d\u5b89\u5168\u548c\u9002\u5f53\u54cd\u5e94\u7684\u7ec6\u5fae\u5dee\u522b\u3002<\/p>\n<p>\u4f8b\u5982\uff0c\u5728\u9762\u5bf9\u4e0d\u5b89\u5168\u8bf7\u6c42\u65f6\uff0c\u62d2\u7edd\uff08\u5982\u300c\u62b1\u6b49\uff0c\u6211\u65e0\u6cd5\u5e2e\u4f60\u300d\uff09\u662f\u4e00\u79cd\u671f\u671b\u7684\u6a21\u578b\u54cd\u5e94\u3002\u76f8\u5173\u89c4\u5219\u5c06\u89c4\u5b9a\uff0c\u62d2\u7edd\u5e94\u300c\u5305\u542b\u7b80\u77ed\u7684\u9053\u6b49\u300d\u5e76\u4e14\u300c\u5e94\u8bf4\u660e\u65e0\u6cd5\u9075\u4ece\u300d\u3002<\/p>\n<p>\u7814\u7a76\u56e2\u961f\u8bbe\u8ba1\u4e86\u4e09\u7c7b\u671f\u671b\u7684\u6a21\u578b\u884c\u4e3a\uff0c\u7528\u4e8e\u5904\u7406\u6709\u5bb3\u6216\u654f\u611f\u7684\u8bdd\u9898\u3002\u6839\u636e\u5b89\u5168\u653f\u7b56\uff0c\u4e0d\u540c\u7684\u8bf7\u6c42\u5bf9\u5e94\u4e0d\u540c\u7684\u6a21\u578b\u54cd\u5e94\u7c7b\u578b\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-08bdadc4cb6f4a691f1432d20f31cf47.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-08bdadc4cb6f4a691f1432d20f31cf47.png\"><\/a><\/p>\n<p><em><sup>\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u56fe\u4e2d\u5185\u5bb9\u7531 AI \u5de5\u5177\u7ffb\u8bd1\uff0c\u4ec5\u4f9b\u53c2\u8003\u3002<\/sup><\/em><\/p>\n<p>\u4ee5\u4e0b\u662f\u4e00\u4e9b\u547d\u9898\u7684\u7b80\u5316\u793a\u4f8b\uff0c\u4ee5\u53ca\u5b83\u4eec\u5982\u4f55<mark data-type=\"concepts\" data-id=\"8ec6a68f-ad96-4b85-ab72-6f8931886922\">\u6620\u5c04<\/mark>\u7406\u60f3\u884c\u4e3a\u6216\u975e\u7406\u60f3\u884c\u4e3a\u5230\u4e0d\u540c\u54cd\u5e94\u7c7b\u578b\u7684\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-2744adbd9dc371c9b8639cddaaf91776.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-2744adbd9dc371c9b8639cddaaf91776.png\"><\/a><\/p>\n<p><em><sup>\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u56fe\u4e2d\u5185\u5bb9\u7531 AI \u5de5\u5177\u7ffb\u8bd1\uff0c\u4ec5\u4f9b\u53c2\u8003\u3002<\/sup><\/em><\/p>\n<p>\u7814\u7a76\u8005\u5728\u4e0b\u8868\u4e2d\u63d0\u4f9b\u4e86\u4e00\u4e9b\u5728\u5b9e\u9a8c\u4e2d\u8bad\u7ec3\u6a21\u578b\u6240\u5b8c\u6210\u7684\u793a\u4f8b\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-d8877e649cfc095d414782c4572d3316.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-d8877e649cfc095d414782c4572d3316.png\"><\/a><\/p>\n<p><em><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-edb3ce86cc56eac12dde2ac389310589.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-edb3ce86cc56eac12dde2ac389310589.png\"><\/a><\/em><\/p>\n<p><em><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-718d29f5a7e35c4a252ab06dcd547106.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-718d29f5a7e35c4a252ab06dcd547106.png\"><\/a><\/em><\/p>\n<p><em><sup>\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u56fe\u4e2d\u5185\u5bb9\u7531 AI \u5de5\u5177\u7ffb\u8bd1\uff0c\u4ec5\u4f9b\u53c2\u8003\u3002<\/sup><\/em><\/p>\n<p>\u8bc4\u4f30\u5668\u662f\u4e00\u4e2a\u56fa\u5b9a\u7684<mark data-type=\"tech_tasks\" data-id=\"bf35ef94-d956-4033-a533-0c0828308c36\">\u8bed\u8a00\u6a21\u578b<\/mark>\uff0c\u6839\u636e\u54cd\u5e94\u9075\u5faa\u89c4\u5219\u7684\u7a0b\u5ea6\u5bf9\u5176\u8fdb\u884c\u8bc4\u5206\uff0c\u4ece\u800c\u4f7f RBR \u65b9\u6cd5\u80fd\u591f\u7075\u6d3b\u9002\u5e94\u65b0\u89c4\u5219\u548c\u5b89\u5168\u653f\u7b56\u3002<\/p>\n<p>RBR \u4f7f\u7528\u8fd9\u4e9b\u8bc4\u5206\u6765\u62df\u5408\u4e00\u4e2a\u7ebf\u6027\u6a21\u578b\uff0c\u8be5\u6a21\u578b\u7684<mark data-type=\"concepts\" data-id=\"149a12cf-10c2-4555-9899-cc6dee319ef5\">\u6743\u91cd<\/mark><mark data-type=\"concepts\" data-id=\"2e982b73-88e2-41e8-a430-f7ae5a9af4bf\">\u53c2\u6570<\/mark>\u662f\u4ece\u4e00\u4e2a\u5df2\u77e5\u7406\u60f3\u54cd\u5e94\u7c7b\u578b\u7684\u5c0f\u6570\u636e\u96c6\uff0c\u4ee5\u53ca\u5bf9\u5e94\u7684\u671f\u671b\u505a\u6cd5\u548c\u4e0d\u671f\u671b\u505a\u6cd5\u4e2d\u5b66\u4e60\u7684\u3002<\/p>\n<p>\u8fd9\u4e9b RBR \u5956\u52b1\u968f\u540e\u4e0e\u6765\u81ea\u300c\u4ec5\u63d0\u4f9b\u5e2e\u52a9\u300d\u7684\u5956\u52b1\u6a21\u578b\u7684\u5956\u52b1\u7ed3\u5408\u8d77\u6765\uff0c\u4f5c\u4e3a PPO \u7b97\u6cd5\u7684\u989d\u5916\u4fe1\u53f7\uff0c\u4ee5\u9f13\u52b1\u6a21\u578b\u9075\u5faa\u5b89\u5168\u884c\u4e3a\u7b56\u7565\u3002<\/p>\n<p>\u8be5\u65b9\u6cd5\u5141\u8bb8\u7814\u7a76\u8005\u5bf9\u6a21\u578b\u7684\u884c\u4e3a\u8fdb\u884c\u7cbe\u7ec6\u63a7\u5236\uff0c\u786e\u4fdd\u5176\u4e0d\u4ec5\u907f\u514d\u6709\u5bb3\u5185\u5bb9\uff0c\u800c\u4e14\u4ee5\u4e00\u79cd\u65e2\u8868\u793a\u5c0a\u91cd\u53c8\u6709\u5e2e\u52a9\u7684\u65b9\u5f0f\u8fdb\u884c\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-585f2f3c33306b82742e75a13dcfdbf9.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-585f2f3c33306b82742e75a13dcfdbf9.png\"><\/a><\/p>\n<p><em><sup>\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0<mark data-type=\"tech_methods\" data-id=\"ee1a8f69-3170-4ddf-b2b6-47d91c844425\">\u5f3a\u5316\u5b66\u4e60<\/mark>\u8fc7\u7a0b\u4e2d RBR \u4e0e\u4f20\u7edf\u5956\u52b1\u6a21\u578b\u7684\u96c6\u6210\u3002<\/sup><\/em><\/p>\n<p><strong>RBR \u597d\u7528\u5417\uff1f\u00a0<\/strong><\/p>\n<p>\u5b9e\u9a8c\u663e\u793a\uff0c\u7ecf\u8fc7 RBR \u8bad\u7ec3\u7684\u6a21\u578b\u8868\u73b0\u51fa\u4e0e\u7ecf\u8fc7\u4eba\u7c7b\u53cd\u9988\u8bad\u7ec3\u7684\u6a21\u578b\u76f8\u5f53\u7684\u5b89\u5168\u6027\u80fd\u3002\u524d\u8005\u8fd8\u51cf\u5c11\u4e86\u9519\u8bef\u5730\u62d2\u7edd\u5b89\u5168\u8bf7\u6c42\uff08\u5373\u8fc7\u5ea6\u62d2\u7edd\uff09\u7684\u60c5\u51b5\u3002<\/p>\n<p>\u6b64\u5916\uff0cRBR \u8fd8\u663e\u8457\u51cf\u5c11\u4e86\u5bf9\u5927\u91cf\u4eba\u5de5\u6570\u636e\u7684\u9700\u6c42\uff0c\u4f7f\u8bad\u7ec3\u8fc7\u7a0b\u66f4\u5feb\u3001\u66f4\u5177\u6210\u672c\u6548\u76ca\u3002<\/p>\n<p>\u968f\u7740\u6a21\u578b\u80fd\u529b\u548c\u5b89\u5168\u51c6\u5219\u7684\u53d1\u5c55\uff0cRBR \u53ef\u4ee5\u901a\u8fc7\u4fee\u6539\u6216\u6dfb\u52a0\u65b0\u89c4\u5219\u5feb\u901f\u66f4\u65b0\uff0c\u800c\u65e0\u9700\u8fdb\u884c\u5927\u91cf\u91cd\u65b0\u8bad\u7ec3\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-bbe130a8115a4919cc3843f7016cd901.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-bbe130a8115a4919cc3843f7016cd901.png\"><\/a><\/p>\n<p><em><sup>\u8be5\u56fe\u663e\u793a\u4e86\u6709\u7528\u6027\uff08\u4ee5\u6a21\u578b\u6b63\u786e\u9075\u5faa\u5b89\u5168\u63d0\u793a\u7684\u767e\u5206\u6bd4\u6765\u8861\u91cf\uff09\u4e0e\u5b89\u5168\u6027\uff08\u4ee5\u6a21\u578b\u6b63\u786e\u62d2\u7edd\u4e0d\u5b89\u5168\u63d0\u793a\u7684\u767e\u5206\u6bd4\u6765\u8861\u91cf\uff09\u4e4b\u95f4\u7684\u6743\u8861\u3002\u5bf9\u4e8e\u8fd9\u4e24\u4e2a\u6307\u6807\uff0c\u503c\u8d8a\u9ad8\u8d8a\u597d\u3002\u53f3\u4e0a\u89d2\u6807\u8bb0\u4e86\u6709\u7528\u6027\u548c\u5b89\u5168\u6027\u4e4b\u95f4\u7684\u5b8c\u7f8e\u5e73\u8861\u3002\u6709\u7528\u6027\u57fa\u7ebf\u4e0d\u4f7f\u7528\u5b89\u5168\u6027 RBR\uff0c\u5f80\u5f80\u66f4\u6709\u7528\u4f46\u5b89\u5168\u6027\u8f83\u4f4e\u3002\u4eba\u7c7b\u57fa\u7ebf\u662f\u5728\u4ec5\u6709\u5e2e\u52a9\u548c\u4eba\u5de5\u6ce8\u91ca\u7684\u5b89\u5168\u6027\u6570\u636e\u4e0a\u8fdb\u884c\u8bad\u7ec3\u7684\uff0c\u5f80\u5f80\u975e\u5e38\u5b89\u5168\u4f46\u6709\u7528\u6027\u8f83\u4f4e\u3002\u501f\u52a9 RBR\uff0cOpenAI \u7684\u76ee\u6807\u662f\u4f7f\u6a21\u578b\u65e2\u5b89\u5168\u53c8\u6709\u7528\u3002<\/sup><\/em><\/p>\n<p><strong>RBR \u6709\u54ea\u4e9b\u5c40\u9650\uff1f\u00a0<\/strong><\/p>\n<p>\u5c3d\u7ba1\u89c4\u5219\u57fa\u7840\u7684\u7cfb\u7edf\uff08RBR\uff09\u5728\u6709\u660e\u786e\u3001\u76f4\u89c2\u89c4\u5219\u7684\u4efb\u52a1\u4e2d\u8868\u73b0\u826f\u597d\uff0c\u4f46\u5728\u66f4\u4e3b\u89c2\u7684\u4efb\u52a1\u4e2d\uff08\u5982\u64b0\u5199\u9ad8\u8d28\u91cf\u7684\u6587\u7ae0\uff09\uff0c\u5e94\u7528 RBR \u53ef\u80fd\u4f1a\u6709\u4e9b\u68d8\u624b\u3002\u7136\u800c\uff0cRBR \u53ef\u4ee5\u4e0e\u4eba\u7c7b\u53cd\u9988\u7ed3\u5408\u8d77\u6765\uff0c\u4ee5\u5e73\u8861\u8fd9\u4e9b\u6311\u6218\u3002\u4f8b\u5982\uff0cRBR \u53ef\u4ee5\u5f3a\u5236\u6267\u884c\u7279\u5b9a\u7684\u51c6\u5219\uff08\u5982\u300c\u4e0d\u8981\u4f7f\u7528\u4fda\u8bed\u300d\u6216\u6a21\u578b\u89c4\u8303\u4e2d\u7684\u89c4\u5219\uff09\uff0c\u800c\u4eba\u7c7b\u53cd\u9988\u53ef\u4ee5\u5e2e\u52a9\u5904\u7406\u66f4\u7ec6\u5fae\u7684\u65b9\u9762\uff08\u5982\u6574\u4f53\u8fde\u8d2f\u6027\uff09\u3002RBR \u7684\u5f3a\u5ea6\u88ab\u4f18\u5316\u4e3a\u65e2\u80fd\u6b63\u786e\u6267\u884c\u5b89\u5168\u504f\u597d\uff0c\u53c8\u4e0d\u4f1a\u8fc7\u5ea6\u5f71\u54cd\u6700\u7ec8\u7684\u5956\u52b1\u8bc4\u5206 \u2014\u2014 \u8fd9\u6837\uff0cRLHF \u5956\u52b1\u6a21\u578b\u4ecd\u7136\u53ef\u4ee5\u5728\u5982\u5199\u4f5c\u98ce\u683c\u7b49\u65b9\u9762\u63d0\u4f9b\u5f3a\u6709\u529b\u7684\u4fe1\u53f7\u3002\u00a0<\/p>\n<p>\u4f26\u7406\u8003\u91cf\uff1a\u5c06\u5b89\u5168\u68c0\u67e5\u4ece\u4eba\u7c7b\u8f6c\u79fb\u5230 AI \u4e0a\u53ef\u80fd\u4f1a\u51cf\u5c11\u5bf9 AI \u5b89\u5168\u7684\u4eba\u5de5\u76d1\u7763\uff0c\u5e76\u4e14\u5982\u679c\u4f7f\u7528\u6709\u504f\u89c1\u7684\u6a21\u578b\u63d0\u4f9b RBR \u5956\u52b1\uff0c\u8fd8\u53ef\u80fd\u653e\u5927\u6f5c\u5728\u7684\u504f\u89c1\u3002\u4e3a\u4e86\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\uff0c\u7814\u7a76\u4eba\u5458\u5e94\u8be5\u4ed4\u7ec6\u8bbe\u8ba1 RBR\uff0c\u4ee5\u786e\u4fdd\u5176\u516c\u5e73\u548c\u51c6\u786e\uff0c\u5e76\u8003\u8651\u7ed3\u5408\u4f7f\u7528 RBR \u548c\u4eba\u7c7b\u53cd\u9988\uff0c\u4ee5\u6700\u5927\u9650\u5ea6\u5730\u51cf\u5c11\u98ce\u9669\u3002\u00a0<\/p>\n<p>OpenAI \u8868\u793a\uff0cRBR \u4e0d\u4ec5\u9650\u4e8e\u5b89\u5168\u8bad\u7ec3\uff0c\u5b83\u4eec\u53ef\u4ee5\u9002\u5e94\u5404\u79cd\u4efb\u52a1\uff0c\u5176\u4e2d\u660e\u786e\u7684\u89c4\u5219\u53ef\u4ee5\u5b9a\u4e49\u6240\u9700\u7684\u884c\u4e3a\uff0c\u4f8b\u5982\u4e3a\u7279\u5b9a\u5e94\u7528\u7a0b\u5e8f\u5b9a\u5236\u6a21\u578b\u54cd\u5e94\u7684\u4e2a\u6027\u6216\u683c\u5f0f\u3002\u4e0b\u4e00\u6b65\uff0cOpenAI \u8fd8\u8ba1\u5212\u8fdb\u884c\u66f4\u5e7f\u6cdb\u7684\u6d88\u878d\u7814\u7a76\uff0c\u4ee5\u66f4\u5168\u9762\u5730\u4e86\u89e3\u4e0d\u540c\u7684 RBR \u7ec4\u4ef6\u3001\u4f7f\u7528\u5408\u6210\u6570\u636e\u8fdb\u884c\u89c4\u5219\u5f00\u53d1\u4ee5\u53ca\u4eba\u5de5\u8bc4\u4f30\uff0c\u4ee5\u9a8c\u8bc1 RBR \u5728\u5305\u62ec\u5b89\u5168\u4ee5\u5916\u7684\u5176\u4ed6\u9886\u57df\u7684\u5404\u79cd\u5e94\u7528\u4e2d\u7684\u6709\u6548\u6027\u3002<\/p>\n<p><em><sup>\u53c2\u8003\u5185\u5bb9\uff1a<\/sup><\/em><\/p>\n<p><em><sup>https:\/\/openai.com\/index\/improving-model-safety-behavior-with-rule-based-rewards\/<\/sup><\/em><\/p>\n<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-07-25-3\" target=\"_blank\">RLHF\u4e0d\u591f\u7528\u4e86\uff0cOpenAI\u8bbe\u8ba1\u51fa\u4e86\u65b0\u7684\u5956\u52b1\u673a\u5236<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:RLHF\u4e0d\u591f\u7528\u4e86\uff0cOp [&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-34738","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>RLHF\u4e0d\u591f\u7528\u4e86\uff0cOpenAI\u8bbe\u8ba1\u51fa\u4e86\u65b0\u7684\u5956\u52b1\u673a\u5236 - \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=34738\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34738#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34738\"},\"author\":{\"name\":\"AI\u5c0f\u52a9\u624b\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\"},\"headline\":\"RLHF\u4e0d\u591f\u7528\u4e86\uff0cOpenAI\u8bbe\u8ba1\u51fa\u4e86\u65b0\u7684\u5956\u52b1\u673a\u5236\",\"datePublished\":\"2024-12-11T17:57:53+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34738\"},\"wordCount\":103,\"commentCount\":0,\"image\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34738#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/07\\\/frc-3833915be2182c19caea6e98194f8efb.png\",\"keywords\":[\"\u5927\u6a21\u578b\"],\"articleSection\":[\"\u884c\u4e1a\u8d44\u8baf\"],\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/17aitech.com\\\/?p=34738#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34738\",\"url\":\"https:\\\/\\\/17aitech.com\\\/?p=34738\",\"name\":\"RLHF\u4e0d\u591f\u7528\u4e86\uff0cOpenAI\u8bbe\u8ba1\u51fa\u4e86\u65b0\u7684\u5956\u52b1\u673a\u5236 - \u4e00\u8d77AI\u6280\u672f\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34738#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34738#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/07\\\/frc-3833915be2182c19caea6e98194f8efb.png\",\"datePublished\":\"2024-12-11T17:57:53+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\"},\"breadcrumb\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34738#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/17aitech.com\\\/?p=34738\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34738#primaryimage\",\"url\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/07\\\/frc-3833915be2182c19caea6e98194f8efb.png\",\"contentUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/07\\\/frc-3833915be2182c19caea6e98194f8efb.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34738#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\\\/\\\/17aitech.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"RLHF\u4e0d\u591f\u7528\u4e86\uff0cOpenAI\u8bbe\u8ba1\u51fa\u4e86\u65b0\u7684\u5956\u52b1\u673a\u5236\"}]},{\"@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":"RLHF\u4e0d\u591f\u7528\u4e86\uff0cOpenAI\u8bbe\u8ba1\u51fa\u4e86\u65b0\u7684\u5956\u52b1\u673a\u5236 - \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=34738","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/17aitech.com\/?p=34738#article","isPartOf":{"@id":"https:\/\/17aitech.com\/?p=34738"},"author":{"name":"AI\u5c0f\u52a9\u624b","@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"headline":"RLHF\u4e0d\u591f\u7528\u4e86\uff0cOpenAI\u8bbe\u8ba1\u51fa\u4e86\u65b0\u7684\u5956\u52b1\u673a\u5236","datePublished":"2024-12-11T17:57:53+00:00","mainEntityOfPage":{"@id":"https:\/\/17aitech.com\/?p=34738"},"wordCount":103,"commentCount":0,"image":{"@id":"https:\/\/17aitech.com\/?p=34738#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-3833915be2182c19caea6e98194f8efb.png","keywords":["\u5927\u6a21\u578b"],"articleSection":["\u884c\u4e1a\u8d44\u8baf"],"inLanguage":"zh-Hans","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/17aitech.com\/?p=34738#respond"]}]},{"@type":"WebPage","@id":"https:\/\/17aitech.com\/?p=34738","url":"https:\/\/17aitech.com\/?p=34738","name":"RLHF\u4e0d\u591f\u7528\u4e86\uff0cOpenAI\u8bbe\u8ba1\u51fa\u4e86\u65b0\u7684\u5956\u52b1\u673a\u5236 - \u4e00\u8d77AI\u6280\u672f","isPartOf":{"@id":"https:\/\/17aitech.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/17aitech.com\/?p=34738#primaryimage"},"image":{"@id":"https:\/\/17aitech.com\/?p=34738#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-3833915be2182c19caea6e98194f8efb.png","datePublished":"2024-12-11T17:57:53+00:00","author":{"@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"breadcrumb":{"@id":"https:\/\/17aitech.com\/?p=34738#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/17aitech.com\/?p=34738"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/?p=34738#primaryimage","url":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-3833915be2182c19caea6e98194f8efb.png","contentUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-3833915be2182c19caea6e98194f8efb.png"},{"@type":"BreadcrumbList","@id":"https:\/\/17aitech.com\/?p=34738#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/17aitech.com\/"},{"@type":"ListItem","position":2,"name":"RLHF\u4e0d\u591f\u7528\u4e86\uff0cOpenAI\u8bbe\u8ba1\u51fa\u4e86\u65b0\u7684\u5956\u52b1\u673a\u5236"}]},{"@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\/34738","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=34738"}],"version-history":[{"count":0,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/34738\/revisions"}],"wp:attachment":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=34738"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=34738"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=34738"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}