{"id":39599,"date":"2025-03-23T22:02:58","date_gmt":"2025-03-23T14:02:58","guid":{"rendered":"https:\/\/17aitech.com\/?p=39599"},"modified":"2025-03-23T22:02:58","modified_gmt":"2025-03-23T14:02:58","slug":"%e7%94%b0%e6%b8%8a%e6%a0%8b%e5%92%8csergey-levine%e5%8f%82%e4%b8%8e%e5%bc%80%e5%8f%91%e6%96%b0%e5%9e%8brl%e7%ae%97%e6%b3%95%ef%bc%8c%e8%83%bd%e9%80%9a%e8%bf%87%e5%a4%9a%e8%bd%ae%e8%ae%ad%e7%bb%83","status":"publish","type":"post","link":"https:\/\/17aitech.com\/?p=39599","title":{"rendered":"\u7530\u6e0a\u680b\u548cSergey Levine\u53c2\u4e0e\u5f00\u53d1\u65b0\u578bRL\u7b97\u6cd5\uff0c\u80fd\u901a\u8fc7\u591a\u8f6e\u8bad\u7ec3\u8ba9\u667a\u80fd\u4f53\u5b66\u4f1a\u534f\u4f5c\u63a8\u7406"},"content":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2025-03-23-3\" target=\"_blank\">\u7530\u6e0a\u680b\u548cSergey Levine\u53c2\u4e0e\u5f00\u53d1\u65b0\u578bRL\u7b97\u6cd5\uff0c\u80fd\u901a\u8fc7\u591a\u8f6e\u8bad\u7ec3\u8ba9\u667a\u80fd\u4f53\u5b66\u4f1a\u534f\u4f5c\u63a8\u7406<\/a><\/p>\n<blockquote data-author-name=\"\" data-content-utf8-length=\"31\" data-source-title=\"\" data-type=\"2\" data-url=\"\">\n<section>\n<section>\u5f3a\u5316\u5b66\u4e60\u63d0\u5347\u4e86 LLM \u5404\u65b9\u9762\u7684\u80fd\u529b\uff0c\u800c\u5f3a\u5316\u5b66\u4e60\u672c\u8eab\u4e5f\u5728\u8fdb\u5316\u3002<\/section>\n<\/section>\n<\/blockquote>\n<section><\/section>\n<section>\u73b0\u5b9e\u4e16\u754c\u4e2d\uff0c\u5f88\u591a\u4efb\u52a1\u5f88\u590d\u6742\uff0c\u9700\u8981\u6267\u884c\u4e00\u7cfb\u5217\u7684\u51b3\u7b56\u3002\u800c\u8981\u8ba9\u667a\u80fd\u4f53\u5728\u8fd9\u4e9b\u4efb\u52a1\u4e0a\u5b9e\u73b0\u6700\u4f73\u6027\u80fd\uff0c\u901a\u5e38\u9700\u8981\u76f4\u63a5\u5728\u591a\u8f6e\u76f8\u5173\u76ee\u6807\uff08\u6bd4\u5982\u6210\u529f\u7387\uff09\u4e0a\u6267\u884c\u4f18\u5316\u3002\u4e0d\u8fc7\uff0c\u76f8\u6bd4\u4e8e\u6a21\u4eff\u6bcf\u4e00\u8f6e\u4e2d\u6700\u53ef\u80fd\u7684\u52a8\u4f5c\uff0c\u8fd9\u79cd\u65b9\u6cd5\u7684\u96be\u5ea6\u8981\u5927\u5f97\u591a\u3002<\/section>\n<section><\/section>\n<section>\u5728\u76f4\u63a5\u4f18\u5316\u591a\u8f6e\u76ee\u6807\u65b9\u9762\uff0c\u4e00\u7c7b\u81ea\u7136\u7684\u65b9\u6cd5\u662f\u5e94\u7528\u5355\u8f6e RLHF \u7b97\u6cd5\uff0c\u4f8b\u5982 RAFT\u3001DPO \u548c PPO \uff0c\u4e0d\u8fc7\u8fd9\u4e9b\u65b9\u6cd5\u4e0d\u4f1a\u5728\u4e0d\u540c\u8f6e\u6b21\u95f4\u6267\u884c\u663e\u5f0f\u7684 credit \u5206\u914d\u3002\u56e0\u6b64\uff0c\u7531\u4e8e\u590d\u6742\u987a\u5e8f\u51b3\u7b56\u4efb\u52a1\u7684\u957f\u671f\u6027\uff0c\u5b83\u4eec\u53ef\u80fd\u4f1a\u51fa\u73b0\u9ad8\u65b9\u5dee\u548c\u8f83\u5dee\u7684\u6837\u672c\u590d\u6742\u6027\u7b49\u95ee\u9898\u3002<\/section>\n<section><\/section>\n<section>\u53e6\u4e00\u79cd\u9009\u62e9\u662f\u5e94\u7528\u4ef7\u503c\u51fd\u6570\u5b66\u4e60\u65b9\u6cd5\uff0c\u4f8b\u5982 TD \u5b66\u4e60\u3002\u7136\u800c\uff0c\u8fd9\u9700\u8981\u5728 LLM \u8868\u5f81\u7684\u57fa\u7840\u4e0a\u8bad\u7ec3\u4e00\u4e2a\u65b0\u7684\u7279\u5b9a\u4e8e\u4efb\u52a1\u7684\u4ef7\u503c\u5934\uff0c\u8fd9\u53ef\u80fd\u65e0\u6cd5\u5728\u6709\u9650\u7684\u5fae\u8c03\u6570\u636e\u4e0b\u5f88\u597d\u5730\u6cdb\u5316\u3002\u56e0\u6b64\uff0c\u76ee\u524d\u5c1a\u4e0d\u6e05\u695a\u54ea\u79cd\u591a\u8f6e RL \u7b97\u6cd5\u6700\u6709\u6548\uff0c\u80fd\u591f\u5145\u5206\u5229\u7528 LLM \u7684\u63a8\u7406\u80fd\u529b\u6765\u8bad\u7ec3\u901a\u7528\u3001\u6709\u80fd\u529b\u548c\u76ee\u6807\u5bfc\u5411\u7684\u667a\u80fd\u4f53\u3002<\/section>\n<section><\/section>\n<section>\u8fd1\u65e5\uff0cMeta FAIR \u548c\u52a0\u5229\u798f\u5c3c\u4e9a\u5927\u5b66\u4f2f\u514b\u5229\u5206\u6821\u4e00\u4e2a\u7814\u7a76\u56e2\u961f\u5728\u8fd9\u4e2a\u7814\u7a76\u8bfe\u9898\u4e0a\u53d6\u5f97\u4e86\u65b0\u7684\u7a81\u7834\u3002\u9996\u5148\uff0c\u4ed6\u4eec\u4e3a\u8be5\u95ee\u9898\u6784\u5efa\u4e86\u4e00\u4e2a\u65b0\u7684\u57fa\u51c6\uff1a<strong>ColBench<\/strong>\uff08Collaborative Agent Benchmark\uff09\u3002\u5728\u6b64\u57fa\u7840\u4e0a\uff0c\u4ed6\u4eec\u8fd8\u63d0\u51fa\u4e86\u4e00\u79cd\u6613\u4e8e\u5b9e\u73b0\u4f46\u975e\u5e38\u6709\u6548\u7684 RL \u7b97\u6cd5\uff1a<strong>SWEET-RL<\/strong>\uff08RL with Step-WisE Evaluation from Training-Time Information\uff09\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-3e6b620684d525a6b7702cdf2a598d43.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-3e6b620684d525a6b7702cdf2a598d43.png\"><\/a><\/section>\n<ul>\n<li>\n<section>\u8bba\u6587\u6807\u9898\uff1aSWEET-RL: Training Multi-Turn LLM Agents on Collaborative Reasoning Tasks<\/section>\n<\/li>\n<li>\n<section>\u8bba\u6587\u5730\u5740\uff1ahttps:\/\/arxiv.org\/pdf\/2503.15478<\/section>\n<\/li>\n<li>\n<section>\u4ee3\u7801\u5730\u5740\uff1ahttps:\/\/github.com\/facebookresearch\/sweet_rl<\/section>\n<\/li>\n<\/ul>\n<section><\/section>\n<section>\u8fd9\u7bc7\u8bba\u6587\u7684\u4e00\u4f5c\u4e3a\u4f2f\u514b\u5229 AI \u7814\u7a76\u6240\uff08BAIR\uff09\u4e8c\u5e74\u7ea7\u535a\u58eb\u751f\u5468\u9038\u98de\uff08Yifei Zhou\uff09\u3002\u5e76\u6709\u591a\u4f4d\u8457\u540d AI \u7814\u7a76\u8005\u53c2\u4e0e\u5176\u4e2d\uff0c\u5305\u62ec\u7530\u6e0a\u680b\u3001Jason Weston \u548c Sergey Levine\u3002<\/section>\n<section><\/section>\n<section>\u4e0b\u9762\u6211\u4eec\u5c31\u6765\u7b80\u5355\u89e3\u8bfb\u4e00\u4e0b\u8fd9\u9879\u7814\u7a76\uff0c\u66f4\u591a\u8be6\u60c5\u8bf7\u53c2\u9605\u539f\u8bba\u6587\u3002<\/section>\n<section><\/section>\n<section><strong>ColBench<\/strong><\/section>\n<section><\/section>\n<section>\u5148\u6765\u770b\u770b\u4ed6\u4eec\u63d0\u51fa\u7684\u65b0\u57fa\u51c6\u3002\u9996\u5148\uff0c\u4e3a\u4e86\u89e3\u51b3\u4e3a LLM \u667a\u80fd\u4f53\u5f00\u53d1\u591a\u8f6e RL \u7b97\u6cd5\u7684\u6311\u6218\uff0c\u8be5\u56e2\u961f\u6784\u5efa\u4e86\u4e00\u4e9b\u57fa\u672c\u8bbe\u8ba1\u539f\u5219\uff0c\u5305\u62ec\uff1a<\/section>\n<section><\/section>\n<ul>\n<li>\n<section>\u5e94\u5177\u6709\u8db3\u591f\u7684\u4efb\u52a1\u590d\u6742\u6027\uff0c\u53ef\u4ee5\u6311\u6218\u667a\u80fd\u4f53\u7684\u63a8\u7406\u548c\u6cdb\u5316\u80fd\u529b\u3002<\/section>\n<\/li>\n<li>\n<section>\u5c3d\u53ef\u80fd\u5730\u964d\u4f4e\u5f00\u9500\uff0c\u4ee5\u652f\u6301\u5feb\u901f\u7814\u7a76\u539f\u578b\u8bbe\u8ba1\u3002<\/section>\n<\/li>\n<li>\n<section>\u5e94\u8be5\u6709\u8db3\u591f\u7684\u4efb\u52a1\u591a\u6837\u6027\uff0c\u4ee5\u4fbf\u5728 RL \u8bad\u7ec3\u65f6\u4e0d\u4f1a\u8fc7\u62df\u5408\u3002<\/section>\n<\/li>\n<\/ul>\n<section><\/section>\n<section>\u63a5\u4e0b\u6765\uff0c\u5177\u4f53\u770b\u770b\u8be5\u57fa\u51c6\u4e2d\u7684\u4e24\u4e2a\u4efb\u52a1\uff1a\u540e\u7aef\u7f16\u7a0b\u548c\u524d\u7aef\u8bbe\u8ba1\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-eec30bcec7ff448325308b6dadd6a0de.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-eec30bcec7ff448325308b6dadd6a0de.png\"><\/a><\/section>\n<section><strong>\u540e\u7aef\u7f16\u7a0b\u534f\u4f5c<\/strong><\/section>\n<section><\/section>\n<section>\u5728\u6b64\u4efb\u52a1\u4e2d\uff0c\u667a\u80fd\u4f53\u9700\u8981\u4e0e\u4eba\u7c7b\u6a21\u62df\u5668\u534f\u4f5c\u7f16\u5199\u81ea\u5b9a\u4e49 Python \u51fd\u6570\uff08\u6700\u591a 50 \u884c\uff09\u3002<\/section>\n<section><\/section>\n<section>\u5728\u534f\u4f5c\u5f00\u59cb\u65f6\uff0c\u667a\u80fd\u4f53\u4f1a\u5148\u83b7\u5f97\u9ad8\u7ea7\u63cf\u8ff0\u548c\u51fd\u6570\u7b7e\u540d\u3002\u4f46\u5e76\u4e0d\u4f1a\u63d0\u4f9b\u5f88\u591a\u5177\u4f53\u7684\u7ec6\u8282\uff0c\u4f8b\u5982\u5e94\u8003\u8651\u54ea\u4e9b\u6761\u4ef6\u4ee5\u53ca\u5728\u8fb9\u7f18\u60c5\u51b5\u4e0b\u8be5\u600e\u4e48\u505a\u3002\u667a\u80fd\u4f53\u5fc5\u987b\u63a8\u7406\u5e76\u51b3\u5b9a\u9700\u8981\u4eba\u7c7b\u6a21\u62df\u5668\u63d0\u4f9b\u54ea\u4e9b\u5177\u4f53\u8bf4\u660e\u3002\u4eba\u7c7b\u6a21\u62df\u5668\u9700\u8981\u6839\u636e\u53ea\u6709\u5b83\u4eec\u81ea\u5df1\u53ef\u89c1\u7684\u53c2\u8003\u4ee3\u7801\uff0c\u7528\u81ea\u7136\u8bed\u8a00\u5bf9\u6bcf\u4e2a\u9700\u8981\u8bf4\u660e\u7684\u95ee\u9898\u63d0\u4f9b\u7b80\u8981\u89e3\u91ca\uff0c\u4f46\u4e0d\u4f1a\u7f16\u5199\u4ee3\u7801\u3002<\/section>\n<section><\/section>\n<section>\u667a\u80fd\u4f53\u548c\u4eba\u7c7b\u6a21\u62df\u5668\u4e4b\u95f4\u7684\u4ea4\u4e92\u4ec5\u9650\u4e8e 10 \u8f6e\u6765\u56de\u3002\u5f53\u667a\u80fd\u4f53\u51b3\u5b9a\u7ed9\u51fa\u6700\u7ec8\u89e3\u51b3\u65b9\u6848\u6216\u8fbe\u5230\u6700\u5927\u8f6e\u6570\u65f6\uff0c\u4ea4\u4e92\u7ed3\u675f\u3002<\/section>\n<section><\/section>\n<section>\u5728\u8bc4\u4f30\u667a\u80fd\u4f53\u662f\u5426\u6210\u529f\u65f6\uff0c\u9700\u8981\u5bf9\u6bcf\u4e2a\u51fd\u6570\u8fdb\u884c 10 \u6b21\u9690\u85cf\u5355\u5143\u6d4b\u8bd5\uff0c\u5e76\u5bf9\u6bcf\u6b21\u534f\u4f5c\u7ed9\u51fa 0 \u6216 1 \u7684\u5956\u52b1\u3002<\/section>\n<section><\/section>\n<section><strong>\u524d\u7aef\u8bbe\u8ba1\u534f\u4f5c<\/strong><\/section>\n<section><\/section>\n<section>\u5728\u6b64\u4efb\u52a1\u4e2d\uff0c\u667a\u80fd\u4f53\u9700\u8981\u4e0e\u4eba\u7c7b\u6a21\u62df\u5668\u534f\u4f5c\uff0c\u901a\u8fc7\u7f16\u5199 HTML \u4ee3\u7801\u7247\u6bb5\uff08\u7ea6 100 \u884c\uff09\u6765\u8bbe\u8ba1\u7f51\u9875\u3002<\/section>\n<section><\/section>\n<section>\u5728\u534f\u4f5c\u5f00\u59cb\u65f6\uff0c\u667a\u80fd\u4f53\u4f1a\u83b7\u5f97\u7f51\u9875\u7684\u9ad8\u7ea7\u63cf\u8ff0\u3002\u540c\u6837\uff0c\u8bb8\u591a\u5177\u4f53\u7ec6\u8282\uff08\u4f8b\u5982\u7f51\u9875\u7684\u5e03\u5c40\u548c\u8c03\u8272\u677f\uff09\u90fd\u7f3a\u5931\uff0c\u53ea\u6709\u4eba\u7c7b\u6a21\u62df\u5668\u624d\u80fd\u770b\u5230\u3002\u5728\u6bcf\u4e00\u8f6e\u4e2d\uff0c\u667a\u80fd\u4f53\u90fd\u6709\u673a\u4f1a\u7f16\u5199 HTML \u7ed3\u679c\u5e76\u901a\u8fc7 Web \u6d4f\u89c8\u5668\u5448\u73b0\u51fa\u6765\u3002\u4eba\u7c7b\u6a21\u62df\u5668\u53ef\u4ee5\u5bf9\u6bd4\u6765\u81ea\u667a\u80fd\u4f53\u7684\u7f51\u9875\u548c\u53c2\u8003\u7f51\u9875\uff0c\u7136\u540e\u5411\u667a\u80fd\u4f53\u63cf\u8ff0\u5b83\u4eec\u7684\u5dee\u5f02\u3002\u4e0e\u540e\u7aef\u7f16\u7a0b\u534f\u4f5c\u7c7b\u4f3c\uff0c\u5f53\u667a\u80fd\u4f53\u51b3\u5b9a\u7ed9\u51fa\u6700\u7ec8\u89e3\u51b3\u65b9\u6848\u6216\u8fbe\u5230\u6700\u5927 10 \u8f6e\u4ea4\u4e92\u65f6\uff0c\u4ea4\u4e92\u7ed3\u675f\u3002<\/section>\n<section><\/section>\n<section>\u8bc4\u4f30\u6307\u6807\u65b9\u9762\uff0c\u4f7f\u7528\u4e86\u667a\u80fd\u4f53\u89e3\u7b54\u4e0e\u53c2\u8003\u7f51\u9875\u4e4b\u95f4\u7684 CLIP \u5d4c\u5165\u7684\u4f59\u5f26\u76f8\u4f3c\u5ea6\u3002\u540c\u6837\uff0c\u534f\u4f5c\u7ed3\u675f\u65f6\uff0c\u4f1a\u53d1\u653e 0 \u6216 1 \u7684\u5956\u52b1\u3002<\/section>\n<section><\/section>\n<section>\u8868 1 \u6bd4\u8f83\u4e86 ColBench \u4e0e\u73b0\u6709\u7684\u5176\u5b83\u57fa\u51c6\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-1569084f6c3062da16a7d32b4919572a.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-1569084f6c3062da16a7d32b4919572a.png\"><\/a><\/section>\n<section><strong>SWEET-RL<\/strong><\/section>\n<section><\/section>\n<section>SWEET-RL \u662f\u4e00\u79cd\u4e24\u9636\u6bb5\u8bad\u7ec3\u65b9\u6cd5\uff0c\u5982\u56fe 2 \u6240\u793a\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-4d78bfdae57917f72d919dda2ead4c66.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-4d78bfdae57917f72d919dda2ead4c66.png\"><\/a><\/section>\n<section><strong>\u7b2c\u4e00\u9636\u6bb5\uff1a\u5b66\u4e60\u5404\u4e2a\u8f6e\u6b21\u7684\u4f18\u52bf\u51fd\u6570<\/strong><\/section>\n<section><\/section>\n<section>\u4e3a\u4e86\u5728\u63a8\u7406\u5bc6\u96c6\u578b\u4efb\u52a1\u4e2d\u6267\u884c\u663e\u5f0f credit \u5206\u914d\uff0c\u4e4b\u524d\u4e00\u4e9b\u7814\u7a76\u4f7f\u7528\u7684\u65b9\u6cd5\u662f\uff1a\u5148\u5b66\u4e60\u4e00\u4e2a\u663e\u5f0f\u7684\u4ef7\u503c\u51fd\u6570\uff0c\u7136\u540e\u4ece\u5b66\u4e60\u5230\u7684\u4ef7\u503c\u51fd\u6570\u4e2d\u5f97\u51fa\u6bcf\u4e2a\u5355\u72ec\u52a8\u4f5c\u7684\u4f18\u52bf\u3002<\/section>\n<section><\/section>\n<section>\u7136\u800c\uff0c\u8be5\u56e2\u961f\u7684\u5b9e\u9a8c\u53d1\u73b0\uff0c\u5f53\u5fae\u8c03\u53ea\u80fd\u4f7f\u7528\u6709\u9650\u6570\u91cf\u7684\u6837\u672c\u65f6\uff0c\u8fd9\u79cd\u4ef7\u503c\u51fd\u6570\u4e0d\u80fd\u5f88\u597d\u5730\u6cdb\u5316\u3002\u4ed6\u4eec\u731c\u60f3\u8fd9\u662f\u56e0\u4e3a\u5728\u63a8\u7406\u5bc6\u96c6\u578b\u4efb\u52a1\u4e2d\u5b66\u4e60\u51c6\u786e\u7684\u4ef7\u503c\u51fd\u6570\u672c\u8eab\u5c31\u662f\u4e00\u9879\u8270\u5de8\u7684\u4efb\u52a1\uff0c\u5e76\u4e14\u4e0d\u80fd\u6709\u6548\u5730\u5229\u7528\u9884\u8bad\u7ec3 LLM \u7684\u63a8\u7406\u548c\u6cdb\u5316\u80fd\u529b\u3002<\/section>\n<section><\/section>\n<section>\u7531\u4e8e\u6267\u884c credit \u5206\u914d\u7684\u6700\u7ec8\u76ee\u6807\u662f\u5f97\u51fa\u6bcf\u4e2a\u52a8\u4f5c\u7684\u4f18\u52bf\uff0c\u8fd9\u5bf9\u4e8e LLM \u6765\u8bf4\u53ef\u80fd\u6bd4\u4f30\u8ba1\u9884\u671f\u7684\u672a\u6765\u56de\u62a5\u66f4\u5bb9\u6613\uff0c\u56e0\u6b64\u8be5\u56e2\u961f\u63d0\u51fa\u76f4\u63a5\u5b66\u4e60\u6bcf\u4e2a\u8f6e\u6b21\u52a8\u4f5c\u7684\u4f18\u52bf\u51fd\u6570\u3002<\/section>\n<section><\/section>\n<section>\u8003\u8651\u5230\u504f\u597d\u4f18\u5316\u5df2\u7ecf\u5728 LLM \u5fae\u8c03\u65b9\u9762\u5f97\u5230\u6210\u529f\u5e94\u7528\uff0c\u56e0\u6b64\u8be5\u56e2\u961f\u63d0\u51fa\u6839\u636e\u8f68\u8ff9\u7684\u504f\u597d\u5bf9\u6765\u8bad\u7ec3\u6bcf\u8f6e\u6b21\u4f18\u52bf\u51fd\u6570\u3002<\/section>\n<section><\/section>\n<section>\u7ed9\u5b9a\u540c\u4e00\u4efb\u52a1\u7684\u4e24\u6761\u8f68\u8ff9\uff0c\u5e76\u9644\u52a0\u8bad\u7ec3\u65f6\u95f4\u4fe1\u606f c\uff0c\u6839\u636e\u5b83\u4eec\u7684\u7d2f\u79ef\u5956\u52b1\u5c06\u5b83\u4eec\u6807\u8bb0\u4e3a\u9009\u53d6 \u03c4+ \u548c\u62d2\u7edd \u03c4\u2212\u3002\u8fd9\u6837\u4e00\u6765\uff0c\u4fbf\u53ef\u4ee5\u91c7\u7528 Bradley-Terry \u76ee\u6807\u8fdb\u884c\u5fae\u8c03\uff1a<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-e07cf54e31bd8ac7432a472eac7a4351.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-e07cf54e31bd8ac7432a472eac7a4351.png\"><\/a><\/section>\n<section>\u7136\u540e\uff0c\u53ef\u4ee5\u4f7f\u7528\u4f18\u52bf\u51fd\u6570\u91cd\u5199\u8fd9\u4e2a\u76ee\u6807\u51fd\u6570\uff1a<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-0d8115c9d5c8f17df85d55c43f57fdc2.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-0d8115c9d5c8f17df85d55c43f57fdc2.png\"><\/a><\/section>\n<section>\u76f4\u89c2\u5730\u8bb2\uff0c\u7c7b\u4f3c\u4e8e\u5355\u8f6e RLHF \u7684\u76ee\u6807\uff0c\u5373\u5b66\u4e60\u6bcf\u4e2a\u9009\u53d6\u54cd\u5e94\u7684\u9ad8\u5956\u52b1\u548c\u6bcf\u4e2a\u62d2\u7edd\u54cd\u5e94\u7684\u4f4e\u5956\u52b1\uff0c2 \u5f0f\u7684\u6548\u679c\u662f\u589e\u52a0\u9009\u53d6\u8f68\u8ff9\u4e2d\u6bcf\u4e2a\u52a8\u4f5c\u7684\u4f18\u52bf\u5e76\u964d\u4f4e\u62d2\u7edd\u8f68\u8ff9\u4e2d\u6bcf\u4e2a\u52a8\u4f5c\u7684\u4f18\u52bf\u3002<\/section>\n<section><\/section>\n<section>\u4e3a\u4e86\u8fdb\u4e00\u6b65\u5c06\u5b66\u4e60\u76ee\u6807\u4e0e\u4e0b\u4e00 token \u9884\u6d4b\u9884\u8bad\u7ec3\u5bf9\u9f50\uff0c\u8be5\u56e2\u961f\u7684\u505a\u6cd5\u662f\u91cd\u65b0\u5229\u7528 LLM \u7684\u73b0\u6709\u8bed\u8a00\u6a21\u578b\u5934\u6765\u53c2\u6570\u5316\u4f18\u52bf\u51fd\u6570\uff1a<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-ca08eb23a8ce58797a1bdcbe9d1ff346.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-ca08eb23a8ce58797a1bdcbe9d1ff346.png\"><\/a><\/section>\n<section><strong>\u7b2c\u4e8c\u9636\u6bb5\uff1a\u901a\u8fc7\u6bcf\u8f6e\u6d41\u7684\u4f18\u52bf\u4f18\u5316\u667a\u80fd\u4f53<\/strong><\/section>\n<section><\/section>\n<section>\u8be5\u56e2\u961f\u5f97\u5230\u7684\u4e00\u4e2a\u91cd\u8981\u89c2\u5bdf\u662f\uff1a\u867d\u7136\u6700\u7ec8\u7b56\u7565 \u03c0_\u03c6 \u4e0d\u80fd\u4ee5\u9690\u85cf\u4fe1\u606f h \u4e3a\u6761\u4ef6\uff0c\u4f46\u6b64\u7c7b\u4fe1\u606f\u5728\u8bad\u7ec3\u671f\u95f4\u662f\u53ef\u7528\u7684\u3002\u7531\u4e8e\u4f18\u52bf LLM \u03c0_\u03b8 \u53ea\u4f1a\u5728\u8bad\u7ec3\u671f\u95f4\u4f7f\u7528\uff0c\u56e0\u6b64\u5b83\u53ef\u4ee5\u5c06 c \u4f5c\u4e3a 3 \u5f0f\u7684\u8f93\u5165\u3002<\/section>\n<section><\/section>\n<section>\u76f4\u89c2\u5730\u8bb2\uff0c\u8bb8\u591a\u73b0\u5b9e\u95ee\u9898\uff08\u4f8b\u5982\u534f\u4f5c\u548c\u6570\u5b66\u63a8\u7406\uff09\u90fd\u5177\u6709\u4e00\u4e9b\u9690\u85cf\u7684\u8bad\u7ec3\u65f6\u95f4\u4fe1\u606f\uff0c\u4f8b\u5982\u53c2\u8003\u89e3\u3002\u5982\u679c\u6bcf\u8f6e\u6b21\u7684\u4f18\u52bf\u51fd\u6570\u53ef\u4ee5\u8bbf\u95ee\u6b64\u7c7b\u8bad\u7ec3\u65f6\u95f4\u4fe1\u606f\uff0c\u90a3\u4e48\u5b83\u5e94\u8be5\u80fd\u591f\u66f4\u597d\u5730\u5224\u65ad\u7b56\u7565\u91c7\u53d6\u7684\u884c\u52a8\u662f\u5426\u5728\u6b63\u786e\u7684\u8f68\u9053\u4e0a\u3002<\/section>\n<section><\/section>\n<section>\u56e0\u6b64\uff0c\u4ed6\u4eec\u4e3a\u6bcf\u8f6e\u6b21\u7684\u4f18\u52bf\u51fd\u6570\u63d0\u4f9b\u4e86\u989d\u5916\u7684\u8bad\u7ec3\u65f6\u95f4\u4fe1\u606f c\uff0c\u800c\u4ec5\u5411\u7b56\u7565\u63d0\u4f9b\u4e86\u4ea4\u4e92\u5386\u53f2 o_t\uff0c\u4ece\u800c\u4ea7\u751f\u4e86\u4e0d\u5bf9\u79f0\u7684 actor-critic \u7ed3\u6784\u3002\u539f\u5219\u4e0a\uff0cRLHF \u6587\u732e\u4e2d\u7684\u4efb\u4f55\u6210\u529f\u7b97\u6cd5\u90fd\u53ef\u7528\u4e8e\u4f18\u5316\u6bcf\u8f6e\u6b21\u7b56\u7565 \u03c0_\u03c6\uff0c\u65b9\u6cd5\u662f\u5c06\u4ea4\u4e92\u5386\u53f2\u89c6\u4e3a\u63d0\u793a\u8bcd\uff0c\u5c06\u6bcf\u8f6e\u6b21\u4f18\u52bf\u51fd\u6570 A_\u03b8 \u89c6\u4e3a\u5956\u52b1\u6a21\u578b\u3002\u5728\u8bad\u7ec3\u7b56\u7565\u7684\u8fd9\u4e2a\u9636\u6bb5\uff0c\u4e0d\u9700\u8981\u4eba\u7c7b\u5408\u4f5c\u8005\u7684\u4e92\u52a8\u3002<\/section>\n<section><\/section>\n<section>\u4e3a\u4e86\u7b80\u5355\uff0c\u8be5\u56e2\u961f\u9009\u62e9\u4f7f\u7528 DPO \u8fdb\u884c\u8bad\u7ec3\u3002\u5bf9\u4e8e\u6bcf\u4e2a\u8f6e\u6b21 t\uff0c\u9996\u5148\u4ece\u7ed9\u5b9a\u4ea4\u4e92\u5386\u53f2 o_t \u7684\u5f53\u524d\u7b56\u7565\u4e2d\u62bd\u53d6\u5019\u9009\u52a8\u4f5c\uff0c\u5e76\u6839\u636e\u5b66\u4e60\u5230\u7684\u6bcf\u8f6e\u6b21\u4f18\u52bf\u51fd\u6570\u5bf9\u5b83\u4eec\u8fdb\u884c\u6392\u5e8f\uff0c\u4ee5\u83b7\u5f97\u8981\u9009\u53d6\u548c\u62d2\u7edd\u7684\u52a8\u4f5c\u3002\u7136\u540e\uff0c\u4f7f\u7528\u6807\u51c6 DPO \u635f\u5931\u4f18\u5316\u6bcf\u4e2a\u8f6e\u6b21\u7684\u7b56\u7565\uff1a<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-84ff9ab593da4a9634981e63007c7f9a.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-84ff9ab593da4a9634981e63007c7f9a.png\"><\/a><\/section>\n<section>\u5728\u5b9e\u8df5\u4e2d\uff0c\u6bcf\u4e2a\u8f6e\u6b21\u90fd\u4f1a\u91c7\u6837 16 \u4e2a\u5019\u9009\u52a8\u4f5c\uff0c\u5e76\u4ece\u524d 50% \u5206\u4f4d\u6570\u4e2d\u968f\u673a\u9009\u62e9\u52a8\u4f5c\u4f5c\u4e3a\u9009\u53d6\u52a8\u4f5c\uff0c\u4ece\u540e 50% \u5206\u4f4d\u6570\u4e2d\u968f\u673a\u9009\u62e9\u52a8\u4f5c\u4f5c\u4e3a\u62d2\u7edd\u52a8\u4f5c\u3002<\/section>\n<section><\/section>\n<section><strong>\u5b9e\u9a8c\u8868\u73b0<\/strong><\/section>\n<section><\/section>\n<section>\u4f5c\u4e3a\u591a\u8f6e\u5f3a\u5316\u5b66\u4e60\u7b97\u6cd5\uff0cSWEET-RL \u7a76\u7adf\u80fd\u4e0d\u80fd\u6709\u6548\u5730\u8bad\u7ec3 LLM \u667a\u80fd\u4f53\u6765\u5b8c\u6210\u590d\u6742\u7684\u534f\u4f5c\u4efb\u52a1\u5462\uff1f\u4e3a\u6b64\uff0c\u8be5\u56e2\u961f\u8fdb\u884c\u4e86\u5b9e\u9a8c\u9a8c\u8bc1\u3002<\/section>\n<section><\/section>\n<section><strong>\u5728 ColBench \u4e0a\u7684\u8868\u73b0<\/strong><\/section>\n<section><\/section>\n<section>\u8868 2 \u5c55\u793a\u4e86\u5728 ColBench \u4e0a\uff0c\u4e0d\u540c LLM \u548c\u591a\u8f6e RL \u7b97\u6cd5\u7684\u6027\u80fd\u60c5\u51b5\u3002<\/section>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-749725cabe7ec738fa3fd36f9c1bc6b4.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-749725cabe7ec738fa3fd36f9c1bc6b4.png\"><\/a><\/section>\n<section>\u9996\u5148\uff0c\u6bd4\u8f83\u300c\u5355\u8f6e\u300d\u7ed3\u679c\u548c\u5176\u4ed6\u534f\u4f5c\u7ed3\u679c\uff0c\u53ef\u4ee5\u770b\u5230\u591a\u8f6e\u534f\u4f5c\u53ef\u4ee5\u901a\u8fc7\u5c06\u6700\u7ec8\u7ed3\u679c\u4e0e\u4eba\u7c7b\u6a21\u62df\u5668\u7684\u53c2\u8003\u300c\u671f\u671b\u300d\u66f4\u7d27\u5bc6\u5730\u7ed3\u5408\u8d77\u6765\uff0c\u4ece\u800c\u5927\u5927\u63d0\u9ad8 LLM \u667a\u80fd\u4f53\u5728 artifact \u521b\u5efa\u65b9\u9762\u7684\u6027\u80fd\u3002\u5982\u679c\u667a\u80fd\u4f53\u5fc5\u987b\u5728\u4e00\u4e2a\u8f6e\u6b21\u5185\u76f4\u63a5\u751f\u4ea7\u6700\u7ec8\u7ed3\u679c\uff0c\u90a3\u4e48\u5373\u4f7f\u662f\u8868\u73b0\u6700\u597d\u7684 GPT-4o \u4e5f\u53ea\u80fd\u8fbe\u5230 16.2%\u3002\u76f8\u6bd4\u4e4b\u4e0b\uff0c<strong>\u5982\u679c\u8ba9\u6240\u6709\u6a21\u578b\u6709\u673a\u4f1a\u4e0e\u4eba\u7c7b\u6a21\u62df\u5668\u8fdb\u884c\u591a\u8f6e\u4ea4\u4e92\u5e76\u6536\u96c6\u66f4\u591a\u4fe1\u606f\uff0c\u5219\u6240\u6709\u6a21\u578b\u7684\u6210\u529f\u7387\u90fd\u80fd\u500d\u589e<\/strong>\uff08\u4f8b\u5982\uff0cLlama-3.1-8B-Instruct \u7684\u6210\u529f\u7387\u4ece 6.9% \u589e\u52a0\u5230 22.4%\uff09\u3002<\/section>\n<section><\/section>\n<section>\u5c3d\u7ba1\u5982\u6b64\uff0c\u5373\u4f7f\u5bf9\u4e8e GPT-4o \u548c o1-mini \u7b49\u4e13\u6709 LLM \u6765\u8bf4\uff0c\u591a\u8f6e\u534f\u4f5c\u4ecd\u7136\u662f\u4e00\u9879\u5177\u6709\u6311\u6218\u6027\u7684\u4efb\u52a1\uff0c\u5b83\u4eec\u7684\u6210\u529f\u7387\u5206\u522b\u53ea\u80fd\u8fbe\u5230 40.4% \u548c 30.3%\u3002\u5c3d\u7ba1 o1-mini \u5728\u6570\u5b66\u548c\u7f16\u7801\u7b49\u7b26\u53f7\u63a8\u7406\u4efb\u52a1\u4e0a\u6709\u6240\u6539\u8fdb\uff0c\u4f46\u8be5\u56e2\u961f\u89c2\u5bdf\u5230\u8fd9\u4e9b\u6539\u8fdb\u5e76\u6ca1\u6709\u76f4\u63a5\u8ba9\u591a\u8f6e\u534f\u4f5c\u667a\u80fd\u4f53\u91c7\u7528\u66f4\u597d\u7684\u7b56\u7565\uff0c\u8fd9\u8868\u660e<strong>\u4e3a\u4e86\u8ba9 LLM \u4f18\u5316\u4e0e\u4eba\u7c7b\u7684\u534f\u4f5c\uff0c\u4e0b\u6e38\u5fae\u8c03\u4ecd\u7136\u662f\u5fc5\u8981\u7684\u3002<\/strong><\/section>\n<section><\/section>\n<section><strong>SWEET-RL \u4e0e\u5176\u5b83\u7b97\u6cd5\u7684\u6bd4\u8f83<\/strong><\/section>\n<section><\/section>\n<section>\u5728\u4f7f\u7528\u4e0b\u6e38\u6570\u636e\u8fdb\u884c\u5fae\u8c03\u540e\uff0c\u53ef\u4ee5\u770b\u5230\u5373\u4f7f\u662f\u6700\u7b80\u5355\u7684 RL \u7b97\u6cd5\u62d2\u7edd\u5f0f\u5fae\u8c03\uff08Rejection Fine-Tuning\uff09\u4e5f\u53ef\u4ee5\u63d0\u9ad8\u5728\u8fd9\u4e24\u9879\u4efb\u52a1\u4e0a\u7684\u6027\u80fd\u3002\u5177\u4f53\u6765\u8bf4\uff0c\u540e\u7aef\u7f16\u7a0b\u6210\u529f\u7387\u548c\u524d\u7aef\u8bbe\u8ba1\u6210\u529f\u7387\u5206\u522b\u63d0\u9ad8\u4e86 5.8% \u548c 4.8%\u3002<\/section>\n<section><\/section>\n<section>\u7136\u800c\uff0c\u8be5\u56e2\u961f\u89c2\u5bdf\u5230\uff0c\u62d2\u7edd\u5f0f\u5fae\u8c03\u5f80\u5f80\u53ea\u662f\u6559 LLM\u300c\u8bb0\u4f4f\u300d\u6bcf\u4e2a\u8bad\u7ec3\u4efb\u52a1\u7684\u89e3\u51b3\u65b9\u6848\uff0c\u800c\u6ca1\u6709\u5b66\u4e60\u4e00\u79cd\u53ef\u6cdb\u5316\u7684\u7b56\u7565\u6765\u5e94\u5bf9\u65b0\u7684\u6d4b\u8bd5\u4efb\u52a1\u3002<\/section>\n<section><\/section>\n<section>\u591a\u8f6e\u6b21 DPO \u901a\u8fc7\u4e3a\u88ab\u62d2\u7edd\u7684\u8f68\u8ff9\u5f15\u5165\u300c\u8d1f\u68af\u5ea6\u300d\u53ef\u4ee5\u7f13\u89e3\u8fd9\u4e2a\u95ee\u9898\uff0c\u4f46\u5982\u679c\u6ca1\u6709\u5728\u957f\u671f\u65f6\u95f4\u4e0a\u8fdb\u884c\u9002\u5f53\u7684 credit \u5206\u914d\uff0c\u6539\u8fdb\u4ecd\u7136\u6709\u9650\u3002<\/section>\n<section><\/section>\n<section>\u800c\u5982\u679c\u5229\u7528\u8bad\u7ec3\u65f6\u95f4\u4fe1\u606f\u663e\u5f0f\u5730\u8bad\u7ec3\u6bcf\u8f6e\u6b21\u7684\u5956\u52b1\u6a21\u578b\u6765\u6267\u884c credit \u5206\u914d\uff0c\u53ef\u4ee5\u89c2\u5bdf\u5230\u76f8\u6bd4\u4e8e\u591a\u8f6e\u6b21 DPO\uff0cSWEET-RL \u5728\u4e24\u4e2a\u4efb\u52a1\u4e0a\u90fd\u6709\u663e\u8457\u63d0\u5347\uff08\u540e\u7aef\u7f16\u7a0b\u6210\u529f\u7387\u63d0\u9ad8 6%\uff0c\u524d\u7aef\u8bbe\u8ba1\u80dc\u7387\u63d0\u9ad8 5.4%\uff09\u3002<\/section>\n<section><\/section>\n<section>\u4e8b\u5b9e\u4e0a\uff0c\u4f7f\u7528 Llama-3.18B-Instruct \u5f97\u5230\u7684 SWEET-RL \u6a21\u578b\u4e0d\u4ec5\u5728\u53c2\u6570\u4e0a\u4e0e Llama-3.1-70B-Instruct \u7684\u6027\u80fd\u76f8\u5f53\uff0c\u800c\u4e14\u5728\u6027\u80fd\u4e0a\u4e5f\u8db3\u4ee5\u4e0e GPT-4o \u548c o1-mini \u7b49 SOTA \u4e13\u6709\u6a21\u578b\u6bd4\u80a9\u3002<\/section>\n<section><\/section>\n<section>\u6709\u5173 SWEET-RL \u7684\u66f4\u591a\u5b9e\u9a8c\u7ec6\u8282\u548c\u5206\u6790\u8bf7\u53c2\u9605\u539f\u8bba\u6587\u3002<\/section>\n<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2025-03-23-3\" target=\"_blank\">\u7530\u6e0a\u680b\u548cSergey Levine\u53c2\u4e0e\u5f00\u53d1\u65b0\u578bRL\u7b97\u6cd5\uff0c\u80fd\u901a\u8fc7\u591a\u8f6e\u8bad\u7ec3\u8ba9\u667a\u80fd\u4f53\u5b66\u4f1a\u534f\u4f5c\u63a8\u7406<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:\u7530\u6e0a\u680b\u548cSergey  [&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":[70],"class_list":["post-39599","post","type-post","status-publish","format-standard","hentry","category-news","tag-agent"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v26.4 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>\u7530\u6e0a\u680b\u548cSergey Levine\u53c2\u4e0e\u5f00\u53d1\u65b0\u578bRL\u7b97\u6cd5\uff0c\u80fd\u901a\u8fc7\u591a\u8f6e\u8bad\u7ec3\u8ba9\u667a\u80fd\u4f53\u5b66\u4f1a\u534f\u4f5c\u63a8\u7406 - \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=39599\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/17aitech.com\/?p=39599\",\"url\":\"https:\/\/17aitech.com\/?p=39599\",\"name\":\"\u7530\u6e0a\u680b\u548cSergey Levine\u53c2\u4e0e\u5f00\u53d1\u65b0\u578bRL\u7b97\u6cd5\uff0c\u80fd\u901a\u8fc7\u591a\u8f6e\u8bad\u7ec3\u8ba9\u667a\u80fd\u4f53\u5b66\u4f1a\u534f\u4f5c\u63a8\u7406 - \u4e00\u8d77AI\u6280\u672f\",\"isPartOf\":{\"@id\":\"https:\/\/17aitech.com\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/17aitech.com\/?p=39599#primaryimage\"},\"image\":{\"@id\":\"https:\/\/17aitech.com\/?p=39599#primaryimage\"},\"thumbnailUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-3e6b620684d525a6b7702cdf2a598d43.png\",\"datePublished\":\"2025-03-23T14:02:58+00:00\",\"author\":{\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d\"},\"breadcrumb\":{\"@id\":\"https:\/\/17aitech.com\/?p=39599#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/17aitech.com\/?p=39599\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/17aitech.com\/?p=39599#primaryimage\",\"url\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-3e6b620684d525a6b7702cdf2a598d43.png\",\"contentUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-3e6b620684d525a6b7702cdf2a598d43.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/17aitech.com\/?p=39599#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/17aitech.com\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u7530\u6e0a\u680b\u548cSergey Levine\u53c2\u4e0e\u5f00\u53d1\u65b0\u578bRL\u7b97\u6cd5\uff0c\u80fd\u901a\u8fc7\u591a\u8f6e\u8bad\u7ec3\u8ba9\u667a\u80fd\u4f53\u5b66\u4f1a\u534f\u4f5c\u63a8\u7406\"}]},{\"@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\":\"https:\/\/17aitech.com\/#\/schema\/person\/image\/\",\"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":"\u7530\u6e0a\u680b\u548cSergey Levine\u53c2\u4e0e\u5f00\u53d1\u65b0\u578bRL\u7b97\u6cd5\uff0c\u80fd\u901a\u8fc7\u591a\u8f6e\u8bad\u7ec3\u8ba9\u667a\u80fd\u4f53\u5b66\u4f1a\u534f\u4f5c\u63a8\u7406 - \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=39599","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/17aitech.com\/?p=39599","url":"https:\/\/17aitech.com\/?p=39599","name":"\u7530\u6e0a\u680b\u548cSergey Levine\u53c2\u4e0e\u5f00\u53d1\u65b0\u578bRL\u7b97\u6cd5\uff0c\u80fd\u901a\u8fc7\u591a\u8f6e\u8bad\u7ec3\u8ba9\u667a\u80fd\u4f53\u5b66\u4f1a\u534f\u4f5c\u63a8\u7406 - \u4e00\u8d77AI\u6280\u672f","isPartOf":{"@id":"https:\/\/17aitech.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/17aitech.com\/?p=39599#primaryimage"},"image":{"@id":"https:\/\/17aitech.com\/?p=39599#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-3e6b620684d525a6b7702cdf2a598d43.png","datePublished":"2025-03-23T14:02:58+00:00","author":{"@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"breadcrumb":{"@id":"https:\/\/17aitech.com\/?p=39599#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/17aitech.com\/?p=39599"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/?p=39599#primaryimage","url":"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-3e6b620684d525a6b7702cdf2a598d43.png","contentUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2025\/03\/frc-3e6b620684d525a6b7702cdf2a598d43.png"},{"@type":"BreadcrumbList","@id":"https:\/\/17aitech.com\/?p=39599#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/17aitech.com\/"},{"@type":"ListItem","position":2,"name":"\u7530\u6e0a\u680b\u548cSergey Levine\u53c2\u4e0e\u5f00\u53d1\u65b0\u578bRL\u7b97\u6cd5\uff0c\u80fd\u901a\u8fc7\u591a\u8f6e\u8bad\u7ec3\u8ba9\u667a\u80fd\u4f53\u5b66\u4f1a\u534f\u4f5c\u63a8\u7406"}]},{"@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":"https:\/\/17aitech.com\/#\/schema\/person\/image\/","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\/39599","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=39599"}],"version-history":[{"count":0,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/39599\/revisions"}],"wp:attachment":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=39599"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=39599"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=39599"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}