{"id":35846,"date":"2025-01-01T23:00:21","date_gmt":"2025-01-01T15:00:21","guid":{"rendered":"https:\/\/17aitech.com\/?p=35846"},"modified":"2025-01-01T23:00:21","modified_gmt":"2025-01-01T15:00:21","slug":"%e7%ab%af%e5%88%b0%e7%ab%af%e4%bc%98%e5%8c%96%e6%89%80%e6%9c%89%e8%83%bd%e5%8a%9b%ef%bc%8c%e5%ad%97%e8%8a%82%e8%b7%b3%e5%8a%a8%e6%8f%90%e5%87%ba%e5%bc%ba%e5%8c%96%e5%ad%a6%e4%b9%a0llm-agent%e6%a1%86","status":"publish","type":"post","link":"https:\/\/17aitech.com\/?p=35846","title":{"rendered":"\u7aef\u5230\u7aef\u4f18\u5316\u6240\u6709\u80fd\u529b\uff0c\u5b57\u8282\u8df3\u52a8\u63d0\u51fa\u5f3a\u5316\u5b66\u4e60LLM Agent\u6846\u67b6AGILE"},"content":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-09-30-4\" target=\"_blank\">\u7aef\u5230\u7aef\u4f18\u5316\u6240\u6709\u80fd\u529b\uff0c\u5b57\u8282\u8df3\u52a8\u63d0\u51fa\u5f3a\u5316\u5b66\u4e60LLM Agent\u6846\u67b6AGILE<\/a><\/p>\n<p>\u5927\u8bed\u8a00\u6a21\u578b\uff08Large Language Models, LLMs\uff09\u7684\u5f3a\u5927\u80fd\u529b\u63a8\u52a8\u4e86 LLM Agent \u7684\u8fc5\u901f\u53d1\u5c55\u3002\u56f4\u7ed5\u589e\u5f3a LLM Agent \u7684\u80fd\u529b\uff0c\u8fd1\u671f\u76f8\u5173\u7814\u7a76\u63d0\u51fa\u4e86\u82e5\u5e72\u5173\u952e\u7ec4\u4ef6\u6216\u5de5\u4f5c\u6d41\u3002\u7136\u800c\uff0c\u5982\u4f55\u5c06\u6838\u5fc3\u8981\u7d20\u96c6\u6210\u5230\u4e00\u4e2a\u7edf\u4e00\u7684\u6846\u67b6\u4e2d\uff0c\u80fd\u591f\u8fdb\u884c\u7aef\u5230\u7aef\u4f18\u5316\uff0c\u4ecd\u7136\u662f\u4e00\u4e2a\u4e9f\u5f85\u89e3\u51b3\u7684\u95ee\u9898\u3002<\/p>\n<p>\u6765\u81ea\u5b57\u8282\u8df3\u52a8 ByteDance Research \u7684\u7814\u7a76\u4eba\u5458\u63d0\u51fa\u4e86\u57fa\u4e8e\u5f3a\u5316\u5b66\u4e60\uff08Reinforcement Learning, RL\uff09\u7684 LLM Agent \u6846\u67b6 \u2014\u2014AGILE\u3002<strong>\u8be5\u6846\u67b6\u4e0b\uff0cAgent \u80fd\u591f\u62e5\u6709\u8bb0\u5fc6\u3001\u5de5\u5177\u4f7f\u7528\u3001\u89c4\u5212\u3001\u53cd\u601d\u3001\u4e0e\u5916\u754c\u73af\u5883\u4ea4\u4e92\u3001\u4e3b\u52a8\u6c42\u52a9\u4e13\u5bb6\u7b49\u591a\u79cd\u80fd\u529b\uff0c\u5e76\u4e14\u901a\u8fc7\u5f3a\u5316\u5b66\u4e60\u5b9e\u73b0\u6240\u6709\u80fd\u529b\u7684\u7aef\u5230\u7aef\u8bad\u7ec3\u3002\u5c24\u5176\u503c\u5f97\u6ce8\u610f\u7684\u662f\uff0cAGILE \u6846\u67b6\u5141\u8bb8 Agent \u5728\u4e0d\u81ea\u4fe1\u65f6\u4e3b\u52a8\u5411\u4eba\u7c7b\u4e13\u5bb6\u5bfb\u6c42\u5efa\u8bae\u3002<\/strong>\u8fd9\u5e26\u6765\u4e86\u4e24\u5927\u4f18\u52bf\uff1a\u9996\u5148\uff0cAgent \u5728\u5904\u7406\u590d\u6742\u95ee\u9898\u65f6\u80fd\u591f\u6301\u7eed\u4fdd\u6301\u9ad8\u51c6\u786e\u7387\uff1b\u5176\u6b21\uff0c\u901a\u8fc7\u5411\u4eba\u7c7b\u5b66\u4e60\uff0c\u589e\u5f3a\u4e86\u5176\u5feb\u901f\u9002\u5e94\u65b0\u4efb\u52a1\u7684\u6cdb\u5316\u80fd\u529b\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-49c515656486c4d41a831c1ed1383baf.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-49c515656486c4d41a831c1ed1383baf.png\"><\/a><\/p>\n<ul>\n<li>\n<p>\u8bba\u6587\u5730\u5740\uff1ahttps:\/\/arxiv.org\/abs\/2405.14751<\/p>\n<\/li>\n<li>\n<p>\u4ee3\u7801\u5730\u5740\uff1ahttps:\/\/github.com\/bytarnish\/AGILE<\/p>\n<\/li>\n<\/ul>\n<p>\u7814\u7a76\u8005\u8ba9 AGILE \u6846\u67b6\u5728\u590d\u6742\u95ee\u7b54\u4efb\u52a1\u4e2d\u8fdb\u884c\u4e86\u9a8c\u8bc1\u3002<strong>\u5728 ProductQA \u4e0e MedMCQA \u4efb\u52a1\u4e0a\uff0c\u7ecf\u8fc7 RL \u8bad\u7ec3\u540e\u7684 13B \u6a21\u578b\u6216\u8005 7B \u6a21\u578b\u80fd\u591f\u8d85\u8d8a\u63d0\u793a\u5de5\u7a0b\u6784\u5efa\u7684 GPT-4 Agent\u3002<\/strong><\/p>\n<p><strong>AGILE Agent \u6846\u67b6<\/strong><\/p>\n<p>\u5982\u56fe 1 (a) \u6240\u793a\uff0cAGILE \u6846\u67b6\u5305\u542b\u56db\u4e2a\u6838\u5fc3\u6a21\u5757\uff1aLLM\u3001\u8bb0\u5fc6\uff08Memory\uff09\u3001\u5de5\u5177\uff08Tools\uff09\u548c\u6267\u884c\u5668\uff08Executor\uff09\u3002LLM \u8d1f\u8d23\u9884\u6d4b\u52a8\u4f5c\uff1b\u8bb0\u5fc6\u6a21\u5757\u8bb0\u5f55 Agent \u7684\u8f68\u8ff9\uff1b\u5de5\u5177\u6a21\u5757\u63d0\u4f9b Agent \u53ef\u4ee5\u8c03\u7528\u7684\u5916\u90e8 API\uff1b\u6267\u884c\u5668\u4f1a\u76d1\u542c LLM \u7684\u8f93\u51fa\uff0c\u6839\u636e LLM \u7684\u8f93\u51fa\u8c03\u7528\u76f8\u5e94\u7684\u6a21\u5757\u6267\u884c\u5177\u4f53\u7684\u64cd\u4f5c\uff0c\u5e76\u5c06\u6267\u884c\u64cd\u4f5c\u5f97\u5230\u7684\u7ed3\u679c\u6dfb\u52a0\u5230 LLM \u7684\u4e0a\u4e0b\u6587\u4e2d\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-9231219aa270b040865918acd4df4d13.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-9231219aa270b040865918acd4df4d13.png\"><\/a><\/p>\n<p>\u56fe 1 (b) \u5c55\u793a\u4e86 AGILE Agent \u5728\u7535\u5546\u95ee\u7b54\u573a\u666f\u4e2d\u7684\u4e00\u4e2a\u793a\u4f8b\u3002\u7535\u5546\u95ee\u7b54\u662f\u4e00\u4e2a\u590d\u6742\u7684\u5b9e\u9645\u5e94\u7528\u573a\u666f\uff0c\u9762\u4e34\u7684\u6311\u6218\u5305\u62ec\u9700\u8981\u6d77\u91cf\u5546\u54c1\u7684\u9886\u57df\u77e5\u8bc6\u3001\u7075\u6d3b\u8fd0\u7528\u5546\u54c1\u68c0\u7d22\u5de5\u5177\u3001\u4ee5\u53ca\u5feb\u901f\u9002\u5e94\u4e0d\u65ad\u6d8c\u73b0\u7684\u65b0\u5546\u54c1\u3002\u5982\u56fe 1 (b) \u6240\u793a\uff0cAGILE Agent \u4f1a\u6839\u636e\u7528\u6237\u7684\u95ee\u9898\u68c0\u7d22\u8bb0\u5fc6\uff0c\u5982\u679c\u65e0\u6cd5\u786e\u5b9a\u95ee\u9898\u7684\u7b54\u6848\uff0cAgent \u4f1a\u5411\u4eba\u7c7b\u4e13\u5bb6\u5bfb\u6c42\u5e2e\u52a9\u3002\u5728\u83b7\u5f97\u4e13\u5bb6\u7684\u53cd\u9988\u540e\uff0cAgent \u4f1a\u53cd\u601d\u5e76\u5c06\u65b0\u638c\u63e1\u7684\u9886\u57df\u77e5\u8bc6\u5b58\u50a8\u5728\u8bb0\u5fc6\u4e2d\u3002\u5728\u672a\u6765\uff0c\u9762\u5bf9\u65b0\u7684\u95ee\u9898\u65f6\uff0cAgent \u80fd\u591f\u4ece\u8bb0\u5fc6\u4e2d\u68c0\u7d22\u5230\u8fd9\u6761\u77e5\u8bc6\uff0c\u5e76\u57fa\u4e8e\u8fd9\u4e9b\u77e5\u8bc6\u76f4\u63a5\u7ed9\u51fa\u51c6\u786e\u7684\u7b54\u6848\u3002\u9664\u6b64\u4e4b\u5916\uff0cAGILE Agent \u4e5f\u4f1a\u6839\u636e\u7528\u6237\u7684\u95ee\u9898\u9009\u62e9\u662f\u5426\u8c03\u7528\u5916\u90e8\u5de5\u5177\uff08\u5982\u641c\u7d22\u3001\u6570\u636e\u5e93\u68c0\u7d22\uff09\uff0c\u8f85\u52a9\u751f\u6210\u6700\u7ec8\u7684\u56de\u7b54\u3002<\/p>\n<p><strong>\u5f3a\u5316\u5b66\u4e60\u5b9a\u4e49\uff1a<\/strong>LLM Agents \u88ab\u5b9a\u4e49\u4e3a\u4e00\u4e2a token-level MDP\uff08Markov Decision Process\uff09\u3002\u52a8\u4f5c\u7a7a\u95f4\uff08Action space\uff09\u7531 LLM \u7684\u8bcd\u8868\u6784\u6210\uff0cLLM \u751f\u6210\u7684\u6bcf\u4e00\u4e2a token \u662f\u4e00\u4e2a\u52a8\u4f5c\uff0cLLM \u672c\u8eab\u5219\u4f5c\u4e3a Agent \u7684\u7b56\u7565\u6a21\u578b\uff08Policy model\uff09\u3002Agent \u7684\u72b6\u6001\uff08State\uff09\u7531 LLM \u4e0a\u4e0b\u6587\u548c\u8bb0\u5fc6\u7ec4\u6210\u3002\u5728\u6bcf\u4e2a\u65f6\u523b\uff0cLLM \u9884\u6d4b\u52a8\u4f5c\uff0c\u6267\u884c\u5668\u6839\u636e\u9884\u5b9a\u4e49\u7684\u903b\u8f91\u5b8c\u6210\u72b6\u6001\u8f6c\u79fb\uff0c\u540c\u65f6\u73af\u5883\u7ed9\u4e88 Agent \u76f8\u5e94\u7684\u5956\u52b1\uff08Reward\uff09\u3002<\/p>\n<p>\u5728 AGILE \u6846\u67b6\u4e0b\uff0cAgent \u6709\u4e24\u79cd\u7b56\u7565\u5b66\u4e60\u65b9\u6cd5\u3002\u7b2c\u4e00\u79cd\u662f\u6a21\u4eff\u5b66\u4e60\uff0c\u901a\u8fc7\u6536\u96c6\u4eba\u7c7b\u8f68\u8ff9\u6570\u636e\u6216\u66f4\u9ad8\u7ea7\u522b Agent \u7684\u8f68\u8ff9\u6570\u636e\uff0c\u5bf9 LLM \u8fdb\u884c SFT \u8bad\u7ec3\u3002\u7b2c\u4e8c\u79cd\u662f\u5f3a\u5316\u5b66\u4e60\uff0c\u901a\u8fc7\u5b9a\u4e49\u5956\u52b1\u51fd\u6570\uff0c\u5229\u7528\u5f3a\u5316\u5b66\u4e60\u7b97\u6cd5\u6765\u8bad\u7ec3 LLM\u3002<\/p>\n<p>\u6b64\u5916\uff0cLLM Agent \u53ef\u80fd\u4f1a\u4ea7\u751f\u8de8\u8d8a\u6570\u767e\u4e07\u4e2a tokens \u7684\u8f68\u8ff9\uff0c\u8fd9\u4e3a\u76f4\u63a5\u8bad\u7ec3\u5e26\u6765\u4e86\u6311\u6218\u3002\u4e3a\u4e86\u5e94\u5bf9\u8fd9\u79cd\u957f\u7a0b\u4f9d\u8d56\u7684\u95ee\u9898\uff0c\u7814\u7a76\u4eba\u5458\u63d0\u51fa\u4e86\u4e00\u79cd\u7247\u6bb5\u7ea7\u522b\u7684\u4f18\u5316\u7b97\u6cd5\u3002<\/p>\n<p><strong>\u4e3b\u52a8\u5bfb\u6c42\u5e2e\u52a9\uff1a<\/strong>AGILE \u6846\u67b6\u5141\u8bb8 Agent \u4e3b\u52a8\u5411\u5916\u90e8\u7684\u4eba\u7c7b\u4e13\u5bb6\u5bfb\u6c42\u5e2e\u52a9\u3002\u8fd9\u79cd\u673a\u5236\u6709\u4e24\u4e2a\u4f18\u52bf\uff1a\u9996\u5148\uff0c\u5f53 Agent \u9047\u5230\u4e0d\u786e\u5b9a\u7684\u60c5\u51b5\u65f6\uff0c\u901a\u8fc7\u6c42\u52a9\u4eba\u7c7b\u4e13\u5bb6\uff0c\u786e\u4fdd\u5176\u5728\u5b9e\u9645\u5e94\u7528\u4e2d\u8fbe\u5230\u9ad8\u51c6\u786e\u7387\u3002\u5176\u6b21\uff0cAgent \u80fd\u591f\u901a\u8fc7\u5bf9\u4eba\u7c7b\u7684\u53cd\u9988\u53cd\u601d\u5e76\u79ef\u7d2f\u77e5\u8bc6\uff0c\u4ece\u800c\u66f4\u5feb\u9002\u5e94\u65b0\u73af\u5883\uff0c\u63d0\u5347\u5176\u6cdb\u5316\u80fd\u529b\u3002\u7136\u800c\u51b3\u5b9a\u4f55\u65f6\u5bfb\u6c42\u5e2e\u52a9\u662f\u4e00\u4e2a\u590d\u6742\u51b3\u7b56\uff0c\u5b83\u6d89\u53ca\u5230 Agent \u7684\u81ea\u6211\u8bc4\u4f30\u3001\u4eba\u7c7b\u53cd\u9988\u5bf9\u672a\u6765\u7684\u4ef7\u503c\u4ee5\u53ca\u4eba\u7c7b\u4e13\u5bb6\u7684\u6210\u672c\u3002\u56e0\u6b64\uff0c\u6807\u6ce8\u4f55\u65f6\u5e94\u8be5\u6c42\u52a9\u662f\u5f88\u96be\u7684\u3002\u4f46\u5728\u5f3a\u5316\u5b66\u4e60\u6846\u67b6\u4e2d\uff0c\u53ef\u4ee5\u901a\u8fc7\u5b9a\u4e49\u76f8\u5173\u5956\u52b1\uff0c\u5c06\u8fd9\u79cd\u6c42\u52a9\u80fd\u529b\u4f5c\u4e3a\u7b56\u7565\u6a21\u578b\u7684\u4e00\u90e8\u5206\uff0c\u5728\u7aef\u5230\u7aef\u8bad\u7ec3\u4e2d\u5f97\u5230\u63d0\u5347\u3002<\/p>\n<p><strong>\u5b9e\u9a8c\u7ed3\u679c<\/strong><\/p>\n<p><strong>ProductQA<\/strong><\/p>\n<p>ProductQA \u662f\u4e00\u4e2a\u5546\u54c1\u95ee\u7b54\u4efb\u52a1\u3002\u8be5\u6570\u636e\u96c6\u5305\u542b 26 \u4e2a\u5bf9\u5e94\u4e0d\u540c\u5546\u54c1\u7c7b\u522b\u7684 QA \u4efb\u52a1\uff0c\u6bcf\u4e2a\u4efb\u52a1\u5e73\u5747\u5305\u542b 3,393 \u4e2a\u95ee\u9898\u3002ProductQA \u5305\u62ec\u57fa\u4e8e\u4e8b\u5b9e\u7684\u95ee\u9898\u3001\u63a8\u7406\u95ee\u9898\u548c\u5546\u54c1\u63a8\u8350\u95ee\u9898\uff0c\u5b83\u80fd\u591f\u5168\u9762\u8bc4\u4f30 Agent \u5904\u7406\u5386\u53f2\u4fe1\u606f\u548c\u7d2f\u79ef\u77e5\u8bc6\u3001\u5229\u7528\u5de5\u5177\u3001\u5411\u4eba\u6c42\u52a9\u3001\u81ea\u6211\u8bc4\u4f30\u548c\u53cd\u601d\u7684\u80fd\u529b\u3002\u6b64\u5916\uff0c\u8bad\u7ec3\u548c\u6d4b\u8bd5\u96c6\u7531\u4e0d\u540c\u7684\u4efb\u52a1\u6784\u6210\uff0c\u4ee5\u8bc4\u4f30 Agent \u5728\u65b0\u5546\u54c1\u95ee\u7b54\u4e0a\u7684\u6cdb\u5316\u80fd\u529b\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-b2a15a8892d1c1feea3a2a37df5869ce.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-b2a15a8892d1c1feea3a2a37df5869ce.png\"><\/a><\/p>\n<p>\u5728\u5546\u54c1\u95ee\u7b54\uff08ProductQA\uff09\u4efb\u52a1\u4e0a\uff0c\u57fa\u4e8e Vicuna-13b \u8bad\u7ec3\u7684 AGILE Agent\uff08agile-vic13b-ppo\uff09\u8868\u73b0\u8d85\u8fc7\u4e86 GPT-4\uff08gpt4-prompt\uff09\u4e0e\u63d0\u5347\u5de5\u7a0b\u6784\u5efa\u7684 GPT-4 Agent\uff08agile-gpt4-prompt\uff09\u3002\u5728\u4f7f\u7528\u4e86\u76f8\u5f53\u7684\u6c42\u52a9\u6bd4\u4f8b\uff08Advice Rate\uff09\u7684\u60c5\u51b5\u4e0b\uff0cagile-vic13b-ppo \u7684 acc \u76f8\u6bd4\u4e8e agile-gpt4-prompt \u63d0\u5347\u4e86 7.4%\uff0c\u5728 Total Score \u4e0a\u63d0\u5347\u4e86 9.2%\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-c5abcf1f58c10fbf69642dff37c54b4e.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-c5abcf1f58c10fbf69642dff37c54b4e.png\"><\/a><\/p>\n<p>\u4ece\u4e0a\u56fe\u53ef\u4ee5\u770b\u51fa\uff0c\u5728\u6267\u884c\u5305\u542b\u4e0a\u5343\u4e2a\u95ee\u7b54\u7684\u4efb\u52a1\u6574\u4e2a\u8fc7\u7a0b\u4e2d\uff0cagile-vic13b-ppo \u7684 acc \u6301\u7eed\u7a33\u5b9a\u5730\u9ad8\u4e8e agile-gpt4-prompt\u3002\u540c\u65f6\u5bfb\u6c42\u4eba\u7c7b\u5e2e\u52a9\u7684\u9891\u7387\uff08Advice Rate\uff09\u968f\u7740\u95ee\u7b54\u8f6e\u6570\u7684\u589e\u52a0\u9010\u6e10\u4e0b\u964d\u3002\u6b64\u5916\uff0c\u901a\u8fc7\u8c03\u6574\u4eba\u7c7b\u7684\u54a8\u8be2\u6210\u672c\uff08Seeking Advice Cost\uff09\u548c\u8fdb\u884c\u5f3a\u5316\u5b66\u4e60\u8bad\u7ec3\uff0cAGILE Agent \u53ef\u4ee5\u6709\u6548\u5b9e\u73b0\u51c6\u786e\u7387\u4e0e\u4e13\u5bb6\u6210\u672c\u7684\u6743\u8861\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-6d5ef6568d5fb65d2081be7c7b18327a.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-6d5ef6568d5fb65d2081be7c7b18327a.png\"><\/a><\/p>\n<p>\u6d88\u878d\u5b9e\u9a8c\u7ed3\u679c\u663e\u793a\uff0c\u8bb0\u5fc6\u3001\u53cd\u601d\u3001\u54a8\u8be2\u4eba\u7c7b\u5efa\u8bae\u3001\u5de5\u5177\u4f7f\u7528\u3001RL \u8bad\u7ec3\u5728\u5b9e\u73b0\u9ad8\u6027\u80fd\u7684 AGILE Agent \u4e2d\u5747\u5177\u6709\u91cd\u8981\u4f5c\u7528\u3002<\/p>\n<p><strong>MedMCQA<\/strong><\/p>\n<p>MedMCQA \u662f\u4e00\u4e2a\u591a\u9879\u9009\u62e9\u7684\u95ee\u7b54\u6570\u636e\u96c6\uff0c\u5176\u95ee\u9898\u6765\u81ea\u533b\u5b66\u9662\u5165\u5b66\u8003\u8bd5\u3002\u5728 MedMCQA \u4efb\u52a1\u4e0a\uff0c\u57fa\u4e8e Meerkat-7b \u8bad\u7ec3\u7684 AGILE Agent\uff08agile-mek7b-ppo\uff09\u8868\u73b0\u4f18\u4e8e GPT-4 Agent\u3002\u51c6\u786e\u7387\u8fbe\u5230\u4e86 85.2%\uff0c\u8d85\u8fc7\u4e86\u4e4b\u524d\u7684 SOTA \u65b9\u6cd5 \u2014\u2014GPT 4-MedPrompt \u7684 79.1%\u3002\u6d88\u878d\u5b9e\u9a8c\u7ed3\u679c\u8fdb\u4e00\u6b65\u9a8c\u8bc1\u4e86\u8bb0\u5fc6\u3001\u54a8\u8be2\u4eba\u7c7b\u5efa\u8bae\u3001\u53cd\u601d\u548c RL \u8bad\u7ec3\u7684\u91cd\u8981\u6027\u3002<\/p>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-c7ae1f466c04afa191991b8dcfc38a1e.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-c7ae1f466c04afa191991b8dcfc38a1e.png\"><\/a><\/p>\n<p>\u66f4\u591a\u7814\u7a76\u7ec6\u8282\uff0c\u53ef\u53c2\u8003\u539f\u8bba\u6587\u3002<\/p>\n<p><strong>\u603b\u7ed3<\/strong><\/p>\n<section>AGILE\u662f\u4e00\u79cd\u57fa\u4e8e\u5f3a\u5316\u5b66\u4e60\u7684LLM Agent\u6846\u67b6\u3002AGILE Agent\u5177\u5907\u62e5\u6709\u957f\u671f\u8bb0\u5fc6\u3001\u4f7f\u7528\u5916\u90e8\u5de5\u5177\u3001\u5411\u4eba\u7c7b\u4e13\u5bb6\u8fdb\u884c\u54a8\u8be2\u3001\u53cd\u601d\u7b49\u80fd\u529b\uff0c\u5e76\u4e14\u6240\u6709\u80fd\u529b\u53ef\u4ee5\u8fdb\u884c\u7aef\u5230\u7aef\u7684\u7edf\u4e00\u4f18\u5316\u3002AGILE\u7684\u4e00\u4e2a\u91cd\u8981\u7279\u70b9\u662fAgent\u5728\u9047\u5230\u65e0\u6cd5\u81ea\u884c\u89e3\u51b3\u7684\u95ee\u9898\u65f6\u53ef\u4ee5\u5411\u4eba\u7c7b\u4e13\u5bb6\u54a8\u8be2\uff0c\u8fd9\u79cd\u673a\u5236\u4fdd\u8bc1\u4e86\u5e94\u7528\u573a\u666f\u4e2d\u5bf9\u9ad8\u51c6\u786e\u7387\u7684\u8981\u6c42\uff0c\u540c\u65f6\u589e\u5f3a\u4e86Agent\u7684\u5b66\u4e60\u4e0e\u6cdb\u5316\u80fd\u529b\u3002\u5b9e\u9a8c\u8868\u660e\uff0c\u7ecf\u8fc7\u5f3a\u5316\u5b66\u4e60\u8bad\u7ec3\u768413B\u62167B\u6a21\u578b\u7684Agent\uff0c\u5176\u80fd\u529b\u53ef\u4ee5\u8d85\u8fc7GPT-4 Agent\u3002<\/section>\n<section><\/section>\n<section><strong>\u56e2\u961f\u4ecb\u7ecd<\/strong><\/section>\n<section><\/section>\n<section>ByteDance Research \u4e13\u6ce8\u4e8e\u4eba\u5de5\u667a\u80fd\u9886\u57df\u7684\u524d\u6cbf\u6280\u672f\u7814\u7a76\uff0c\u6db5\u76d6\u4e86\u673a\u5668\u7ffb\u8bd1\u3001\u89c6\u9891\u751f\u6210\u57fa\u7840\u6a21\u578b\u3001\u673a\u5668\u4eba\u7814\u7a76\u3001\u673a\u5668\u5b66\u4e60\u516c\u5e73\u6027\u3001\u91cf\u5b50\u5316\u5b66\u3001AI \u5236\u836f\u3001\u5206\u5b50\u52a8\u529b\u5b66\u7b49\u591a\u6280\u672f\u7814\u7a76\u9886\u57df\uff0c\u540c\u65f6\u81f4\u529b\u4e8e\u5c06\u7814\u7a76\u6210\u679c\u843d\u5730\uff0c\u4e3a\u516c\u53f8\u73b0\u6709\u7684\u4ea7\u54c1\u548c\u4e1a\u52a1\u63d0\u4f9b\u6838\u5fc3\u6280\u672f\u652f\u6301\u548c\u670d\u52a1\u3002<\/section>\n<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-09-30-4\" target=\"_blank\">\u7aef\u5230\u7aef\u4f18\u5316\u6240\u6709\u80fd\u529b\uff0c\u5b57\u8282\u8df3\u52a8\u63d0\u51fa\u5f3a\u5316\u5b66\u4e60LLM Agent\u6846\u67b6AGILE<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:\u7aef\u5230\u7aef\u4f18\u5316\u6240\u6709\u80fd\u529b\uff0c\u5b57 [&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":[70,55],"class_list":["post-35846","post","type-post","status-publish","format-standard","hentry","category-news","tag-agent","tag-55"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v26.4 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>\u7aef\u5230\u7aef\u4f18\u5316\u6240\u6709\u80fd\u529b\uff0c\u5b57\u8282\u8df3\u52a8\u63d0\u51fa\u5f3a\u5316\u5b66\u4e60LLM Agent\u6846\u67b6AGILE - \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=35846\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/17aitech.com\/?p=35846\",\"url\":\"https:\/\/17aitech.com\/?p=35846\",\"name\":\"\u7aef\u5230\u7aef\u4f18\u5316\u6240\u6709\u80fd\u529b\uff0c\u5b57\u8282\u8df3\u52a8\u63d0\u51fa\u5f3a\u5316\u5b66\u4e60LLM Agent\u6846\u67b6AGILE - \u4e00\u8d77AI\u6280\u672f\",\"isPartOf\":{\"@id\":\"https:\/\/17aitech.com\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/17aitech.com\/?p=35846#primaryimage\"},\"image\":{\"@id\":\"https:\/\/17aitech.com\/?p=35846#primaryimage\"},\"thumbnailUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-49c515656486c4d41a831c1ed1383baf.png\",\"datePublished\":\"2025-01-01T15:00:21+00:00\",\"author\":{\"@id\":\"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d\"},\"breadcrumb\":{\"@id\":\"https:\/\/17aitech.com\/?p=35846#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/17aitech.com\/?p=35846\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\/\/17aitech.com\/?p=35846#primaryimage\",\"url\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-49c515656486c4d41a831c1ed1383baf.png\",\"contentUrl\":\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-49c515656486c4d41a831c1ed1383baf.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/17aitech.com\/?p=35846#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\/\/17aitech.com\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u7aef\u5230\u7aef\u4f18\u5316\u6240\u6709\u80fd\u529b\uff0c\u5b57\u8282\u8df3\u52a8\u63d0\u51fa\u5f3a\u5316\u5b66\u4e60LLM Agent\u6846\u67b6AGILE\"}]},{\"@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":"\u7aef\u5230\u7aef\u4f18\u5316\u6240\u6709\u80fd\u529b\uff0c\u5b57\u8282\u8df3\u52a8\u63d0\u51fa\u5f3a\u5316\u5b66\u4e60LLM Agent\u6846\u67b6AGILE - \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=35846","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/17aitech.com\/?p=35846","url":"https:\/\/17aitech.com\/?p=35846","name":"\u7aef\u5230\u7aef\u4f18\u5316\u6240\u6709\u80fd\u529b\uff0c\u5b57\u8282\u8df3\u52a8\u63d0\u51fa\u5f3a\u5316\u5b66\u4e60LLM Agent\u6846\u67b6AGILE - \u4e00\u8d77AI\u6280\u672f","isPartOf":{"@id":"https:\/\/17aitech.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/17aitech.com\/?p=35846#primaryimage"},"image":{"@id":"https:\/\/17aitech.com\/?p=35846#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-49c515656486c4d41a831c1ed1383baf.png","datePublished":"2025-01-01T15:00:21+00:00","author":{"@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"breadcrumb":{"@id":"https:\/\/17aitech.com\/?p=35846#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/17aitech.com\/?p=35846"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/?p=35846#primaryimage","url":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-49c515656486c4d41a831c1ed1383baf.png","contentUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/09\/frc-49c515656486c4d41a831c1ed1383baf.png"},{"@type":"BreadcrumbList","@id":"https:\/\/17aitech.com\/?p=35846#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/17aitech.com\/"},{"@type":"ListItem","position":2,"name":"\u7aef\u5230\u7aef\u4f18\u5316\u6240\u6709\u80fd\u529b\uff0c\u5b57\u8282\u8df3\u52a8\u63d0\u51fa\u5f3a\u5316\u5b66\u4e60LLM Agent\u6846\u67b6AGILE"}]},{"@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\/35846","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=35846"}],"version-history":[{"count":0,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/35846\/revisions"}],"wp:attachment":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=35846"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=35846"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=35846"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}