{"id":34611,"date":"2024-12-10T00:56:56","date_gmt":"2024-12-09T16:56:56","guid":{"rendered":"https:\/\/17aitech.com\/?p=34611"},"modified":"2024-12-10T00:56:56","modified_gmt":"2024-12-09T16:56:56","slug":"acl-2024-%e5%af%b925%e4%b8%aa%e5%bc%80%e9%97%ad%e6%ba%90%e6%a8%a1%e5%9e%8b%e6%95%b0%e5%ad%a6%e8%af%84%e6%b5%8b%ef%bc%8cgpt-3-5-turbo%e6%89%8d%e5%8b%89%e5%bc%ba%e5%8f%8a%e6%a0%bc","status":"publish","type":"post","link":"https:\/\/17aitech.com\/?p=34611","title":{"rendered":"ACL 2024 | \u5bf925\u4e2a\u5f00\u95ed\u6e90\u6a21\u578b\u6570\u5b66\u8bc4\u6d4b\uff0cGPT-3.5-Turbo\u624d\u52c9\u5f3a\u53ca\u683c"},"content":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-07-18-2\" target=\"_blank\">ACL 2024 | \u5bf925\u4e2a\u5f00\u95ed\u6e90\u6a21\u578b\u6570\u5b66\u8bc4\u6d4b\uff0cGPT-3.5-Turbo\u624d\u52c9\u5f3a\u53ca\u683c<\/a><\/p>\n<section><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-0e61ecaceafa415da61123235006bd58.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-0e61ecaceafa415da61123235006bd58.png\"><\/a><\/section>\n<blockquote data-author-name=\"\" data-content-utf8-length=\"167\" data-source-title=\"\" data-type=\"2\" data-url=\"\">\n<section>\n<section>AIxiv\u4e13\u680f\u662f<mark data-type=\"institutions\" data-id=\"9e96434f-5827-41e3-9b3c-36d39bc0d446\">\u673a\u5668\u4e4b\u5fc3<\/mark>\u53d1\u5e03\u5b66\u672f\u3001\u6280\u672f\u5185\u5bb9\u7684\u680f\u76ee\u3002\u8fc7\u53bb\u6570\u5e74\uff0c<mark data-type=\"institutions\" data-id=\"9e96434f-5827-41e3-9b3c-36d39bc0d446\">\u673a\u5668\u4e4b\u5fc3<\/mark>AIxiv\u4e13\u680f\u63a5\u6536\u62a5\u9053\u4e862000\u591a\u7bc7\u5185\u5bb9\uff0c\u8986\u76d6\u5168\u7403\u5404\u5927\u9ad8\u6821\u4e0e\u4f01\u4e1a\u7684\u9876\u7ea7\u5b9e\u9a8c\u5ba4\uff0c\u6709\u6548\u4fc3\u8fdb\u4e86\u5b66\u672f\u4ea4\u6d41\u4e0e\u4f20\u64ad\u3002\u5982\u679c\u60a8\u6709\u4f18\u79c0\u7684\u5de5\u4f5c\u60f3\u8981\u5206\u4eab\uff0c\u6b22\u8fce\u6295\u7a3f\u6216\u8005\u8054\u7cfb\u62a5\u9053\u3002\u6295\u7a3f\u90ae\u7bb1\uff1aliyazhou@jiqizhixin.com\uff1bzhaoyunfeng@jiqizhixin.com<\/section>\n<\/section>\n<\/blockquote>\n<section><\/section>\n<section><strong>\u672c\u6587\u4f5c\u8005\u6765\u81ea\u9999\u6e2f\u5927\u5b66\u548c<mark data-type=\"institutions\" data-id=\"e3f47c14-ff22-49c7-8e82-4745dc37b451\">\u817e\u8baf<\/mark>\u3002<\/strong><strong>\u4f5c\u8005\u5217\u8868\uff1a\u674e\u6c81\u6850\uff0cLeyang Cui\uff0c\u8d75\u5b66\u4eae\uff0c<mark data-type=\"experts\" data-id=\"01f6cdd5-5b12-48e6-8b18-d860a8b9711b\">\u5b54\u4ee4\u9e4f<\/mark>\uff0cWei Bi\u3002\u5176\u4e2d\uff0c\u7b2c\u4e00\u4f5c\u8005\u674e\u6c81\u6850\u662f\u9999\u6e2f\u5927\u5b66<mark data-type=\"tech_tasks\" data-id=\"c8ff5114-6cbb-49ca-8a89-3ee2826be0b4\">\u81ea\u7136\u8bed\u8a00\u5904\u7406<\/mark>\u5b9e\u9a8c\u5ba4\u7684\u535a\u58eb\u751f\uff0c\u7814\u7a76\u65b9\u5411\u6d89\u53ca<mark data-type=\"tech_tasks\" data-id=\"f04b46c6-6e93-4912-822d-bcecd5528d65\">\u81ea\u7136\u8bed\u8a00\u751f\u6210<\/mark>\u548c\u6587\u672c\u63a8\u7406\uff0c\u4e0e\u535a\u58eb\u751f\u8d75\u5b66\u4eae\u5171\u540c\u5e08\u4ece<mark data-type=\"experts\" data-id=\"01f6cdd5-5b12-48e6-8b18-d860a8b9711b\">\u5b54\u4ee4\u9e4f<\/mark>\u6559\u6388\u3002Leyang Cui \u548c Wei Bi \u662f<mark data-type=\"institutions\" data-id=\"e3f47c14-ff22-49c7-8e82-4745dc37b451\">\u817e\u8baf<\/mark>\u9ad8\u7ea7\u7814\u7a76\u5458\u3002<\/strong><\/section>\n<section><\/section>\n<section><strong>\u524d\u8a00<\/strong><\/section>\n<section><\/section>\n<section>\u5927\u578b<mark data-type=\"tech_tasks\" data-id=\"bf35ef94-d956-4033-a533-0c0828308c36\">\u8bed\u8a00\u6a21\u578b<\/mark>\uff08LLMs\uff09\u5728\u89e3\u51b3\u95ee\u9898\u65b9\u9762\u7684\u975e\u51e1\u80fd\u529b\u65e5\u76ca\u663e\u73b0\u3002\u6700\u8fd1\uff0c\u4e00\u4e2a\u503c\u5f97\u5173\u6ce8\u7684\u73b0\u8c61\u662f\uff0c\u8fd9\u4e9b\u6a21\u578b\u5728\u591a\u9879\u6570\u5b66\u63a8\u7406\u7684<mark data-type=\"concepts\" data-id=\"308c3a45-0fee-4ec6-858e-85b15f440fc0\">\u57fa\u51c6<\/mark>\u6d4b\u8bd5\u4e2d\u83b7\u5f97\u4e86\u60ca\u4eba\u7684\u6210\u7ee9\u3002\u4ee5 GPT-4 \u4e3a\u4f8b\uff0c\u5728\u9ad8\u96be\u5ea6\u5c0f\u5b66\u5e94\u7528\u9898\u6d4b\u8bd5\u96c6 GSM8K [1] \u4e2d\u8868\u73b0\u4f18\u5f02\uff0c<mark data-type=\"concepts\" data-id=\"8be77eae-12da-4e9e-9a88-b7f5bae98c2e\">\u51c6\u786e\u7387<\/mark>\u9ad8\u8fbe 90% \u4ee5\u4e0a\u3002\u540c\u65f6\uff0c\u8bb8\u591a\u5f00\u6e90\u6a21\u578b\u4e5f\u5c55\u73b0\u51fa\u4e86\u4e0d\u4fd7\u7684\u5b9e\u529b\uff0c<mark data-type=\"concepts\" data-id=\"8be77eae-12da-4e9e-9a88-b7f5bae98c2e\">\u51c6\u786e\u7387<\/mark>\u8d85\u8fc7 80%\u3002<\/section>\n<section><\/section>\n<section>\u7136\u800c\u5728\u4f7f\u7528\u4e2d\u6211\u4eec\u7ecf\u5e38\u4f1a\u53d1\u73b0\uff0c\u5f53\u6570\u5b66\u95ee\u9898\u7a0d\u4f5c\u6539\u53d8\u65f6\uff0cLLMs \u53ef\u80fd\u4f1a\u51fa\u73b0\u4e00\u4e9b\u4f4e\u7ea7\u9519\u8bef\uff0c\u5982\u4e0b\u56fe\u6240\u793a\uff1a<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-74fec63c3a1c70304e559ee618bc19bd.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-74fec63c3a1c70304e559ee618bc19bd.png\"><\/a><\/p>\n<section><em><sup>\u56fe 1\uff1aGPT-3.5-Turbo \u6b63\u786e\u89e3\u7b54\u4e86\u4e00\u4e2a\u6570\u5b66\u95ee\u9898\uff08\u5de6\uff09\uff0c\u4f46\u5f53\u5728\u539f\u95ee\u9898\u7684\u57fa\u7840\u4e0a\u6dfb\u52a0\u4e00\u4e2a\u9650\u5236\u6761\u4ef6\uff08\u53f3\uff09\u65f6\uff0cTurbo \u56e0\u4e3a\u6ca1\u6709\u6b63\u786e\u533a\u5206 \u201c\u79bb\u5f00\u201d \u548c \u201c\u8fd4\u56de\u201d \u7684\u65b9\u5411\uff0c\u800c\u8bef\u7528\u8fd0\u7b97\u7b26\u51fa\u9519\u3002<\/sup><\/em><\/section>\n<section><\/section>\n<section>\u6211\u4eec\u4e0d\u7981\u8981\u95ee\uff1a\u5927\u578b<mark data-type=\"tech_tasks\" data-id=\"bf35ef94-d956-4033-a533-0c0828308c36\">\u8bed\u8a00\u6a21\u578b<\/mark>\u662f\u5426\u771f\u7684\u638c\u63e1\u4e86\u6570\u5b66\u77e5\u8bc6\u7684\u7cbe\u9ad3\uff1f\u5b83\u4eec\u662f\u5982\u4f55\u5728\u8fd9\u4e9b\u6d4b\u8bd5\u4e2d\u53d6\u5f97\u5982\u6b64\u9ad8\u5206\u7684\uff1f\u96be\u9053\u4ec5\u4ec5\u662f\u56e0\u4e3a\u6a21\u4eff\u4e86\u5927\u91cf\u8bad\u7ec3\u6570\u636e\u4e2d\u7684\u8868\u9762\u63a8\u7406\u6a21\u5f0f\u5417\uff1fLLMs \u662f\u5426\u771f\u6b63\u7406\u89e3\u6570\u5b66\u6982\u5ff5\uff0c\u4ecd\u662f\u4e00\u4e2a\u503c\u5f97\u63a2\u8ba8\u7684\u95ee\u9898\u3002<\/section>\n<section><\/section>\n<section>\u4e3a\u4e86\u63a2\u7a76\u8fd9\u4e00\u95ee\u9898\uff0c\u672c\u6587\u4f5c\u8005\u8bbe\u8ba1\u4e86\u4e00\u4e2a\u8bc4\u4f30<mark data-type=\"concepts\" data-id=\"308c3a45-0fee-4ec6-858e-85b15f440fc0\">\u57fa\u51c6<\/mark> <strong>GSM-Plus<\/strong>\u3002\u8fd9\u4e2a\u6d4b\u8bd5\u65e8\u5728\u5bf9\u4e00\u4e2a\u95ee\u9898\u8fdb\u884c 8 \u79cd\u4e0d\u540c\u7684\u7ec6\u7c92\u5ea6\u6570\u5b66\u53d8\u6362\uff0c\u7cfb\u7edf\u5730\u8bc4\u4f30\u5f53\u524d LLMs \u5728\u5904\u7406\u57fa\u7840\u6570\u5b66\u5e94\u7528\u9898\u65f6\u7684\u80fd\u529b\u3002\u5728\u8fd9\u4e00\u5168\u65b0\u7684<mark data-type=\"concepts\" data-id=\"308c3a45-0fee-4ec6-858e-85b15f440fc0\">\u57fa\u51c6<\/mark>\u6d4b\u8bd5\u4e2d\uff0c\u8bba\u6587\u5bf9 25 \u4e2a\u4e0d\u540c\u7684 LLMs \u8fdb\u884c\u4e86\u4e25\u683c\u8bc4\u6d4b\uff0c\u5305\u62ec\u4e1a\u754c\u7684\u5f00\u6e90\u548c\u95ed\u6e90\u6a21\u578b\u3002<\/section>\n<section><\/section>\n<section>\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u5bf9\u4e8e\u5927\u591a\u6570 LLMs \u6765\u8bf4\uff0cGSM-Plus \u662f\u4e00\u4e2a\u5177\u6709\u6311\u6218\u6027\u7684<mark data-type=\"concepts\" data-id=\"308c3a45-0fee-4ec6-858e-85b15f440fc0\">\u57fa\u51c6<\/mark>\u6d4b\u8bd5\u3002\u5373\u4fbf\u662f\u5728 GSM8K \u4e0a\uff0cGPT-3.5-Turbo \u5df2\u80fd\u53d6\u5f97 73.62% \u7684<mark data-type=\"concepts\" data-id=\"8be77eae-12da-4e9e-9a88-b7f5bae98c2e\">\u51c6\u786e\u7387<\/mark>\uff0c\u4f46\u5728 GSM-Plus \u4e0a\u4ec5\u80fd\u8fbe\u5230 61.19% \u7684<mark data-type=\"concepts\" data-id=\"8be77eae-12da-4e9e-9a88-b7f5bae98c2e\">\u51c6\u786e\u7387<\/mark>\u3002\u672c\u6587\u5de5\u4f5c\u5df2\u7ecf\u4ee54\uff0c4\uff0c 4.5\u5206\u88abACL2024\u5f55\u7528\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-14ac15985d71494c1641cb2eefaa84ba.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-14ac15985d71494c1641cb2eefaa84ba.png\"><\/a><\/p>\n<ul>\n<li>\n<section>\u8bba\u6587\u6807\u9898\uff1aGSM-Plus: A Comprehensive Benchmark for Evaluating the Robustness of LLMs as Mathematical Problem Solvers<\/section>\n<\/li>\n<li>\n<section>\u8bba\u6587\u5730\u5740\uff1ahttps:\/\/arxiv.org\/pdf\/2402.19255<\/section>\n<\/li>\n<li>\n<section>\u8bba\u6587\u4e3b\u9875\uff1ahttps:\/\/qtli.github.io\/GSM-Plus\/<\/section>\n<\/li>\n<\/ul>\n<section><\/section>\n<section><strong>\u80cc\u666f<\/strong><\/section>\n<section><\/section>\n<section>\u6570\u5b66\u63a8\u7406\u662f<mark data-type=\"concepts\" data-id=\"2d28aa9c-942d-471d-bd96-8bfefb7144e0\">\u4eba\u5de5\u667a\u80fd<\/mark>\u53d1\u5c55\u7684\u91cd\u8981\u8bc1\u660e\u3002\u5b83\u9700\u8981\u4e25\u683c\u7684\u95ee\u9898\u7406\u89e3\u3001\u7b56\u7565\u5236\u5b9a\u548c\u8ba1\u7b97\u6267\u884c\u80fd\u529b\u3002\u5728\u8fc7\u53bb\u51e0\u5e74\u4e2d\uff0c\u8bf8\u591a\u516c\u5f00\u6570\u636e\u96c6\u88ab\u7528\u4e8e\u8bc4\u4f30<mark data-type=\"concepts\" data-id=\"2d28aa9c-942d-471d-bd96-8bfefb7144e0\">\u4eba\u5de5\u667a\u80fd<\/mark>\u7cfb\u7edf\u7684\u6570\u5b66\u63a8\u7406\u80fd\u529b\u3002\u65e9\u671f\u7684\u6570\u5b66\u6570\u636e\u96c6\u4fa7\u91cd\u4e8e\u57fa\u4e8e\u65b9\u7a0b\u7684\u6570\u5b66\u95ee\u9898\u3002\u968f\u540e\uff0c\u66f4\u96be\u7684\u6570\u636e\u96c6\u88ab\u5f15\u5165\uff0c\u6db5\u76d6\u4e86\u5c0f\u5b66\u3001\u9ad8\u4e2d\u548c\u5927\u5b66\u6c34\u5e73\u7684\u6570\u5b66\u95ee\u9898\u3002<\/section>\n<section><\/section>\n<section>\u968f\u7740\u8bc4\u6d4b\u6570\u636e\u96be\u5ea6\u7684\u4e0d\u65ad\u63d0\u9ad8\uff0cLLMs \u7684\u53d1\u5c55\u4e5f\u53d8\u5f97\u5341\u5206\u8fc5\u901f\u3002\u4e3a\u4e86\u63d0\u5347 LLMs \u5728\u6570\u5b66\u9886\u57df\u7684\u6027\u80fd\uff0c\u53ef\u4ee5\u901a\u8fc7\u5728\u591a\u6837\u5316\u7684\u4efb\u52a1\u6570\u636e\u4e0a\u8fdb\u884c\u8bad\u7ec3\uff0c\u4f7f\u7528\u76d1\u7763\u5fae\u8c03\uff08SFT\uff09\u6765\u5feb\u901f\u5e2e\u52a9 LLMs \u9002\u5e94\u5230\u6570\u5b66\u9886\u57df\u3002\u5728\u63a8\u7406\u9636\u6bb5\uff0c\u901a\u8fc7\u8bbe\u8ba1\u5de7\u5999\u7684\u8f93\u5165\u63d0\u793a\uff08\u4f8b\u5982\uff0cChain-of-Thought \u548c Program-of-Thought\uff09\u4e5f\u53ef\u4ee5\u6709\u6548\u6fc0\u53d1 LLMs \u7684\u6570\u5b66\u80fd\u529b\u3002<\/section>\n<section><\/section>\n<section>\u5bf9\u4e8e\u5927\u591a\u6570 LLMs \u800c\u8a00\uff0c\u9762\u5bf9\u9ad8\u4e2d\u53ca\u4ee5\u4e0a\u7684\u6570\u5b66\u95ee\u9898\u4ecd\u6709\u5f88\u5927\u7684\u63d0\u5347\u7a7a\u95f4\u3002\u7136\u800c\uff0c\u5728\u5c0f\u5b66\u6570\u5b66\u9886\u57df\uff0cLLMs \u5df2\u7ecf\u5c55\u73b0\u51fa\u5de8\u5927\u7684\u6f5c\u529b\u3002<strong>\u8fd9\u8ba9\u6211\u4eec\u4e0d\u7981\u601d\u8003\uff0c<\/strong><strong>\u5728\u73b0\u5b9e\u73af\u5883\u4e2d LLMs \u662f\u5426\u80fd\u4f9d\u7136\u4fdd\u6301\u9ad8\u6027\u80fd\uff1f<\/strong><\/section>\n<section><\/section>\n<section><strong>\u5bf9\u6297\u6027\u8bc4\u6d4b\u6570\u636e\u96c6 GSM-Plus<\/strong><\/section>\n<section><\/section>\n<section>\u672c\u7814\u7a76\u65e8\u5728\u63a8\u51fa\u4e00\u4e2a\u7efc\u5408\u6027<mark data-type=\"concepts\" data-id=\"308c3a45-0fee-4ec6-858e-85b15f440fc0\">\u57fa\u51c6<\/mark>\u6d4b\u8bd5 GSM-Plus\uff0c\u4ee5\u7cfb\u7edf\u5730\u68c0\u9a8c LLMs \u5728\u89e3\u51b3\u57fa\u7840\u6570\u5b66\u95ee\u9898\u65f6\u7684\u9c81\u68d2\u6027\u3002\u53d7 Polya \u539f\u5219 [2] \u4e2d\u89e3\u51b3\u6570\u5b66\u95ee\u9898\u7684\u80fd\u529b\u5206\u7c7b\u6cd5\u7684\u542f\u53d1\uff0c\u672c\u6587\u786e\u5b9a\u4e86\u4e94\u4e2a\u65b9\u9762\u7684\u6307\u5bfc\u539f\u5219\u7528\u4e8e\u6784\u5efa GSM-Plus \u6570\u636e\u96c6\uff1a<\/section>\n<section><\/section>\n<section>\u4e3a\u4e86\u4fbf\u4e8e\u7406\u89e3\uff0c\u6b64\u5904\u4ee5\u300c \u73cd\u59ae\u7279\u7684\u9e2d\u5b50\u6bcf\u5929\u4e0b 16 \u4e2a\u86cb\u3002\u5979\u6bcf\u5929\u65e9\u4e0a\u5403\u4e09\u4e2a\u86cb\u4f5c\u4e3a\u65e9\u9910\uff0c\u5e76\u4e14\u7528\u56db\u4e2a\u86cb\u70e4\u677e\u997c\u7ed9\u5979\u7684\u670b\u53cb\u3002\u5979\u6bcf\u5929\u4ee5\u6bcf\u4e2a\u9e2d\u86cb 2 \u7f8e\u5143\u7684\u4ef7\u683c\u5728\u519c\u8d38\u5e02\u573a\u51fa\u552e\u5269\u4f59\u7684\u86cb\u3002\u5979\u6bcf\u5929\u5728\u519c\u8d38\u5e02\u573a\u4e0a\u8d5a\u591a\u5c11\u7f8e\u5143\uff1f\u300d\u95ee\u9898\u4e3a\u4f8b\u3002<\/section>\n<section><\/section>\n<section><strong>\uff081\uff09<\/strong><strong>\u6570\u503c\u53d8\u5316<\/strong>\uff1a\u6307\u6539\u53d8\u6570\u503c\u6570\u636e\u6216\u5176\u7c7b\u578b\uff0c\u672c\u6587\u5b9a\u4e49\u4e86\u4e09\u4e2a\u5b50\u7c7b\u522b\uff1a<\/section>\n<section><\/section>\n<ul>\n<li>\n<section>\u6570\u503c\u66ff\u6362\uff1a\u5728\u540c\u7b49\u6570\u4f4d\u548c\u7c7b\u578b\u4e0b\u66ff\u6362\u6570\u503c\uff0c\u4f8b\u5982\u5c06\u95ee\u9898\u4e2d\u7684 \u201c16\u201d \u66ff\u6362\u4e3a \u201c20\u201d\u3002<\/section>\n<\/li>\n<li>\n<section>\u6570\u4f4d\u6269\u5c55\uff1a\u589e\u52a0\u6570\u503c\u7684\u4f4d\u6570\uff0c\u4f8b\u5982\u5c06 \u201c16\u201d \u66ff\u6362\u4e3a \u201c1600\u201d\u3002<\/section>\n<\/li>\n<li>\n<section>\u6574\u6570 &#8211; \u5c0f\u6570 &#8211; \u5206\u6570\u8f6c\u6362\uff1a\u5c06\u6574\u6570\u66f4\u6362\u4e3a\u5c0f\u6570\u6216\u5206\u6570\uff0c\u4f8b\u5982\u5c06 \u201c2\u201d \u8f6c\u6362\u4e3a \u201c2.5\u201d\u3002<\/section>\n<\/li>\n<\/ul>\n<section><\/section>\n<section><strong>\uff082\uff09<mark data-type=\"tech_tasks\" data-id=\"c1ae9de3-3159-4e1a-8bff-4ec534f247a7\">\u7b97\u672f<\/mark>\u53d8\u5316<\/strong>\uff1a\u6307\u5bf9\u6570\u5b66\u95ee\u9898\u5f15\u5165\u989d\u5916\u7684\u8fd0\u7b97\u6216\u8005\u8fdb\u884c\u53cd\u8f6c\uff0c\u4f46\u53ea\u9650\u4e8e\u52a0\u3001\u51cf\u3001\u4e58\u3001\u9664\u8fd0\u7b97\uff1a<\/section>\n<section><\/section>\n<ul>\n<li>\n<section>\u8fd0\u7b97\u6269\u5145\uff1a\u5728\u539f\u95ee\u9898\u57fa\u7840\u4e0a\u589e\u52a0\u9650\u5236\u6761\u4ef6\u3002\u4f8b\u5982\uff0c\u589e\u52a0\u65b0\u6761\u4ef6\u201c\u5979\u6bcf\u5929\u8fd8\u4f1a\u4f7f\u7528\u4e24\u4e2a\u9e21\u86cb\u81ea\u5236\u53d1\u819c\u201d\u3002<\/section>\n<\/li>\n<li>\n<section>\u8fd0\u7b97\u9006\u8f6c\uff1a\u5c06\u539f\u95ee\u9898\u7684\u67d0\u4e2a\u5df2\u77e5\u6761\u4ef6\u8f6c\u6362\u4e3a GSM-Plus \u53d8\u4f53\u95ee\u9898\u7684\u5f85\u6c42\u89e3\u53d8\u91cf\u3002\u4f8b\u5982\uff0c\u56fe 2 \u4e2d\u539f\u95ee\u9898\u7684\u9648\u8ff0 \u201c\u6bcf\u4e2a\u9e2d\u86cb 2 \u7f8e\u5143\u201d \u8f6c\u6362\u4e3a\u65b0\u95ee\u9898\u7684\u7591\u95ee\u53e5 \u201c\u6bcf\u4e2a\u9e2d\u86cb\u7684\u4ef7\u683c\u662f\u591a\u5c11\uff1f\u201d\uff0c\u800c\u539f\u95ee\u9898\u7591\u95ee\u53e5\u201d \u6bcf\u5929\u5728\u519c\u8d38\u5e02\u573a\u4e0a\u8d5a\u591a\u5c11\u7f8e\u5143\uff1f\u201d \u5219\u8f6c\u6362\u4e3a\u65b0\u95ee\u9898\u7684\u5df2\u77e5\u6761\u4ef6\u201d \u5979\u6bcf\u5929\u5728\u519c\u8d38\u5e02\u573a\u8d5a 18 \u7f8e\u5143\u201d<\/section>\n<\/li>\n<\/ul>\n<section><\/section>\n<section><strong>\uff083\uff09\u95ee\u9898\u7406\u89e3<\/strong>\uff1a\u6307\u5728\u610f\u601d\u4e0d\u53d8\u7684\u524d\u63d0\u4e0b\uff0c\u7528\u4e0d\u540c\u8bcd\u53e5\u91cd\u65b0\u8868\u8ff0\u6570\u5b66\u95ee\u9898\uff0c\u5982\u201d \u73cd\u59ae\u7279\u517b\u4e86\u4e00\u7fa4\u9e2d\u5b50\uff0c\u8fd9\u4e9b\u9e2d\u5b50\u6bcf\u5929\u4ea7 16 \u4e2a\u9e2d\u86cb\u3002\u5979\u65e9\u9910\u6d88\u8017\u4e09\u4e2a\u9e2d\u86cb\uff0c\u7136\u540e\u6d88\u8017\u56db\u4e2a\u9e2d\u86cb\u70e4\u677e\u997c\u7ed9\u5979\u7684\u670b\u53cb\u3002\u73cd\u59ae\u7279\u5728\u519c\u8d38\u5e02\u573a\u4e0a\u4ee5\u6bcf\u4e2a\u65b0\u9c9c\u7684\u9e2d\u86cb 2 \u7f8e\u5143\u7684\u4ef7\u683c\u5c06\u5269\u4f59\u7684\u9e2d\u86cb\u5168\u90e8\u51fa\u552e\u3002\u5979\u6bcf\u5929\u901a\u8fc7\u5728\u519c\u8d38\u5e02\u573a\u51fa\u552e\u9e2d\u86cb\u8d5a\u591a\u5c11\u94b1\uff1f\u201d<\/section>\n<section><\/section>\n<section><strong>\uff084\uff09\u5e72\u6270\u9879\u63d2\u5165<\/strong>\uff1a\u6307\u5c06\u4e0e\u4e3b\u9898\u76f8\u5173\u3001\u5305\u542b\u6570\u503c\u4f46\u5bf9\u6c42\u89e3\u65e0\u7528\u7684\u53e5\u5b50\u63d2\u5165\u5230\u539f\u95ee\u9898\u4e2d\uff0c\u5982\u201d \u73cd\u59ae\u7279\u8fd8\u60f3\u7528\u4e24\u4e2a\u9e2d\u86cb\u5582\u517b\u5979\u7684\u5ba0\u7269\u9e66\u9e49\uff0c\u6240\u5e78\u5979\u7684\u90bb\u5c45\u6bcf\u5929\u9001\u5979\u4e24\u4e2a\u9e2d\u86cb\u7528\u4e8e\u5582\u517b\u9e66\u9e49\u201d\u3002<\/section>\n<section><\/section>\n<section><strong>\uff085\uff09\u6279\u5224\u6027\u601d\u7ef4<\/strong>\uff1a\u4fa7\u91cd\u4e8e\u5f53\u6570\u5b66\u95ee\u9898\u7f3a\u4e4f\u5fc5\u8981\u6761\u4ef6\u65f6\uff0cLLMs \u662f\u5426\u5177\u6709\u63d0\u95ee\u6216\u6000\u7591\u80fd\u529b\uff0c\u4f8b\u5982\u201d \u73cd\u59ae\u7279\u7684\u9e2d\u5b50\u6bcf\u5929\u90fd\u4f1a\u4e0b\u86cb\u3002\u5979\u6bcf\u5929\u65e9\u4e0a\u5403\u4e09\u4e2a\u86cb\u4f5c\u4e3a\u65e9\u9910\uff0c\u5e76\u4e14\u6bcf\u5929\u7528\u56db\u4e2a\u86cb\u70e4\u677e\u997c\u7ed9\u5979\u7684\u670b\u53cb\u3002\u5979\u6bcf\u5929\u4ee5\u6bcf\u4e2a\u9e2d\u86cb 2 \u7f8e\u5143\u7684\u4ef7\u683c\u5728\u519c\u8d38\u5e02\u573a\u51fa\u552e\u5269\u4f59\u7684\u86cb\u3002\u5979\u6bcf\u5929\u5728\u519c\u8d38\u5e02\u573a\u4e0a\u8d5a\u591a\u5c11\u7f8e\u5143\uff1f\u201d\u3002<\/section>\n<section><\/section>\n<section>\u57fa\u4e8e GSM8K \u7684 1,319 \u4e2a\u6d4b\u8bd5\u95ee\u9898\uff0c\u672c\u6587\u4e3a\u6bcf\u4e2a\u95ee\u9898\u521b\u5efa\u4e86\u516b\u4e2a\u53d8\u4f53\uff0c\u4ece\u800c\u751f\u6210\u4e86\u5305\u542b 10,552 \u4e2a\u95ee\u9898\u53d8\u4f53\u7684 GSM-Plus \u6570\u636e\u96c6\uff08\u672c\u6587\u8fd8\u63d0\u4f9b\u4e86\u4e00\u4e2a\u5305\u542b 2,400 \u4e2a\u95ee\u9898\u53d8\u4f53\u7684\u6d4b\u8bd5\u5b50\u96c6\uff0c\u4ee5\u4fbf\u5feb\u901f\u8bc4\u6d4b\uff09\u3002\u901a\u8fc7\u4f7f\u7528\u6bcf\u4e2a\u95ee\u9898\u53ca\u5176\u516b\u4e2a\u53d8\u4f53\u6d4b\u8bd5 LLMs\uff0cGSM-Plus \u53ef\u4ee5\u5e2e\u52a9\u7814\u7a76\u4eba\u5458\u5168\u9762\u8bc4\u4f30 LLMs \u5728\u89e3\u51b3\u6570\u5b66\u95ee\u9898\u4e2d\u7684\u9c81\u68d2\u6027\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-8fa6c16b53a79e28d39a7bc5a7f8ffdd.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-8fa6c16b53a79e28d39a7bc5a7f8ffdd.png\"><\/a><\/p>\n<section><em><sup>\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u56fe 2\uff1a\u57fa\u4e8e\u4e00\u4e2a\u79cd\u5b50\u6570\u5b66\u9898\uff0c\u4f7f\u7528 5 \u4e2a\u89d2\u5ea6\u7684 8 \u79cd\u6270\u52a8\u751f\u6210\u95ee\u9898\u53d8\u4f53\u3002\u4e3b\u8981\u4fee\u6539\u5185\u5bb9\u4ee5\u7eff\u8272\u6807\u51fa\u3002<\/sup><\/em><\/section>\n<section><\/section>\n<section>\u901a\u8fc7\u4f7f\u7528 GSM-Plus \u8bc4\u4f30 25 \u4e2a\u4e0d\u540c\u89c4\u6a21\u3001\u4e0d\u540c\u9884\u8bad\u7ec3\u65b9\u5f0f\u3001\u4e0d\u540c\u4efb\u52a1\u5fae\u8c03\u7684 LLMs\uff0c\u4ee5\u53ca\u7ec4\u5408 4 \u79cd\u5e38\u7528\u7684\u63d0\u793a\u6280\u672f\uff0c\u672c\u6587\u53d1\u73b0 LLMs \u6574\u4f53\u4e0a\u53ef\u4ee5\u51c6\u786e\u89e3\u51b3 GSM8K \u95ee\u9898\uff0c\u4f46\u5728\u56de\u7b54 GSM-Plus \u4e2d\u7684\u53d8\u4f53\u95ee\u9898\u65f6\u4f1a\u9047\u5230\u660e\u663e\u56f0\u96be\u3002\u4e3b\u8981\u53d1\u73b0\u5982\u4e0b\uff1a<\/section>\n<section><\/section>\n<ul>\n<li>\n<section>\u4efb\u52a1\u7279\u5b9a\u7684\u4f18\u5316\uff0c\u5373\u5728\u6570\u5b66\u76f8\u5173\u7684\u6570\u636e\u96c6\u4e0a\u5fae\u8c03\uff0c\u901a\u5e38\u53ef\u4ee5\u63d0\u9ad8\u4e0b\u6e38\u4efb\u52a1\u51c6\u786e\u6027\uff1b\u800c\u9c81\u68d2\u6027\u7684\u9ad8\u4f4e\u66f4\u591a\u5730\u53d6\u51b3\u4e8e\u57fa\u7840\u6a21\u578b\u548c\u5fae\u8c03\u6570\u636e\u96c6\u7684\u9009\u62e9\u3002<\/section>\n<\/li>\n<li>\n<section>\u5f53\u9700\u8981 \u201c\u6279\u5224\u6027\u601d\u7ef4\u201d\u3001\u6d89\u53ca \u201c<mark data-type=\"tech_tasks\" data-id=\"c1ae9de3-3159-4e1a-8bff-4ec534f247a7\">\u7b97\u672f<\/mark>\u53d8\u5316\u201d \u548c \u201c\u5e72\u6270\u56e0\u7d20\u63d2\u5165\u201d \u65f6\uff0cLLMs \u7684\u6027\u80fd\u4f1a\u8fc5\u901f\u4e0b\u964d\uff1b\u4f46\u5bf9\u4e8e \u201c\u6570\u503c\u53d8\u5316\u201d \u548c \u201c\u95ee\u9898\u7406\u89e3\u201d \u7684\u6270\u52a8\uff0cLLMs \u7684\u6027\u80fd\u6bd4\u8f83\u7a33\u5b9a\u3002<\/section>\n<\/li>\n<li>\n<section>\u5148\u524d\u7684\u63d0\u793a\u6280\u672f\uff08\u4f8b\u5982\uff0cCoT\uff0cPoT\uff0cLtM \u548c Complexity-based CoT\uff09\u5bf9\u4e8e\u9c81\u68d2\u6027\u589e\u5f3a\u4f5c\u7528\u4e0d\u663e\u8457\uff0c\u7279\u522b\u662f\u5bf9\u4e8e \u201c<mark data-type=\"tech_tasks\" data-id=\"c1ae9de3-3159-4e1a-8bff-4ec534f247a7\">\u7b97\u672f<\/mark>\u53d8\u5316 \u201c\u548c\u201d \u6279\u5224\u6027\u601d\u7ef4\u201d\u3002\u5728\u524d\u4eba\u5de5\u4f5c\u7684\u57fa\u7840\u4e0a\uff0c\u672c\u6587\u8fdb\u4e00\u6b65\u63a2\u7d22\u4e86\u4e00\u79cd\u7ec4\u5408\u63d0\u793a\u65b9\u6cd5\uff0c\u901a\u8fc7\u8fed\u4ee3\u751f\u6210\u548c\u9a8c\u8bc1\u6bcf\u4e2a\u63a8\u7406\u601d\u7ef4\uff0c\u53ef\u4ee5\u540c\u65f6\u63d0\u5347 LLMs \u5728 GSM8K \u548c GSM-Plus \u4e0a\u7684\u6027\u80fd\u3002 \u00a0<\/section>\n<\/li>\n<\/ul>\n<section><\/section>\n<section><strong>GSM-Plus \u7279\u70b9<\/strong><\/section>\n<section><\/section>\n<ol>\n<li>\n<section><strong>\u8d28\u91cf\u4fdd\u8bc1<\/strong>\uff1a\u91c7\u7528\u4e24\u9636\u6bb5\u751f\u6210 GSM-Plus \u8bc4\u6d4b\u9898\u3002\u9996\u5148\uff0c\u5229\u7528 GPT-4 \u7684\u95ee\u9898\u6539\u5199\u80fd\u529b\u751f\u6210\u95ee\u9898\u53d8\u4f53\uff0c\u7136\u540e\u4e3a\u8fd9\u4e9b\u53d8\u4f53\u751f\u6210\u5019\u9009\u7b54\u6848\uff1b\u4e3a\u786e\u4fdd\u6570\u636e\u8d28\u91cf\uff0c\u6240\u6709\u7531 GPT-4 \u751f\u6210\u7684\u95ee\u9898\u53d8\u4f53\u548c\u7b54\u6848\u90fd\u8981\u7ecf\u8fc7\u4eba\u5de5\u6807\u6ce8\u56e2\u961f\u8fdb\u884c\u4e25\u683c\u68c0\u67e5\u3002\u4eba\u5de5\u6807\u6ce8\u56e2\u961f\u4fee\u6b63\u4e86 18.85% \u7684 GPT-4 \u6539\u5199\u7684\u95ee\u9898\u3002<\/section>\n<\/li>\n<li>\n<section><strong>\u7ec6\u7c92\u5ea6\u8bc4\u4f30<\/strong>\uff1a\u5bf9\u4e8e\u4e3b\u6d41\u8bc4\u6d4b\u6570\u636e\u96c6 GSM8K \u7684\u6bcf\u4e2a\u6d4b\u8bd5\u9898\uff0cGSM-Plus \u63d0\u4f9b\u4e86 8 \u4e2a\u6270\u52a8\u65b9\u5411\u7684\u53d8\u4f53\u95ee\u9898\uff0c\u5145\u5206\u6d4b\u8bd5\u4e86\u5728\u4e0d\u540c\u4e0a\u4e0b\u6587\u4e0b\uff0c\u5927\u6a21\u578b\u7075\u6d3b\u89e3\u51b3\u6570\u5b66\u5e94\u7528\u9898\u7684\u80fd\u529b\u3002<\/section>\n<\/li>\n<li>\n<section><strong>\u6311\u6218\u6027<\/strong>\uff1a\u76f8\u6bd4\u4e8e GSM8K\uff0cGSM-Plus \u7684\u95ee\u9898\u53d8\u4f53\u66f4\u5177\u6311\u6218\u6027\uff0c\u6240\u6709\u53c2\u4e0e\u8bc4\u4f30\u7684 LLMs \u7684\u6027\u80fd\u90fd\u663e\u8457\u4e0b\u964d\u3002\u5728\u63a5\u4e0b\u6765\u7684\u5206\u6790\u4e2d\uff0c\u672c\u6587\u4f1a\u7279\u522b\u5206\u6790 LLMs \u5728\u4e0d\u540c\u7c7b\u578b\u6270\u52a8\u4e0b\u7684\u89e3\u9898\u9c81\u68d2\u6027\u3002<\/section>\n<\/li>\n<\/ol>\n<section><\/section>\n<section><strong>\u4e0e\u5176\u4ed6\u5c0f\u5b66\u6570\u5b66\u5e94\u7528\u9898\u6570\u636e\u7684\u6bd4\u8f83<\/strong><\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-a1c398b2ea4d31f397e39e2e1d36fa4a.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-a1c398b2ea4d31f397e39e2e1d36fa4a.png\"><\/a><\/p>\n<section><em><sup>\u8868 1\uff1a\u4e0d\u540c\u989c\u8272\u4ee3\u8868\u4e0d\u540c\u7684\u6270\u52a8\u7c7b\u578b\uff1a<\/sup><\/em><sup><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-ebf32256c9b72b27bf24d66b96d02f48.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-ebf32256c9b72b27bf24d66b96d02f48.png\"><\/a><em>\u6570\u503c\u66ff\u6362\uff0c<\/em><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-135783144685b72d39109356f455d9d7.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-135783144685b72d39109356f455d9d7.png\"><\/a><em>\u6570\u4f4d\u6269\u5c55\uff0c<\/em><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-daa5110198e30ac6c968bc7553342b34.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-daa5110198e30ac6c968bc7553342b34.png\"><\/a><em>\u6574\u6570 &#8211; \u5c0f\u6570 &#8211; \u5206\u6570\u8f6c\u6362\uff0c<\/em><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-e226cfb485851fe639fb998e258a035f.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-e226cfb485851fe639fb998e258a035f.png\"><\/a><em>\u8fd0\u7b97\u6269\u5145\uff0c<\/em><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-c3c6aae354ced2b7a3d9c8fd384dfc0c.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-c3c6aae354ced2b7a3d9c8fd384dfc0c.png\"><\/a><em>\u8fd0\u7b97\u9006\u8f6c\uff0c<\/em><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-320b0271057b28abb80483fd44cec5da.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-320b0271057b28abb80483fd44cec5da.png\"><\/a><em>\u95ee\u9898\u7406\u89e3\uff0c<\/em><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-215e7a96acd97662fc334f7da4dda6ec.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-215e7a96acd97662fc334f7da4dda6ec.png\"><\/a><em>\u5e72\u6270\u9879\u63d2\u5165\uff0c<\/em><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-8d7fd07c612f9cbbafd440a873865aff.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-8d7fd07c612f9cbbafd440a873865aff.png\"><\/a><\/sup><em><sup>\u6279\u5224\u6027\u601d\u7ef4\u3002<\/sup><\/em><\/section>\n<section><\/section>\n<section>\u4ece\u4e0a\u8868\u53ef\u4ee5\u770b\u51fa\uff0c\u5148\u524d\u7684\u7814\u7a76\u4f7f\u7528\u4e0d\u540c\u7684\u6270\u52a8\u6765\u68c0\u9a8c\u6570\u5b66\u63a8\u7406\u7684\u9c81\u68d2\u6027\uff0c\u4f46\u662f\u8bc4\u4f30\u8bbe\u7f6e\u4ec5\u6db5\u76d6\u90e8\u5206\u6270\u52a8\u7c7b\u578b\uff0c\u4e14\u5927\u591a\u662f\u901a\u8fc7\u81ea\u52a8\u65b9\u6cd5\u6784\u5efa\u5f15\u5165\u6270\u52a8\uff0c\u8d28\u91cf\u96be\u4ee5\u4fdd\u8bc1\u3002\u76f8\u6bd4\u4e4b\u4e0b\uff0cGSM-Plus \u4f7f\u7528\u516b\u79cd\u4e0d\u540c\u7684\u6570\u5b66\u63a8\u7406\u6280\u80fd\u5bf9\u5355\u4e00\u95ee\u9898\u8fdb\u884c\u6270\u52a8\uff0c\u8986\u76d6\u9762\u66f4\u5168\uff0c\u4e14\u7ecf\u8fc7\u4e25\u683c\u7684\u8d28\u91cf\u63a7\u5236\u3002<\/section>\n<section><\/section>\n<section><strong>\u5b9e\u9a8c\u5206\u6790<\/strong><\/section>\n<section><\/section>\n<section><strong>\u8bc4\u6d4b\u6307\u6807<\/strong><\/section>\n<section><\/section>\n<ul>\n<li>\n<section><strong>\u6027\u80fd\u4e0b\u964d\u7387\uff08PDR\uff09<\/strong>\uff1a\u4e0e\u539f\u95ee\u9898\u76f8\u6bd4\uff0cLLMs \u5728\u6270\u52a8\u540e\u7684\u95ee\u9898\u4e0a\u7684\u6027\u80fd\u4e0b\u964d\u7a0b\u5ea6\u3002<\/section>\n<\/li>\n<li>\n<section><strong>\u540c\u65f6\u89e3\u51b3\u7684\u95ee\u9898\u5bf9\u7684\u767e\u5206\u6bd4\uff08ASP\uff09<\/strong>\uff1a\u539f\u95ee\u9898\u53ca\u5176\u5bf9\u5e94\u7684\u67d0\u4e2a\u95ee\u9898\u53d8\u4f53\u5747\u88ab LLMs \u6b63\u786e\u89e3\u7b54\u7684\u6bd4\u4f8b\u3002<\/section>\n<\/li>\n<\/ul>\n<section><\/section>\n<section><strong>\u6574\u4f53\u6027\u80fd<\/strong><\/section>\n<section><\/section>\n<section>\u5982\u4e0b\u8868\u6240\u793a\uff0c\u76f8\u8f83\u4e8e GSM8K\uff0c\u5927\u591a\u6570 LLMs \u5728 GSM-Plus \u4e0a\u7684\u6027\u80fd\u90fd\u5927\u5e45\u4e0b\u964d\u3002\u00a0<\/section>\n<section><\/section>\n<section>GPT-4 \u8868\u73b0\u51fa\u6700\u9ad8\u7684\u9c81\u68d2\u6027\uff0c\u5176 PDR \u6700\u5c0f\u4ec5\u4e3a 8.23\uff05\u3002\u800c\u00a0CodeLlama \u7684 PDR \u6700\u5927\uff0c\u5176\u4e2d 7B\u300113B \u548c 34B \u7684\u6a21\u578b\u5206\u522b\u4e3a 40.56\uff05\u300139.71\uff05\u548c 34.27\uff05\uff0c\u8d85\u8fc7\u4e86\u5176\u57fa\u5ea7\u6a21\u578b LLaMA-2-7B\uff0839.49\uff05\uff09\uff0c\u4ee5\u53ca\u5728\u5176\u4e0a\u5fae\u8c03\u7684\u6570\u5b66 SFT \u6a21\u578b\uff0c\u5982 SEGO-7B\uff0834.91\uff05\uff09\u3002\u8fd9\u8868\u660e\u4ec5\u4f7f\u7528\u7a0b\u5e8f\u8bed\u8a00\u63a8\u7406\u5bf9\u4e8e\u6270\u52a8\u662f\u5f88\u8106\u5f31\u7684\u3002<\/section>\n<section><\/section>\n<section>\u5728\u9762\u5bf9\u6570\u5b66\u6270\u52a8\u65f6\uff0c\u6a21\u578b\u89c4\u6a21\u8d8a\u5927\uff0c\u6027\u80fd\u8d8a\u7a33\u5b9a\u3002\u867d\u7136\u76d1\u7763\u5fae\u8c03\u53ef\u4ee5\u63d0\u9ad8\u5728\u4e0b\u6e38\u4efb\u52a1\u4e0a\u7684<mark data-type=\"concepts\" data-id=\"8be77eae-12da-4e9e-9a88-b7f5bae98c2e\">\u51c6\u786e\u7387<\/mark>\uff0c\u4f46\u5e76\u4e0d\u80fd\u663e\u8457\u589e\u5f3a\u6a21\u578b\u5bf9\u4e8e\u6270\u52a8\u7684\u9c81\u68d2\u6027\uff08\u5373\u66f4\u4f4e\u7684 PDR\uff09\u3002\u76d1\u7763\u5fae\u8c03\u7684\u6570\u636e\u5bf9\u4e8e\u9c81\u68d2\u6027\u975e\u5e38\u91cd\u8981\u3002\u540c\u6837\u662f\u57fa\u4e8e LLaMA-2 \u8fdb\u884c\u5fae\u8c03\uff0c\u4f7f\u7528\u4e0d\u540c\u7684\u6570\u636e\uff0c\u4f1a\u5bfc\u81f4\u6a21\u578b\u7684<mark data-type=\"concepts\" data-id=\"8be77eae-12da-4e9e-9a88-b7f5bae98c2e\">\u51c6\u786e\u7387<\/mark>\u548c\u9c81\u68d2\u6027\u5177\u6709\u8f83\u5927\u5dee\u5f02\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-ed1f6d3409044d56ba698cec82e6f3af.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-ed1f6d3409044d56ba698cec82e6f3af.png\"><\/a><\/p>\n<section><em><sup>\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u8868 2\uff1a\u6574\u4f53\u6027\u80fd<\/sup><\/em><\/section>\n<section><\/section>\n<section><strong>\u7ec6\u7c92\u5ea6\u5b9e\u9a8c\u5206\u6790<\/strong><\/section>\n<section><\/section>\n<section>\u4e0d\u540c\u6270\u52a8\u4e0b LLMs \u7684\u6027\u80fd\u8868\u73b0<\/section>\n<section><\/section>\n<section>\u672c\u6587\u8fdb\u4e00\u6b65\u8bc4\u4f30\u4e86 LLMs \u5728 8 \u79cd\u95ee\u9898\u53d8\u4f53\u4e0b\u7684\u6027\u80fd\u7a33\u5b9a\u6027\u3002\u4e0e\u4eba\u7c7b\u57fa\u7ebf\u76f8\u6bd4\uff0c\u5bf9\u4e8e \u201c\u6279\u5224\u6027\u601d\u7ef4\u201d\uff08\u7d2b\u8272\uff09\u3001\u201c\u8fd0\u7b97\u6269\u5145\u201d \u548c \u201c\u8fd0\u7b97\u9006\u8f6c\u201d\uff08\u84dd\u8272\uff09\u3001\u201c\u5e72\u6270\u9879\u63d2\u5165\u201d\uff08\u7c89\u8272\uff09\u4ee5\u53ca \u201c\u6574\u6570 &#8211; \u5c0f\u6570 &#8211; \u5206\u6570\u8f6c\u6362\u201d\uff08\u6a59\u8272\uff09\u6270\u52a8\uff0cLLMs \u6027\u80fd\u4e0b\u964d\u660e\u663e\u3002\u800c\u5bf9\u4e8e \u201c\u6570\u503c\u66ff\u6362\u201d \u548c \u201c\u95ee\u9898\u7406\u89e3\u201d\uff0cLLMs \u7684\u6027\u80fd\u7a33\u5b9a\uff0c\u751a\u81f3\u6709\u8f7b\u5fae\u7684\u63d0\u5347\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-5e9b1c9020472cd3e9ad310611eb1626.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-5e9b1c9020472cd3e9ad310611eb1626.png\"><\/a><\/p>\n<section><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\u56fe 3\uff1a\u7ec6\u7c92\u5ea6\u5b9e\u9a8c\u5206\u6790<\/sup><\/em><\/section>\n<section><\/section>\n<section>\u6570\u5b66\u63a8\u7406\u80fd\u529b\u7684\u8fc1\u79fb\u6027<\/section>\n<section><\/section>\n<section>\u524d\u9762\u7684\u5206\u6790\u4e3b\u8981\u57fa\u4e8e\u6570\u636e\u96c6\u6574\u4f53\u3002\u63a5\u4e0b\u6765\uff0c\u672c\u6587\u6839\u636e\u6570\u5b66\u9898\u662f\u5426\u88ab\u6b63\u786e\u56de\u7b54\u5c06 2 \u4e2a\u6570\u636e\u96c6\u5206\u5272\uff0c\u5206\u6790\u5f53 LLMs \u6210\u529f\u89e3\u51b3 GSM8K \u95ee\u9898\u65f6\uff0c\u662f\u5426\u610f\u5473\u7740\u6b63\u786e\u56de\u7b54 GSM-Plus \u53d8\u4f53\u95ee\u9898\u7684\u53ef\u80fd\u6027\u53d8\u9ad8\uff08\u5373\u9ad8 ASP \u503c\uff09\uff0c\u53cd\u4e4b\u4ea6\u7136\u3002\u5982\u679c\u8fd9\u79cd\u65ad\u8a00\u6210\u7acb\uff0c\u53ef\u4ee5\u8ba4\u4e3a LLMs \u5728\u8fd9\u7c7b\u7279\u5b9a\u7684\u6570\u5b66\u9898\u5b50\u96c6\u4e0a\u6027\u80fd\u7a33\u5b9a\uff0c\u5373\u4f7f\u5728\u6574\u4e2a\u6570\u636e\u96c6\u4e0a\u5e76\u975e\u5982\u6b64\u3002\u5728\u5b9e\u9a8c\u8bbe\u7f6e\u4e2d\uff0c\u6bcf\u4e2a GSM8K \u95ee\u9898\u53ca\u5176\u5728 GSM-Plus \u4e2d\u7684\u53d8\u4f53\u8f6c\u5316\u4e3a 8 \u4e2a\u95ee\u9898\u5bf9\uff0c\u7ed3\u679c\u5982\u56fe 4 \u6240\u793a\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-0a0740f2067e0af7a6f9b8f7f28cd2a1.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-0a0740f2067e0af7a6f9b8f7f28cd2a1.png\"><\/a><\/p>\n<section><em><sup>\u56fe 4\uff1aLLMs \u5728 GSM8K \u548c GSM-Plus \u95ee\u9898\u5bf9\u4e4b\u95f4\u7684\u63a8\u7406\u53ef\u8fc1\u79fb\u6027\u3002\u7d2b\u8272\uff08\u5747\u6b63\u786e\uff09\u548c\u84dd\u8272\uff08\u5747\u9519\u8bef\uff09\u7684\u6761\u5f62\u56fe\u8868\u793a\u4e00\u81f4\u7684\u6a21\u578b\u884c\u4e3a\uff0c\u800c\u7ea2\u8272\uff08GSM8K \u6b63\u786e &amp; GSM-Plus \u9519\u8bef\uff09\u548c\u9ec4\u8272\uff08GSM8K \u9519\u8bef &amp; GSM-Plus \u6b63\u786e\uff09\u7684\u6761\u5f62\u56fe\u5219\u8868\u793a\u4e0d\u4e00\u81f4\u7684\u6a21\u578b\u884c\u4e3a\u3002\u7d2b\u8272\u548c\u7ea2\u8272\u6761\u5f62\u56fe\u7684\u9ad8\u5ea6\u548c\u8868\u793a LLMs \u6b63\u786e\u89e3\u51b3 GSM8K \u95ee\u9898\u7684\u6570\u91cf\u3002<\/sup><\/em><\/section>\n<section><\/section>\n<section>\u7ea2\u8272\u6761\u5f62\u56fe\u7684\u5b58\u5728\uff08LLMs \u6b63\u786e\u56de\u7b54\u539f\u95ee\u9898\uff0c\u4f46\u672a\u89e3\u51b3\u53d8\u4f53\u95ee\u9898\uff09\uff0c\u8868\u660e\u5927\u591a\u6570\u6a21\u578b\u7684\u6027\u80fd\u53ef\u8fc1\u79fb\u6027\u6709\u9650\u3002\u867d\u7136 LLMs \u5728 GSM8K \u95ee\u9898\u4e0a\u6027\u80fd\u6709\u6240\u5dee\u5f02\uff08\u7d2b\u8272\u548c\u7ea2\u8272\u6761\u5f62\u56fe\u7684\u9ad8\u5ea6\uff09\uff0c\u4f46\u6027\u80fd\u53ef\u8fc1\u79fb\u6027\u76f8\u4f3c\uff08\u7ea2\u8272\u6761\u5f62\u56fe\u7684\u9ad8\u5ea6\uff09\u3002\u8fd9\u610f\u5473\u7740\u73b0\u6709\u7684<mark data-type=\"concepts\" data-id=\"308c3a45-0fee-4ec6-858e-85b15f440fc0\">\u57fa\u51c6<\/mark>\u6d4b\u8bd5\u65e0\u6cd5\u51c6\u786e\u8bc4\u4f30\u6a21\u578b\u5728\u6570\u5b66\u63a8\u7406\u65b9\u9762\u7684\u771f\u5b9e\u80fd\u529b\u3002\u9ad8<mark data-type=\"concepts\" data-id=\"8be77eae-12da-4e9e-9a88-b7f5bae98c2e\">\u51c6\u786e\u7387<\/mark>\u5e76\u4e0d\u7b49\u4ef7\u4e8e\u5f3a\u5927\u7684\u63a8\u7406\u9c81\u68d2\u6027\u3002<\/section>\n<section><\/section>\n<section>\u63d0\u793a\u5bf9\u4e8e LLMs \u6027\u80fd\u9c81\u68d2\u6027\u7684\u5e2e\u52a9<\/section>\n<section><\/section>\n<section>\u5148\u524d\u7684\u5de5\u4f5c\u8868\u660e\uff0c\u826f\u597d\u7684\u63d0\u793a\u6307\u4ee4\u5bf9\u4e8e\u6fc0\u53d1<mark data-type=\"tech_tasks\" data-id=\"bf35ef94-d956-4033-a533-0c0828308c36\">\u8bed\u8a00\u6a21\u578b<\/mark>\u7684\u6570\u5b66\u80fd\u529b\u5341\u5206\u91cd\u8981\u3002\u672c\u6587\u9009\u62e9\u4e86 4 \u4e2a\u4ee3\u8868\u6027\u6a21\u578b\uff0c\u5e76\u6d4b\u8bd5\u5b83\u4eec\u5728\u4e0d\u540c\u7684\u63d0\u793a\u6307\u4ee4\u4e0b\u89e3\u9898\u7684\u8868\u73b0\u3002\u5982\u4e0b\u56fe\u6240\u793a\uff0c\u5f53\u9762\u5bf9\u5e72\u6270\u65f6\uff0c\u4f7f\u7528\u590d\u6742\u7684\u793a\u4f8b\u4f5c\u4e3a\u4e0a\u4e0b\u6587\u6f14\u793a\uff08Complexity-based CoT\uff09\u65f6\uff0cLLMs \u8868\u73b0\u6700\u4e3a\u7a33\u5b9a\uff1b\u76f8\u6bd4\u4e4b\u4e0b\uff0c\u4ec5\u4f7f\u7528\u7a0b\u5e8f\u8bed\u8a00\u8868\u793a\u4e2d\u95f4\u63a8\u7406\uff08Program-of-Thought\uff09\u65f6\uff0cLLMs \u66f4\u5bb9\u6613\u53d7\u5230\u5e72\u6270\u7684\u5f71\u54cd\u3002\u603b\u4f53\u800c\u8a00\uff0c\u8fd9\u4e9b\u63d0\u793a\u6280\u5de7\u90fd\u4e0d\u8db3\u4ee5\u8ba9 LLMs \u5728 GSM-Plus \u4e0a\u7ef4\u6301\u4e0e GSM8K \u76f8\u540c\u7684\u6027\u80fd\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-f362d879e633d9d962300c830b5014d1.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-f362d879e633d9d962300c830b5014d1.png\"><\/a><\/p>\n<section><em><sup>\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u56fe 5\uff1a\u63d0\u793a\u5bf9\u4e8e LLMs \u6027\u80fd\u9c81\u68d2\u6027\u7684\u5f71\u54cd<\/sup><\/em><\/section>\n<section><\/section>\n<section>\u7ec4\u5408\u63d0\u793a\u662f\u5426\u6709\u6548\uff1f<\/section>\n<section><\/section>\n<section><strong>\u5982\u4f55\u57fa\u4e8e\u73b0\u6709\u7684\u63d0\u793a\u65b9\u6cd5\u589e\u5f3a LLMs \u7684\u9c81\u68d2\u6027\u5462\uff1f<\/strong>\u672c\u6587\u53d1\u73b0 LLMs \u5728\u89e3\u9898\u8fc7\u7a0b\u4e2d\u5e38\u5e38\u4f1a\u5ffd\u7565\u91cd\u8981\u6761\u4ef6\u6216\u51fa\u73b0\u8ba1\u7b97\u9519\u8bef\u3002\u4e3a\u6b64\uff0c\u672c\u6587\u63a2\u7d22\u4e86\u4e00\u79cd\u7ec4\u5408\u63d0\u793a\u65b9\u6cd5 Comp\u3002\u8be5\u65b9\u6cd5\u9996\u5148\u63d0\u793a LLMs \u63d0\u53d6\u95ee\u9898\u4e2d\u4e0e\u6570\u503c\u76f8\u5173\u7684\u5fc5\u8981\u6761\u4ef6\uff08Prompt1\uff09\u3002\u63a5\u7740\uff0c\u6839\u636e\u95ee\u9898\u548c\u5173\u952e\u6761\u4ef6\uff0c\u6307\u793a LLMs \u8fed\u4ee3\u5730\u751f\u6210\u63a8\u7406\u76ee\u6807\uff08Prompt2\uff09\u548c\u8ba1\u7b97\u76ee\u6807\uff08Prompt3\uff09\uff0c\u5e76\u8ba9\u5176\u4e3a\u751f\u6210\u7684\u5386\u53f2\u89e3\u9898\u6b65\u9aa4\u63d0\u4f9b\u53cd\u9988\uff0c\u4ee5\u786e\u5b9a\u662f\u5426\u83b7\u5f97\u4e86\u6700\u7ec8\u7b54\u6848\uff08Prompt4\uff09\u3002\u5177\u4f53\u5b9e\u73b0\u5982\u56fe 6 \u6240\u793a\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-cdb44910772a6cde852918304e9693f6.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-cdb44910772a6cde852918304e9693f6.png\"><\/a><\/p>\n<section><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 \u56fe 6\uff1aComp \u8fed\u4ee3\u63d0\u793a\u65b9\u5f0f\u7684\u793a\u610f\u56fe<\/sup><\/em><\/section>\n<section><\/section>\n<section>\u53ef\u4ee5\u770b\u51fa\uff0cComp \u901a\u8fc7\u8fed\u4ee3\u751f\u6210\u548c\u81ea\u6211\u9a8c\u8bc1\u53ef\u4ee5\u6539\u5584 LLMs \u5728\u5404\u79cd\u95ee\u9898\u53d8\u5316\u7c7b\u578b\u4e0b\u7684\u6027\u80fd\uff0c\u4f46\u5b83\u4ecd\u7136\u65e0\u6cd5\u5f25\u5408 LLMs \u5728\u6807\u51c6\u6d4b\u8bd5\u96c6\u548c\u5bf9\u6297\u6027\u6d4b\u8bd5\u96c6\u4e4b\u95f4\u7684\u6027\u80fd\u5dee\u8ddd\u3002\u8be5\u7814\u7a76\u671f\u5f85\u672a\u6765\u6709\u66f4\u591a\u7684\u65b9\u6cd5\u8fdb\u4e00\u6b65\u63d0\u5347\u6a21\u578b\u7684\u9c81\u68d2\u6027\uff0c\u63a8\u52a8 LLMs \u5728\u6570\u5b66\u63a8\u7406\u9886\u57df\u7684\u8fdb\u4e00\u6b65\u53d1\u5c55\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-783000735fa6bc1909003bbf6a34b172.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-783000735fa6bc1909003bbf6a34b172.png\"><\/a><\/p>\n<section><em><sup>\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0\u8868 3\uff1aComp \u8fed\u4ee3\u63d0\u793a\u7684\u6027\u80fd<\/sup><\/em><\/section>\n<section><\/section>\n<section>\u751f\u6210\u793a\u4f8b<\/section>\n<section><\/section>\n<section>\u4e0b\u56fe\u5c55\u793a\u4e86\u5728 GSM8K \u95ee\u9898\u548c\u57fa\u4e8e \u201c\u8fd0\u7b97\u9006\u8f6c\u201d \u7684 GSM-Plus \u6539\u5199\u95ee\u9898\u4e0a\uff0c\u4e0d\u540c\u63d0\u793a\u6280\u672f\u4e0b GPT-3.5-Turbo \u7684\u8868\u73b0\u3002\u867d\u7136\u6240\u6709\u63d0\u793a\u90fd\u53ef\u4ee5\u6fc0\u53d1 Turbo \u51c6\u786e\u56de\u7b54 GSM8K \u95ee\u9898\uff0c\u4f46\u53ea\u6709 Comp \u80fd\u591f\u5e2e\u52a9 Turbo \u5728 GSM-Plus \u53d8\u4f53\u95ee\u9898\u4e0a\u751f\u6210\u6b63\u786e\u7684\u7b54\u6848\u3002<\/section>\n<p><a href=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-15a225c54ff9c6448bcf2582208a7281.png\" data-fancybox=\"images\" data-fancybox=\"gallery\"><img decoding=\"async\" src=\"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-15a225c54ff9c6448bcf2582208a7281.png\"><\/a><\/p>\n<section><em><sup>\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u56fe 7\uff1a\u5728\u4e0d\u540c\u63d0\u793a\u8bbe\u7f6e\u4e0b\uff0c\u6a21\u578b\u56de\u7b54\u6570\u5b66\u95ee\u9898\u7684\u793a\u4f8b<\/sup><\/em><\/section>\n<section><\/section>\n<section><strong>\u7ed3\u8bed<\/strong><\/section>\n<section><\/section>\n<section>\u672c\u6587\u4ecb\u7ecd\u4e86\u4e00\u4e2a\u5bf9\u6297\u6027\u5c0f\u5b66\u6570\u5b66\u5e94\u7528\u9898\u8bc4\u6d4b\u96c6 GSM-Plus\uff0c\u65e8\u5728\u7cfb\u7edf\u5206\u6790 LLMs \u5728\u89e3\u51b3\u6570\u5b66\u5e94\u7528\u9898\u4e2d\u7684\u9c81\u68d2\u6027\u3002\u5b9e\u9a8c\u5206\u6790\u53d1\u73b0\uff0c\u5927\u591a\u6570 LLMs \u5728\u9762\u4e34\u6270\u52a8\u65f6\uff0c\u6027\u80fd\u76f8\u8f83\u4e8e\u5b83\u4eec\u5728\u6807\u51c6<mark data-type=\"concepts\" data-id=\"308c3a45-0fee-4ec6-858e-85b15f440fc0\">\u57fa\u51c6<\/mark>\u4e0a\u7684\u8868\u73b0\u663e\u8457\u4e0b\u964d\uff0c\u8fdc\u8fdc\u8fbe\u4e0d\u5230\u4eba\u7c7b\u7684\u8868\u73b0\u6c34\u5e73\u3002\u7814\u7a76\u8005\u671f\u671b\u672c\u6587\u7684\u5de5\u4f5c\u80fd\u591f\u4fc3\u8fdb\u66f4\u591a\u672a\u6765\u7814\u7a76\uff0c\u5305\u62ec\u4f46\u4e0d\u9650\u4e8e\uff1a\uff081\uff09\u5bf9 LLMs \u7684\u6570\u5b66\u6280\u80fd\u8fdb\u884c\u7cfb\u7edf\u8bc4\u4f30\uff1b\uff082\uff09\u6784\u5efa\u80fd\u591f\u7075\u6d3b\u8fdb\u884c\u6570\u5b66\u63a8\u7406\u7684\u6a21\u578b\u3002<\/section>\n<section><\/section>\n<section><em><sup>\u53c2\u8003\u94fe\u63a5<\/sup><\/em><\/section>\n<section><sup><em>[1] Cobbe, Karl, et al. &#8220;Training verifiers to solve math word problems.&#8221; arXiv preprint arXiv:2110.14168 (2021). https:\/\/paperswithcode.com\/sota\/arithmetic-reasoning-on-gsm8k<\/em><\/sup><\/section>\n<section><em><sup>[2] George Polya. 2004. How to solve it: A new aspect of mathematical method, volume 85. Princeton university press.<\/sup><\/em><\/section>\n<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:<a href=\"https:\/\/www.jiqizhixin.com\/articles\/2024-07-18-2\" target=\"_blank\">ACL 2024 | \u5bf925\u4e2a\u5f00\u95ed\u6e90\u6a21\u578b\u6570\u5b66\u8bc4\u6d4b\uff0cGPT-3.5-Turbo\u624d\u52c9\u5f3a\u53ca\u683c<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:ACL 2024 |  [&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":[68,63,67],"class_list":["post-34611","post","type-post","status-publish","format-standard","hentry","category-news","tag-68","tag-63","tag-67"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.9 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>ACL 2024 | \u5bf925\u4e2a\u5f00\u95ed\u6e90\u6a21\u578b\u6570\u5b66\u8bc4\u6d4b\uff0cGPT-3.5-Turbo\u624d\u52c9\u5f3a\u53ca\u683c - \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=34611\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34611#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34611\"},\"author\":{\"name\":\"AI\u5c0f\u52a9\u624b\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\"},\"headline\":\"ACL 2024 | \u5bf925\u4e2a\u5f00\u95ed\u6e90\u6a21\u578b\u6570\u5b66\u8bc4\u6d4b\uff0cGPT-3.5-Turbo\u624d\u52c9\u5f3a\u53ca\u683c\",\"datePublished\":\"2024-12-09T16:56:56+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34611\"},\"wordCount\":260,\"commentCount\":0,\"image\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34611#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/07\\\/frc-0e61ecaceafa415da61123235006bd58.png\",\"keywords\":[\"\u5927\u6a21\u578b\",\"\u6570\u5b66\u77e5\u8bc6\",\"\u81ea\u7136\u8bed\u8a00\u5904\u7406\"],\"articleSection\":[\"\u884c\u4e1a\u8d44\u8baf\"],\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/17aitech.com\\\/?p=34611#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34611\",\"url\":\"https:\\\/\\\/17aitech.com\\\/?p=34611\",\"name\":\"ACL 2024 | \u5bf925\u4e2a\u5f00\u95ed\u6e90\u6a21\u578b\u6570\u5b66\u8bc4\u6d4b\uff0cGPT-3.5-Turbo\u624d\u52c9\u5f3a\u53ca\u683c - \u4e00\u8d77AI\u6280\u672f\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34611#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34611#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/07\\\/frc-0e61ecaceafa415da61123235006bd58.png\",\"datePublished\":\"2024-12-09T16:56:56+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\"},\"breadcrumb\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34611#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/17aitech.com\\\/?p=34611\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34611#primaryimage\",\"url\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/07\\\/frc-0e61ecaceafa415da61123235006bd58.png\",\"contentUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/07\\\/frc-0e61ecaceafa415da61123235006bd58.png\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=34611#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\\\/\\\/17aitech.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"ACL 2024 | \u5bf925\u4e2a\u5f00\u95ed\u6e90\u6a21\u578b\u6570\u5b66\u8bc4\u6d4b\uff0cGPT-3.5-Turbo\u624d\u52c9\u5f3a\u53ca\u683c\"}]},{\"@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":"ACL 2024 | \u5bf925\u4e2a\u5f00\u95ed\u6e90\u6a21\u578b\u6570\u5b66\u8bc4\u6d4b\uff0cGPT-3.5-Turbo\u624d\u52c9\u5f3a\u53ca\u683c - \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=34611","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/17aitech.com\/?p=34611#article","isPartOf":{"@id":"https:\/\/17aitech.com\/?p=34611"},"author":{"name":"AI\u5c0f\u52a9\u624b","@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"headline":"ACL 2024 | \u5bf925\u4e2a\u5f00\u95ed\u6e90\u6a21\u578b\u6570\u5b66\u8bc4\u6d4b\uff0cGPT-3.5-Turbo\u624d\u52c9\u5f3a\u53ca\u683c","datePublished":"2024-12-09T16:56:56+00:00","mainEntityOfPage":{"@id":"https:\/\/17aitech.com\/?p=34611"},"wordCount":260,"commentCount":0,"image":{"@id":"https:\/\/17aitech.com\/?p=34611#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-0e61ecaceafa415da61123235006bd58.png","keywords":["\u5927\u6a21\u578b","\u6570\u5b66\u77e5\u8bc6","\u81ea\u7136\u8bed\u8a00\u5904\u7406"],"articleSection":["\u884c\u4e1a\u8d44\u8baf"],"inLanguage":"zh-Hans","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/17aitech.com\/?p=34611#respond"]}]},{"@type":"WebPage","@id":"https:\/\/17aitech.com\/?p=34611","url":"https:\/\/17aitech.com\/?p=34611","name":"ACL 2024 | \u5bf925\u4e2a\u5f00\u95ed\u6e90\u6a21\u578b\u6570\u5b66\u8bc4\u6d4b\uff0cGPT-3.5-Turbo\u624d\u52c9\u5f3a\u53ca\u683c - \u4e00\u8d77AI\u6280\u672f","isPartOf":{"@id":"https:\/\/17aitech.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/17aitech.com\/?p=34611#primaryimage"},"image":{"@id":"https:\/\/17aitech.com\/?p=34611#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-0e61ecaceafa415da61123235006bd58.png","datePublished":"2024-12-09T16:56:56+00:00","author":{"@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"breadcrumb":{"@id":"https:\/\/17aitech.com\/?p=34611#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/17aitech.com\/?p=34611"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/?p=34611#primaryimage","url":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-0e61ecaceafa415da61123235006bd58.png","contentUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/07\/frc-0e61ecaceafa415da61123235006bd58.png"},{"@type":"BreadcrumbList","@id":"https:\/\/17aitech.com\/?p=34611#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/17aitech.com\/"},{"@type":"ListItem","position":2,"name":"ACL 2024 | \u5bf925\u4e2a\u5f00\u95ed\u6e90\u6a21\u578b\u6570\u5b66\u8bc4\u6d4b\uff0cGPT-3.5-Turbo\u624d\u52c9\u5f3a\u53ca\u683c"}]},{"@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\/34611","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=34611"}],"version-history":[{"count":0,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/34611\/revisions"}],"wp:attachment":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=34611"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=34611"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=34611"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}