{"id":22310,"date":"2024-09-23T12:14:37","date_gmt":"2024-09-23T04:14:37","guid":{"rendered":"https:\/\/aif.amtbbs.org\/?p=22310"},"modified":"2024-09-23T12:14:37","modified_gmt":"2024-09-23T04:14:37","slug":"cot%e7%a5%9e%e8%af%9d%e7%a0%b4%e7%81%ad%ef%bc%8c%e5%b9%b6%e9%9d%9ellm%e6%a0%87%e9%85%8d%ef%bc%81%e4%b8%89%e5%a4%a7%e5%ad%a6%e5%ba%9c%e6%9c%ba%e6%9e%84%e8%81%94%e6%89%8b%e8%af%81%e5%ae%9e%ef%bc%8ccot","status":"publish","type":"post","link":"https:\/\/aif.amtbbs.org\/index.php\/2024\/09\/23\/22310\/","title":{"rendered":"CoT\u795e\u8bdd\u7834\u706d\uff0c\u5e76\u975eLLM\u6807\u914d\uff01\u4e09\u5927\u5b66\u5e9c\u673a\u6784\u8054\u624b\u8bc1\u5b9e\uff0cCoT\u4ec5\u5728\u6570\u5b66\u7b26\u53f7\u63a8\u7406\u6709\u7528"},"content":{"rendered":"<div><img data-dominant-color=\"525b5e\" data-has-transparency=\"false\" style=\"--dominant-color: #525b5e;\" loading=\"lazy\" decoding=\"async\" class=\"not-transparent alignnone size-full wp-image-22312\" src=\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2024\/09\/98630c9424e1b1baa3c473fc733e4e8aea6c29-300x167-1.jpg\" width=\"300\" height=\"167\" alt=\"\" srcset=\"https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2024\/09\/98630c9424e1b1baa3c473fc733e4e8aea6c29-300x167-1.jpg 300w, https:\/\/aiforumimage.oss-cn-shanghai.aliyuncs.com\/wp-content\/uploads\/2024\/09\/98630c9424e1b1baa3c473fc733e4e8aea6c29-300x167-1-150x84.jpg 150w\" sizes=\"auto, (max-width: 300px) 100vw, 300px\" \/><\/div>\n<div><\/div>\n<div class=\"article-desc\">CoT\u53ea\u5bf9\u6570\u5b66\u3001\u7b26\u53f7\u63a8\u7406\u624d\u8d77\u4f5c\u7528\uff0c\u5176\u4ed6\u7684\u4efb\u52a1\u51e0\u4e4e\u6ca1\u4ec0\u4e48\u5375\u7528\uff01\u8fd9\u662f\u6765\u81eaUT-Austin\u3001\u970d\u666e\u91d1\u65af\u3001\u666e\u6797\u65af\u987f\u4e09\u5927\u673a\u6784\u7814\u7a76\u4eba\u5458\u8054\u624b\uff0c\u5206\u6790\u4e86100+\u7bc7\u8bba\u658714\u7c7b\u4efb\u52a1\u5f97\u51fa\u7684\u7ed3\u8bba\u3002\u770b\u6765\uff0cCoT\u5e76\u975e\u662f\u6240\u6709\u5927\u6a21\u578b\u6807\u914d\u3002<\/div>\n<div id=\"postspictures\" class=\"article-content\">\n<div id=\"container\" class=\"container am-engine\" data-v-1d7a5742=\"\" data-element=\"root\">\n<p>\u8c37\u6b4cCoT\u5f00\u5c71\u4e4b\u4f5c\uff0c\u518d\u6b21\u6210\u4e3aOpenAI o1\u6a21\u578b\u7684\u5229\u5668\u3002<\/p>\n<p>LLM\u590d\u6742\u63a8\u7406\u80fd\u529b\u7684\u5b9e\u73b0\uff0c\u5c31\u79bb\u4e0d\u5f00\u4e00\u6b65\u4e00\u6b65\u601d\u8003\uff0c\u4f46\u662f\u8fd9\u79cd\u300c\u601d\u8003\u300d\u7a76\u7adf\u5bf9\u4e8e\u4ec0\u4e48\u7c7b\u578b\u7684\u4efb\u52a1\u6709\u5e2e\u52a9\u5462\uff1f<\/p>\n<p>\u6765\u81eaUT-Austin\u3001\u970d\u666e\u91d1\u65af\u3001\u666e\u6797\u65af\u987f\u7684\u7814\u7a76\u4eba\u5458\uff0c\u4f7f\u7528CoT\u5bf9100\u591a\u7bc7\u8bba\u6587\uff0c\u8fdb\u884c\u4e86\u5b9a\u91cf\u5143\u5206\u6790\u3002<\/p>\n<p>&nbsp;<\/p>\n<p>\u8bba\u6587\u5730\u5740\uff1ahttps:\/\/arxiv.org\/abs\/2409.12183<\/p>\n<p>\u540c\u65f6\uff0c\u4ed6\u4eec\u8fd8\u5bf914\u4e2a\u6a21\u578b\u768420\u4e2a\u6570\u636e\u96c6\uff0c\u8fdb\u884c\u4e86\u8bc4\u4f30\u3002<\/p>\n<p>\u7ed3\u679c\u663e\u793a\uff0cCoT\u5728\u6d89\u53ca\u6570\u5b66\u3001\u903b\u8f91\u4efb\u52a1\u4e2d\uff0c\u80fd\u591f\u589e\u5f3aLLM\u6027\u80fd\uff0c\u4f46\u5728\u5176\u4ed6\u7c7b\u578b\u4efb\u52a1\u4e0a\uff0c\u6536\u76ca\u8f83\u5c0f\u3002<\/p>\n<p>&nbsp;<\/p>\n<p>\u5728MMLU\u4e2d\uff0c\u9664\u975e\u95ee\u9898\u6216\u6a21\u578b\u56de\u7b54\u4e2d\u5305\u542b\u201c=\u201d\uff08\u8868\u793a\u7b26\u53f7\u8fd0\u7b97\u548c\u63a8\u7406\uff09\uff0c\u5426\u5219\u76f4\u63a5\u751f\u6210\u7b54\u6848\uff0c\u800c\u4e0d\u4f7f\u7528CoT\uff0c\u80fd\u8fbe\u5230\u4e0e\u4f7f\u7528CoT\u76f8\u540c\u7684\u51c6\u786e\u7387\u3002<\/p>\n<p>\u57fa\u4e8e\u8fd9\u4e00\u53d1\u73b0\uff0c\u4f5c\u8005\u901a\u8fc7\u5206\u79bb\u300c\u89c4\u5212\u300d\u548c\u300c\u6267\u884c\u300d\uff0c\u5e76\u5de5\u5177\u589e\u5f3a\u7684LLM\u8fdb\u884c\u6bd4\u8f83\uff0c\u6765\u5206\u6790CoT\u5728\u8fd9\u4e9b\u4efb\u52a1\u4e2d\u7684\u884c\u4e3a\u3002<\/p>\n<p>\u7ed3\u679c\u8bc1\u5b9e\uff0cCoT\u7684\u5927\u90e8\u5206\u6536\u76ca\uff0c\u6765\u81ea\u4e8e\u7b26\u53f7\u6267\u884c\u7684\u6539\u8fdb\uff0c\u4f46\u76f8\u5bf9\u4e8e\u4f7f\u7528\u7b26\u53f7\u6c42\u89e3\u5668\u6765\u8bf4\uff0c\u8868\u73b0\u8f83\u5dee\u3002<\/p>\n<p>&nbsp;<\/p>\n<p>\u603b\u800c\u8a00\u4e4b\uff0c\u4f5c\u8005\u5e0c\u671b\u901a\u8fc7\u7814\u7a76\u544a\u8bc9\u6211\u4eec\uff0cCoT\u5e76\u975e\u662f\u4e07\u80fd\u7684\u3002<\/p>\n<p>\u300cCoT\u53ef\u4ee5\u6709\u9009\u62e9\u6027\u5730\u5e94\u7528\uff0c\u4ee5\u5e73\u8861\u6027\u80fd\u548c\u63a8\u7406\u8ba1\u7b97\u6210\u672c\u300d\u3002<\/p>\n<p>\u53e6\u5916\uff0c\u672a\u6765\u7814\u7a76\u4e2d\u8fd8\u9700\u8d85\u8d8a\u57fa\u4e8e\u63d0\u793a\u7684CoT\uff0c\u8f6c\u5411\u5229\u7528\u6574\u4e2aLLM\u4e2d\u95f4\u8ba1\u7b97\u7684\u65b0\u8303\u5f0f\u3002<\/p>\n<h3>\u7528CoT\uff0c\u8fd8\u662f\u4e0d\u7528CoT\uff0c\u8fd9\u662f\u4e00\u4e2a\u95ee\u9898<\/h3>\n<p>o1\u6210\u4e3a\u5f53\u7ea2\u70b8\u5b50\u9e21\uff0c\u4e00\u5927\u539f\u56e0\u4fbf\u662f\u91c7\u7528\u4e86CoT\u67b6\u6784\u3002<\/p>\n<p>\u4f46\u662f\uff0c\u5343\u4e07\u4e0d\u8981\u88abOpenAI\u8bef\u5bfc\u4e86\u3002<\/p>\n<p>&nbsp;<\/p>\n<p>\u5176\u5b9e\uff0c\u5148\u524d\u5c31\u6709\u7814\u7a76\u79f0\uff0cCoT\u5728\u6570\u5b66\u4ee5\u5916\u7684\u9886\u57df\uff0c\u5e76\u6ca1\u6709\u90a3\u4e48\u6709\u7528\uff0c\u751a\u81f3\u6709\u65f6\u4f1a\u635f\u5bb3\u6a21\u578b\u6027\u80fd\u3002<\/p>\n<p>\u6700\u65b0\u7814\u7a76\u4e2d\uff0c\u4f5c\u8005\u7684\u76ee\u6807\u662f\u53bb\u8bc4\u4f30\uff0c\u57fa\u4e8e\u63d0\u793a\u7684CoT\u5728\u54ea\u4e9b\u9886\u57df\u53d1\u6325\u4f5c\u7528\u6700\u5927\uff0c\u4ee5\u53ca\u5177\u4f53\u7684\u539f\u56e0\u3002<\/p>\n<h3>110\u7bc7\u8bba\u6587\uff0c14\u5927\u7c7b\u522b<\/h3>\n<p>\u5bf9\u6b64\uff0c\u7814\u7a76\u4eba\u5458\u4ece\u673a\u5668\u5b66\u4e60\u9876\u4f1aICLR 2024\u3001\u4e24\u4e2aNLP\u9876\u4f1aEACL 2024\u548cNAACL 2024\u4e2d\uff0c\u7b5b\u9009\u4e86\u6240\u67094642\u7bc7\u8bba\u6587\uff082259\u7bc7\u6765\u81eaICLR 2024\uff0c2,382\u7bc7\u6765\u81ea\u4e24\u4e2aACL\u9644\u5c5e\u4f1a\u8bae\uff09\u3002<\/p>\n<p>\u5728\u8fd9\u4e9b\u8bba\u6587\u4e2d\uff0c\u901a\u8fc7\u81ea\u52a8\u7b5b\u9009\u51fa\u73b0CoT\u3001chain-of-thought\u6216chain of thought\u4e24\u6b21\u7684\u6587\u7ae0\uff0c\u5171\u5f97\u5230516\u4e2a\u6837\u672c\u3002<\/p>\n<p>\u7136\u540e\uff0c\u4e0b\u4e00\u6b65\u901a\u8fc7\u624b\u52a8\u7b5b\u9009\uff0c\u5f97\u5230\u5bf9\u300cCoT\u63d0\u793a\u4e0e\u76f4\u63a5\u63d0\u793a\u8fdb\u884c\u6bd4\u8f83\u300d\u7684\u8bba\u6587\u3002<\/p>\n<p>\u7ecf\u8fc7\u4e00\u7cfb\u5217\u6392\u9664\uff0c\u6700\u7ec8\u6536\u96c6\u4e86110\u7bc7\u8bba\u6587\u6837\u672c\uff0c\u6db5\u76d6\u4e86264\u4e2a\u6570\u636e\u96c6\u3002<\/p>\n<p>\u6700\u540e\uff0c\u7814\u7a76\u4eba\u5458\u5c06\u6240\u83b7\u5f97\u6837\u672c\u4e2d\u7684\u4e0d\u540c\u4efb\u52a1\uff0c\u5206\u621014\u4e2a\u7c7b\u522b\u3002<\/p>\n<p>&nbsp;<\/p>\n<p>\u5982\u4e0b\u56fe2\u6240\u793a\uff0c\u5728\u4e0d\u6587\u732e\u4e2d\uff0c\u4f5c\u8005\u53d1\u73b0CoT\u5728\u4efb\u52a1\u5206\u7c7b\u4e2d\u7684\u6027\u80fd\u589e\u76ca\u3002<\/p>\n<p>\u53ef\u89c1\uff0cCoT\u5728\u7b26\u53f7\u63a8\u7406\u3001\u6570\u5b66\u3001\u903b\u8f91\u63a8\u7406\u4e09\u5927\u7c7b\u522b\u4efb\u52a1\u4e2d\uff0c\u6027\u80fd\u6700\u4f18\uff0c\u5e73\u5747\u63d0\u9ad8\u5206\u522b\u4e3a14.2%\u300112.3%\u548c6.9%\u3002<\/p>\n<p>\u4f7f\u7528CoT\u7684\u524d\u4e09\u9879\u4efb\u52a1\uff0c\u5e73\u5747\u6027\u80fd\u4e3a56.9%\uff0c\u800c\u6ca1\u6709CoT\u7684\u6027\u80fd\u4e3a45.5%\u3002<\/p>\n<p>\u5bf9\u4e8e\u5176\u4ed6\u7c7b\u522b\u4efb\u52a1\uff0c\u4f7f\u7528CoT\u7684\u5e73\u5747\u6027\u80fd\u4e3a56.8%\uff0c\u800c\u6ca1\u6709\u4f7f\u7528CoT\u7684\u5e73\u5747\u6027\u80fd\u4e3a56.1%\u3002<\/p>\n<p>&nbsp;<\/p>\n<p>\u56fe2\u53f3\u4fa7\u663e\u793a\u4e86\uff0c\u5728\u6570\u5b66\u3001\u7b26\u53f7\u6216\u903b\u8f91\u63a8\u7406\u4ee5\u5916\u7684\u4efb\u52a1\u5b9e\u9a8c\u4e2d\uff0c\u5e73\u5747CoT\u589e\u91cf\u8f83\u9ad810\u4e2a\u5f02\u5e38\u503c\u3002<\/p>\n<p>\u867d\u7136\u8fd9\u4e9b\u8bba\u6587\u6ca1\u6709\u88ab\u5f52\u7c7b\u4e3a\u6570\u5b66\u903b\u8f91\uff0c\u4f46\u5176\u4e2d\u4e00\u4e9b\u5728\u67d0\u79cd\u7a0b\u5ea6\u4e0a\u4e0e\u903b\u8f91\u3001\u6570\u5b66\u6216\u7b26\u53f7\u63a8\u7406\u76f8\u5173\u3002<\/p>\n<p>\u4ece\u8fd9\u4e2a\u5217\u8868\u4e2d\u53ef\u4ee5\u770b\u51fa\uff0c\u4eceCoT\u4e2d\u83b7\u76ca\u6700\u591a\u7684\u6570\u636e\u96c6\u662fBIG-bench Hard\u3002\u8fd8\u6709BIG-bench Temporal\u3001MMLU-Moral Scenarios\u90fd\u6d89\u53ca\u5230\u4e86\u57fa\u672c\u7b80\u5355\u95ee\u9898\u7684\u7ec4\u5408\u3002<\/p>\n<p>&nbsp;<\/p>\n<p>\u540c\u65f6\uff0c\u8fd8\u6709\u51e0\u4e2a\u5f02\u5e38\u503c\uff0c\u4e5f\u9690\u7ea6\u9075\u5faa\u8fd9\u4e00\u8d8b\u52bf\u3002\u6bd4\u5982\uff0cScienceQ\u662f\u7531\u4e00\u7cfb\u5217\u81ea\u7136\u548c\u793e\u4f1a\u79d1\u5b66\u79d1\u5b66\u9009\u62e9\u9898\u7ec4\u6210\uff0c\u4f46\u5982\u679c\u4e0d\u6309\u5b66\u79d1\/\u95ee\u9898\u7c7b\u578b\u7ec6\u5206\u6210\u7ee9\uff0c\u5f88\u96be\u89e3\u91ca\u5176\u6536\u76ca\u3002<\/p>\n<p>\u5176\u6b21\uff0c\u5728\u4e00\u4e9b\u8bba\u6587\u5206\u6790\u4e2d\uff0c\u5176\u4ed6\u8bba\u6587\u8bc4\u5206\u7ed3\u679c\u5e76\u672a\u663e\u793a\u51faCoT\u5e26\u6765\u7684\u6539\u8fdb\u3002<\/p>\n<h3>\u5b9e\u9a8c\u7ed3\u679c<\/h3>\n<p>\u66f4\u8fdb\u4e00\u6b65\u7684\uff0c\u56e2\u961f\u5728\u96f6\u6837\u672c\u5b66\u4e60\u548c\u5c11\u6837\u672c\u5b66\u4e60\u7684\u8bbe\u7f6e\u4e0b\uff0c\u5bf914\u4e2a\u6a21\u578b\u768420\u4e2a\u6570\u636e\u96c6\u8fdb\u884c\u4e86\u4e00\u7cfb\u5217\u5b9e\u9a8c\uff0c\u4ee5\u6bd4\u8f83\u6027\u80fd\u3002<\/p>\n<h3>\u4e0eCoT\u76f8\u6bd4\uff0c\u96f6\u6837\u672cCoT\u6709\u4f55\u6539\u8fdb\uff1f<\/h3>\n<p>\u5982\u4e0b\u56fe3\u6240\u793a\uff0c\u5177\u4f53\u663e\u793a\u4e86\u56fe1\u4e2d\uff0c\u6bcf\u4e2a\u63a8\u7406\u7c7b\u522b\u7684\u5e73\u5747CoT\u6027\u80fd\u6539\u8fdb\u3002\u53f3\u4fa7\u5448\u73b0\u7684\u662f\uff0c\u5bf9\u6bcf\u4e2a\u6570\u636e\u96c6\u4f7f\u7528CoT\u6240\u5e26\u6765\u7684\u6027\u80fd\u589e\u76ca\uff0c\u8fd9\u662f\u6240\u6709\u6a21\u578b\u548c\u5355\u4e2a\u6a21\u578b\u9009\u62e9\u7684\u5e73\u5747\u503c\u3002<\/p>\n<p>\u5728\u975e\u7b26\u53f7\u63a8\u7406\u7c7b\u522b\u548c\u6570\u636e\u96c6\u4e0a\uff0c\u7279\u522b\u662f\u90a3\u4e9b\u5305\u542b\u4e3b\u8981\u6d89\u53ca\u5e38\u8bc6\uff08CSOA\u3001PIOA\u3001SiOA\uff09\u3001\u8bed\u8a00\u7406\u89e3\uff08WinoGrande\uff09\u548c\u9605\u8bfb\u7406\u89e3\uff08(AGILSAT\u3001ARC-Easy\u3001ARC-Challenge\uff09\u95ee\u9898\u7684\u6570\u636e\u96c6\u3002<\/p>\n<p>\u96f6\u6837\u672cCoT\u548c\u96f6\u6837\u672c\u76f4\u63a5\u56de\u7b54\u7684\u6027\u80fd\u4e4b\u95f4\uff0c\u51e0\u4e4e\u6ca1\u6709\u533a\u522b\u3002<\/p>\n<p>\u5c3d\u7ba1\u8fd9\u4e9b\u6570\u636e\u96c6\u4e2d\u6d89\u53ca\u63a8\u7406\uff0c\u4f46\u662fCoT\u5e76\u6ca1\u6709\u5e26\u6765\u663e\u8457\u7684\u6539\u8fdb\u3002<\/p>\n<p>&nbsp;<\/p>\n<p>\u76f8\u8f83\u4e4b\u4e0b\uff0c\u6570\u5b66\u548c\u7b26\u53f7\u7c7b\u522b\uff0c\u4e0e\u7b26\u53f7\u548c\u8bb8\u591a\u7b26\u53f7\u6570\u636e\u96c6\u4e00\u8d77\u5f97\u5230\u4e86\u5927\u5e45\u6539\u8fdb\u3002<\/p>\n<p>MATH\u548cGSM8k\u7684\u589e\u5e45\u5206\u522b\u9ad8\u8fbe41.6%\u548c66.9%\u3002\u5bf9\u4e8e\u534a\u7b26\u53f7\u6570\u636e\u96c6\u5982Mvsteries\u7b49\uff0c\u7ed3\u679c\u663e\u793a\u51fa\u9002\u5ea6\u7684\u589e\u5e45\u3002<\/p>\n<p>\u4f8b\u5982\uff0c\u4ece\u7b80\u5355\u7684\u81ea\u7136\u8bed\u8a00(ContextHub)\u6216\u66f4\u590d\u6742\u7684\u5e38\u8bc6\u9648\u8ff0(MuSR \u8c0b\u6740\u4e4b\u8c1c)\u89e3\u6790\u51fa\u4e00\u9636\u903b\u8f91\u3002\u6240\u6709\u7ed3\u679c\u5747\u663e\u793a\u5728\u9644\u5f55 C.1\u4e2d\uff0c\u88687\u4e2d\u8fd8\u663e\u793a\u4e86 CoT \u548c\u76f4\u63a5\u7b54\u6848\u63d0\u793a\u7684\u5b8c\u6574\u6570\u5b57\u7ed3\u679c\u5217\u8868\u3002\u6211\u4eec\u8fd8\u63a2\u7d22\u4e86\u5c11\u6570\u955c\u5934\u8bbe\u7f6e\uff0c\u53d1\u73b0\u5b83\u5bf9 CoT \u4f55\u65f6\u63d0\u4f9b\u5e2e\u52a9\u7684\u5f71\u54cd\u4e0d\u5927;\u53c2\u89c1\u9644\u5f55 B\u3002<\/p>\n<h3>\u7b54\u6848\u5f62\u5f0f\uff0c\u662f\u5426\u4f1a\u5f71\u54cd\u5230CoT\u5e2e\u52a9\u7684\u8303\u56f4\uff1f<\/h3>\n<p>\u7b54\u6848\u2014\u2014\u4e0d\u591a\uff0c\u9884\u5148\u89c4\u5212\u6216\u63a8\u7406\u6b63\u786e\u53cd\u5e94\uff0c\u53ef\u80fd\u4f1a\u963b\u788d\u6a21\u578b\u81ea\u4e3b\u53cd\u5e94\u80fd\u529b\u3002<\/p>\n<p>\u9664\u4e86\u6570\u5b66\u4e4b\u5916\uff0c\u8bb8\u591a\u5e38\u7528\u7684\u95ee\u9898\u6570\u636e\u96c6\u662f\u591a\u9879\u9009\u62e9\u3002<\/p>\n<p>\u5bf9\u4e8e\u4e24\u4e2a\u975e\u591a\u9879\u9009\u62e9\uff0c\u4e14\u5305\u542b\u4e0d\u540c\u7ea7\u522b\u7684\u975e\u7b26\u53f7\u63a8\u7406\u6765\u56de\u7b54\u95ee\u9898\u6570\u636e\u96c6\uff0cCoT\u5177\u6709\u4e0e\u8de8\u6a21\u578b\u76f4\u63a5\u56de\u7b54\u76f8\u4f3c\u7684\u6027\u80fd\u3002<\/p>\n<p>\u5176\u6b21\uff0cBiGGen Bench\u4f7f\u7528\u81ea\u7531\u5f0f\u56de\u7b54\u4f5c\u4e3a\u95ee\u9898\u7b54\u6848\uff0c\u5e76\u4f7f\u7528LLM\u4f5c\u4e3a\u6cd5\u5b98\uff0c\u4ee51-5\u7ea7\u6765\u8bc4\u4f30\u8fd9\u4e9b\u56de\u7b54\u3002<\/p>\n<p>\u5f97\u5230\u7684\u7b54\u6848\uff0c\u672c\u8d28\u4e0a\u6a21\u7cca\u4e86CoT\u548c\u76f4\u63a5\u7b54\u6848\u4e4b\u95f4\u7684\u754c\u9650\u3002<\/p>\n<p>\u4e3a\u6b64\uff0c\u7814\u7a76\u4eba\u5458\u8bbe\u7f6e\u4e86\u4e00\u4e2a\u65b0\u7684CoT\u63d0\u793a\uff0c\u8981\u6c42\u8bed\u8a00\u6a21\u578b\u751f\u6210\u81ea\u7531\u5f62\u5f0f\u54cd\u5e94\u7684\u89c4\u5212\uff0c\u7136\u540e\u8981\u6c42\u5176\u5728\u751f\u6210\u5b8c\u6574\u7684\u54cd\u5e94\u3002<\/p>\n<h3>\u77e5\u8bc6\u3001\u8f6f\u63a8\u7406\u548c\u5e38\u8bc6\u65b9\u9762\uff0c\u6027\u80fd\u63d0\u5347\u663e\u8457\u5417\uff1f<\/h3>\n<p>\u9664\u4e86MMLU\u3001StrategyQA\u548cMuSR\u5916\uff0c\u5927\u591a\u6570\u60c5\u51b5\u4e0b\u7b54\u6848\u662f\u5426\u5b9a\u7684\u3002<\/p>\n<p>\u4f5c\u8005\u4f7f\u7528\u914d\u5bf9\u5f15\u5bfc\u6cd5\u5728\u77e5\u8bc6\u3001\u8f6f\u63a8\u7406\u548c\u5e38\u8bc6\u63a8\u7406\u7c7b\u522b\u768413\u4e2a\u6570\u636e\u96c6\u4e0a\uff0c\u6d4b\u8bd5\u4e86CoT\u6539\u8fdb\u7684\u663e\u8457\u6027\u3002<\/p>\n<p>\u7ed3\u679c\u5f97\u51fa\uff0c\u5927\u7ea6 38%\u7684\u6570\u636e\u96c6\u663e\u793a\u51fa\uff0c\u8fd9\u4e09\u4e2a\u63a8\u7406\u7c7b\u522b\u7684\u6548\u76ca\u662f\u663e\u8457\u7684\u3002<\/p>\n<h3>MMLU\u548cMMLU PRO<\/h3>\n<p>MMLU\u548cMMLU Pro\u663e\u793a\u4e86\uff0c\u4f7f\u7528CoT\u5e26\u6765\u6536\u76ca\uff0c\u4f46\u7531\u4e8e\u8fd9\u4e9b\u6570\u636e\u96c6\u975e\u5e38\u5e7f\u6cdb\uff0c\u56e0\u6b64\u65e0\u6cd5\u8fdb\u884c\u7b80\u5355\u7684\u8868\u5f81\u3002<\/p>\n<p>\u7814\u7a76\u4eba\u5458\u63a2\u7d22\u4e86MMLU\u6bcf\u4e2a\u7c7b\u522b\u4e0a\u7684CoT\u6027\u80fd\uff0c\u4ee5\u4e86\u89e3\u8fd9\u4e9b\u9886\u57df\u4e4b\u95f4CoT\u6027\u80fd\u7684\u5dee\u5f02\u3002<\/p>\n<p>\u5bf9\u6b64\uff0c\u4ed6\u4eec\u5217\u51fa\u4e863\u4e2a\u7c7b\u522b\uff0c\u5176\u4e2dCoT\u5728MMLU\u548cMMLU Pro\u4e0a\uff0cLlama 3.1 8B\u548c70B\u6700\u5927\u7684\u8bef\u5dee\u51cf\u5c11\u3002<\/p>\n<p>&nbsp;<\/p>\n<p>\u5176\u4e2d\uff0c\u4e00\u4e9b\u4efb\u52a1\u7c7b\u522b\u660e\u663e\u662f\u6570\u5b66\u6027\u8d28\uff0c\u6b63\u5982\u56fe8\u4e2d\u6240\u5448\u73b0\u7684\u90a3\u6837\u3002<\/p>\n<p>\u6b64\u5916\uff0c\u6211\u4eec\u8fd8\u53ef\u4ee5\u770b\u5230\uff0cCoT\u5728\u5546\u4e1a\u4efb\u52a1\u4e0a\uff0c\u80fd\u591f\u63d0\u4f9b\u4e00\u5b9a\u7684\u5e2e\u52a9\u3002\u7ecf\u8fc7\u4ed4\u7ec6\u6838\u67e5\uff0c\u8fd9\u4e9b\u4efb\u52a1\u4e5f\u4f1a\u5e38\u5e38\u6d89\u53ca\u6570\u5b66\u7b49\u4e00\u4e9b\u5185\u5bb9\u3002<\/p>\n<p>&nbsp;<\/p>\n<p>\u5b9e\u9a8c\u7ed3\u679c\u5982\u4e0b\u56fe4\u6240\u793a\uff0c\u5f53\u95ee\u9898\u6216\u751f\u6210\u7ed3\u679c\u6c47\u603b\u5305\u542b\u201c=\u201d\uff0c\u4ee5\u53ca\u4e0d\u5305\u542b\u201c=\u201d\u65f6\uff0c\u4f7f\u7528CoT\u65f6\u7684\u589e\u76ca\u6548\u679c\u3002<\/p>\n<p>&nbsp;<\/p>\n<h3>CoT\u5728\u516c\u5f0f\u63a8\u7406\u4e2d\u7684\u4f18\u7f3a\u70b9<\/h3>\n<p>\u524d\u9762\u4e3b\u8981\u8bf4\u660e\u4e86CoT\u4e3b\u8981\u5728\u7b26\u53f7\u63a8\u7406\u4efb\u52a1\u53d1\u6325\u4f5c\u7528\uff0c\u4f46\u6ca1\u6709\u8bf4\u660e\u5177\u4f53\u539f\u56e0\u3002<\/p>\n<p>\u5728\u7b26\u53f7\u4efb\u52a1\u4e0a\uff0c\u7814\u7a76\u4eba\u5458CoT\u7684\u6027\u80fd\u63d0\u5347\u5f52\u56e0\u4e8e\u4e24\u4e2a\u9636\u6bb5\uff1a\u89c4\u5212\u9636\u6bb5\u548c\u6267\u884c\u9636\u6bb5\u3002<\/p>\n<p>&nbsp;<\/p>\n<p>\u4e0b\u56fe6\u663e\u793a\u4e86\uff0c\u4ee3\u8868\u6027\u6a21\u578b\u9009\u62e9\u7684\u7ed3\u679c\u3002<\/p>\n<p>\u5c06\u76f4\u63a5\u7b54\u6848\u4e0ePlan+ Direct\u6c42\u89e3\u5668\u548cPlan+CoT\u6c42\u89e3\u5668\u8fdb\u884c\u6bd4\u8f83\u65f6\uff0c\u53ef\u4ee5\u6ce8\u610f\u5230\uff0c\u5bf9\u4e8e\u8bb8\u591a\u6570\u636e\u96c6\u548c\u6a21\u578b\uff0c\u4ec5\u6709\u89c4\u5212\u5e76\u4e0d\u80fd\u89e3\u51b3\u5927\u90e8\u5206\u6027\u80fd\u589e\u76ca\u3002<\/p>\n<p>\u4e0e\u76f4\u63a5\u7b54\u6848\u76f8\u6bd4\uff0c\u9700\u8981CoT\u6216Plan+CoT\u6c42\u89e3\u5668\uff0c\u6765\u83b7\u5f97\u5f3a\u5927\u7684\u6027\u80fd\u3002<\/p>\n<p>\u5c3d\u7ba1CoT\u548cPlan+CoT\u6c42\u89e3\u5668\uff0c\u6bd4\u76f4\u63a5\u7b54\u6848\u548cPlan+Direct\u6c42\u89e3\u5668\u6709\u4f18\u52bf\uff0c\u4f46\u5728\u5927\u591a\u6570\u8bbe\u7f6e\u4e2d\uff0c\u4ecd\u4ee5Plan+Tool\u6c42\u89e3\u5668\u4e3a\u4e3b\u3002<\/p>\n<p>\u4e0e\u7b26\u53f7\u6c42\u89e3\u5668\u76f8\u6bd4\uff0cLLM\u6267\u884c\u548c\u8ffd\u8e2a\u6b65\u9aa4\u7684\u80fd\u529b\u53d7\u5230\u9650\u5236\u3002<\/p>\n<p>&nbsp;<\/p>\n<p>\u9274\u4e8e\u4ee5\u4e0a\u7684\u53d1\u73b0\uff0c\u7814\u7a76\u56e2\u961f\u8ba4\u4e3aCoT\u5e94\u8be5\u6709\u9009\u62e9\u6027\u5730\u5e94\u7528\uff0c\u5c24\u5176\u662f\u5728\u9700\u8981\u5904\u7406\u6570\u5b66\u3001\u903b\u8f91\u63a8\u7406\u7684\u4efb\u52a1\u3002<\/p>\n<p>\u800c\u4e0d\u5206\u9752\u7ea2\u7682\u767d\u5730\u4f7f\u7528CoT\uff0c\u53ef\u80fd\u4f1a\u5bfc\u81f4\u63a8\u7406\u6210\u672c\u589e\u52a0\u3002<\/p>\n<p>\u4ed6\u4eec\u8fd8\u5efa\u8bae\uff0c\u82e5\u60f3\u8fdb\u4e00\u6b65\u63d0\u5347\u6a21\u578b\u63a8\u7406\u80fd\u529b\uff0c\u8fd8\u9700\u8981\u8d85\u8d8a\u57fa\u4e8e\u63d0\u793a\u7684CoT\u3002<\/p>\n<p>\u603b\u800c\u8a00\u4e4b\uff0cCoT is not all you need\u3002<\/p>\n<\/div>\n<\/div>\n<div class=\"pvc_clear\"><\/div>\n<p id=\"pvc_stats_22310\" class=\"pvc_stats total_only  \" data-element-id=\"22310\" style=\"\"><i class=\"pvc-stats-icon medium\" aria-hidden=\"true\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" version=\"1.0\" viewBox=\"0 0 502 315\" preserveAspectRatio=\"xMidYMid meet\"><g transform=\"translate(0,332) scale(0.1,-0.1)\" fill=\"\" stroke=\"none\"><path d=\"M2394 3279 l-29 -30 -3 -207 c-2 -182 0 -211 15 -242 39 -76 157 -76 196 0 15 31 17 60 15 243 l-3 209 -33 29 c-26 23 -41 29 -80 29 -41 0 -53 -5 -78 -31z\"\/><path d=\"M3085 3251 c-45 -19 -58 -50 -96 -229 -47 -217 -49 -260 -13 -295 52 -53 146 -42 177 20 16 31 87 366 87 410 0 70 -86 122 -155 94z\"\/><path d=\"M1751 3234 c-13 -9 -29 -31 -37 -50 -12 -29 -10 -49 21 -204 19 -94 39 -189 45 -210 14 -50 54 -80 110 -80 34 0 48 6 76 34 21 21 34 44 34 59 0 14 -18 113 -40 219 -37 178 -43 195 -70 221 -36 32 -101 37 -139 11z\"\/><path d=\"M1163 3073 c-36 -7 -73 -59 -73 -102 0 -56 133 -378 171 -413 34 -32 83 -37 129 -13 70 36 67 87 -16 290 -86 209 -89 214 -129 231 -35 14 -42 15 -82 7z\"\/><path d=\"M3689 3066 c-15 -9 -33 -30 -42 -48 -48 -103 -147 -355 -147 -375 0 -98 131 -148 192 -74 13 15 57 108 97 206 80 196 84 226 37 273 -30 30 -99 39 -137 18z\"\/><path d=\"M583 2784 c-38 -19 -67 -74 -58 -113 9 -42 211 -354 242 -373 16 -10 45 -18 66 -18 51 0 107 52 107 100 0 39 -1 41 -124 234 -80 126 -108 162 -133 173 -41 17 -61 16 -100 -3z\"\/><path d=\"M4250 2784 c-14 -9 -74 -91 -133 -183 -95 -150 -107 -173 -107 -213 0 -55 33 -94 87 -104 67 -13 90 8 211 198 130 202 137 225 78 284 -27 27 -42 34 -72 34 -22 0 -50 -8 -64 -16z\"\/><path d=\"M2275 2693 c-553 -48 -1095 -270 -1585 -649 -135 -104 -459 -423 -483 -476 -23 -49 -22 -139 2 -186 73 -142 361 -457 571 -626 285 -228 642 -407 990 -497 242 -63 336 -73 660 -74 310 0 370 5 595 52 535 111 1045 392 1455 803 122 121 250 273 275 326 19 41 19 137 0 174 -41 79 -309 363 -465 492 -447 370 -946 591 -1479 653 -113 14 -422 18 -536 8z m395 -428 c171 -34 330 -124 456 -258 112 -119 167 -219 211 -378 27 -96 24 -300 -5 -401 -72 -255 -236 -447 -474 -557 -132 -62 -201 -76 -368 -76 -167 0 -236 14 -368 76 -213 98 -373 271 -451 485 -162 444 86 934 547 1084 153 49 292 57 452 25z m909 -232 c222 -123 408 -262 593 -441 76 -74 138 -139 138 -144 0 -16 -233 -242 -330 -319 -155 -123 -309 -223 -461 -299 l-81 -41 32 46 c18 26 49 83 70 128 143 306 141 649 -6 957 -25 52 -61 116 -79 142 l-34 47 45 -20 c26 -10 76 -36 113 -56z m-2057 25 c-40 -58 -105 -190 -130 -263 -110 -324 -59 -707 132 -981 25 -35 42 -64 37 -64 -19 0 -241 119 -326 174 -188 122 -406 314 -532 468 l-58 71 108 103 c185 178 428 349 672 473 66 33 121 60 123 61 2 0 -10 -19 -26 -42z\"\/><path d=\"M2375 1950 c-198 -44 -350 -190 -395 -379 -18 -76 -8 -221 19 -290 114 -284 457 -406 731 -260 98 52 188 154 231 260 27 69 37 214 19 290 -38 163 -166 304 -326 360 -67 23 -215 33 -279 19z\"\/><\/g><\/svg><\/i> <img loading=\"lazy\" decoding=\"async\" width=\"16\" height=\"16\" alt=\"Loading\" src=\"https:\/\/aif.amtbbs.org\/wp-content\/plugins\/page-views-count\/ajax-loader-2x.gif\" border=0 \/><\/p>\n<div class=\"pvc_clear\"><\/div>\n","protected":false},"excerpt":{"rendered":"<p>CoT\u53ea\u5bf9\u6570\u5b66\u3001\u7b26\u53f7\u63a8\u7406\u624d\u8d77\u4f5c\u7528\uff0c\u5176\u4ed6\u7684\u4efb\u52a1\u51e0\u4e4e\u6ca1\u4ec0\u4e48\u5375\u7528\uff01\u8fd9\u662f\u6765\u81eaUT-Austin\u3001\u970d\u666e\u91d1\u65af\u3001\u666e\u6797\u65af\u987f\u4e09\u5927 [&hellip;]<\/p>\n<div class=\"pvc_clear\"><\/div>\n<p id=\"pvc_stats_22310\" class=\"pvc_stats total_only  \" data-element-id=\"22310\" style=\"\"><i class=\"pvc-stats-icon medium\" aria-hidden=\"true\"><svg xmlns=\"http:\/\/www.w3.org\/2000\/svg\" version=\"1.0\" viewBox=\"0 0 502 315\" preserveAspectRatio=\"xMidYMid meet\"><g transform=\"translate(0,332) scale(0.1,-0.1)\" fill=\"\" stroke=\"none\"><path d=\"M2394 3279 l-29 -30 -3 -207 c-2 -182 0 -211 15 -242 39 -76 157 -76 196 0 15 31 17 60 15 243 l-3 209 -33 29 c-26 23 -41 29 -80 29 -41 0 -53 -5 -78 -31z\"\/><path d=\"M3085 3251 c-45 -19 -58 -50 -96 -229 -47 -217 -49 -260 -13 -295 52 -53 146 -42 177 20 16 31 87 366 87 410 0 70 -86 122 -155 94z\"\/><path d=\"M1751 3234 c-13 -9 -29 -31 -37 -50 -12 -29 -10 -49 21 -204 19 -94 39 -189 45 -210 14 -50 54 -80 110 -80 34 0 48 6 76 34 21 21 34 44 34 59 0 14 -18 113 -40 219 -37 178 -43 195 -70 221 -36 32 -101 37 -139 11z\"\/><path d=\"M1163 3073 c-36 -7 -73 -59 -73 -102 0 -56 133 -378 171 -413 34 -32 83 -37 129 -13 70 36 67 87 -16 290 -86 209 -89 214 -129 231 -35 14 -42 15 -82 7z\"\/><path d=\"M3689 3066 c-15 -9 -33 -30 -42 -48 -48 -103 -147 -355 -147 -375 0 -98 131 -148 192 -74 13 15 57 108 97 206 80 196 84 226 37 273 -30 30 -99 39 -137 18z\"\/><path d=\"M583 2784 c-38 -19 -67 -74 -58 -113 9 -42 211 -354 242 -373 16 -10 45 -18 66 -18 51 0 107 52 107 100 0 39 -1 41 -124 234 -80 126 -108 162 -133 173 -41 17 -61 16 -100 -3z\"\/><path d=\"M4250 2784 c-14 -9 -74 -91 -133 -183 -95 -150 -107 -173 -107 -213 0 -55 33 -94 87 -104 67 -13 90 8 211 198 130 202 137 225 78 284 -27 27 -42 34 -72 34 -22 0 -50 -8 -64 -16z\"\/><path d=\"M2275 2693 c-553 -48 -1095 -270 -1585 -649 -135 -104 -459 -423 -483 -476 -23 -49 -22 -139 2 -186 73 -142 361 -457 571 -626 285 -228 642 -407 990 -497 242 -63 336 -73 660 -74 310 0 370 5 595 52 535 111 1045 392 1455 803 122 121 250 273 275 326 19 41 19 137 0 174 -41 79 -309 363 -465 492 -447 370 -946 591 -1479 653 -113 14 -422 18 -536 8z m395 -428 c171 -34 330 -124 456 -258 112 -119 167 -219 211 -378 27 -96 24 -300 -5 -401 -72 -255 -236 -447 -474 -557 -132 -62 -201 -76 -368 -76 -167 0 -236 14 -368 76 -213 98 -373 271 -451 485 -162 444 86 934 547 1084 153 49 292 57 452 25z m909 -232 c222 -123 408 -262 593 -441 76 -74 138 -139 138 -144 0 -16 -233 -242 -330 -319 -155 -123 -309 -223 -461 -299 l-81 -41 32 46 c18 26 49 83 70 128 143 306 141 649 -6 957 -25 52 -61 116 -79 142 l-34 47 45 -20 c26 -10 76 -36 113 -56z m-2057 25 c-40 -58 -105 -190 -130 -263 -110 -324 -59 -707 132 -981 25 -35 42 -64 37 -64 -19 0 -241 119 -326 174 -188 122 -406 314 -532 468 l-58 71 108 103 c185 178 428 349 672 473 66 33 121 60 123 61 2 0 -10 -19 -26 -42z\"\/><path d=\"M2375 1950 c-198 -44 -350 -190 -395 -379 -18 -76 -8 -221 19 -290 114 -284 457 -406 731 -260 98 52 188 154 231 260 27 69 37 214 19 290 -38 163 -166 304 -326 360 -67 23 -215 33 -279 19z\"\/><\/g><\/svg><\/i> <img loading=\"lazy\" decoding=\"async\" width=\"16\" height=\"16\" alt=\"Loading\" src=\"https:\/\/aif.amtbbs.org\/wp-content\/plugins\/page-views-count\/ajax-loader-2x.gif\" border=0 \/><\/p>\n<div class=\"pvc_clear\"><\/div>\n","protected":false},"author":56,"featured_media":22312,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[23,17,20,80],"tags":[1328],"class_list":["post-22310","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-23","category-17","category-20","category-80","tag-ai-"],"_links":{"self":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/22310","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/users\/56"}],"replies":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/comments?post=22310"}],"version-history":[{"count":1,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/22310\/revisions"}],"predecessor-version":[{"id":22313,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/posts\/22310\/revisions\/22313"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/media\/22312"}],"wp:attachment":[{"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/media?parent=22310"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/categories?post=22310"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/aif.amtbbs.org\/index.php\/wp-json\/wp\/v2\/tags?post=22310"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}