{"id":30582,"date":"2025-04-20T12:34:48","date_gmt":"2025-04-20T04:34:48","guid":{"rendered":"https:\/\/www.aisharenet.com\/?p=30582"},"modified":"2025-04-20T12:34:48","modified_gmt":"2025-04-20T04:34:48","slug":"pinggudayuyanmoan","status":"publish","type":"post","link":"https:\/\/www.kdjingpai.com\/ja\/pinggudayuyanmoan\/","title":{"rendered":"\u8bc4\u4f30\u5927\u8bed\u8a00\u6a21\u578b\u7684\u521b\u9020\u529b\uff1a\u8d85\u8d8a\u9009\u62e9\u9898\u7684 LoTbench \u8303\u5f0f"},"content":{"rendered":"<p>\u5728\u5927\u8bed\u8a00\u6a21\u578b\uff08\u00a0<code>LLM<\/code>\u00a0\uff09\u7684\u7814\u7a76\u9886\u57df\uff0c\u6a21\u578b\u7684\u00a0<code>Leap-of-Thought<\/code>\u00a0\u80fd\u529b\uff0c\u5373\u521b\u9020\u529b\uff0c\u5176\u91cd\u8981\u6027\u4e0d\u4e9a\u4e8e\u4ee5\u00a0<code>Chain-of-Thought<\/code>\u00a0\u4e3a\u4ee3\u8868\u7684\u903b\u8f91\u63a8\u7406\u80fd\u529b\u3002\u7136\u800c\uff0c\u76ee\u524d\u9488\u5bf9\u00a0<code>LLM<\/code>\u00a0\u521b\u9020\u529b\u7684\u6df1\u5165\u8ba8\u8bba\u548c\u6709\u6548\u8bc4\u4f30\u65b9\u6cd5\u4ecd\u7136\u76f8\u5bf9\u532e\u4e4f\uff0c\u8fd9\u5728\u4e00\u5b9a\u7a0b\u5ea6\u4e0a\u5236\u7ea6\u4e86\u00a0<code>LLM<\/code>\u00a0\u5728\u521b\u610f\u5e94\u7528\u65b9\u9762\u7684\u53d1\u5c55\u6f5c\u529b\u3002<\/p>\n<p>\u9020\u6210\u8fd9\u4e00\u73b0\u72b6\u7684\u4e3b\u8981\u539f\u56e0\u5728\u4e8e\uff0c\u4e3a\u201c\u521b\u9020\u529b\u201d\u8fd9\u4e00\u62bd\u8c61\u6982\u5ff5\u6784\u5efa\u4e00\u4e2a\u5ba2\u89c2\u3001\u81ea\u52a8\u5316\u4e14\u53ef\u9760\u7684\u8bc4\u4f30\u6d41\u7a0b\u6781\u5176\u56f0\u96be\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/04\/1db383b44b90172.png\" alt=\"\u56fe 1\" \/><\/p>\n<p>\u8fc7\u53bb\u8bb8\u591a\u5bf9\u00a0<code>LLM<\/code>\u00a0\u521b\u9020\u529b\u7684\u6d4b\u8bc4\u5c1d\u8bd5\uff0c\u5982\u56fe 1 \u6240\u793a\uff0c\u4f9d\u7136\u6cbf\u7528\u8bc4\u4f30\u903b\u8f91\u601d\u7ef4\u80fd\u529b\u65f6\u5e38\u7528\u7684\u9009\u62e9\u9898\u3001\u6392\u5e8f\u9898\u7b49\u5f62\u5f0f\u3002\u8fd9\u7c7b\u65b9\u6cd5\u64c5\u957f\u8003\u5bdf\u6a21\u578b\u662f\u5426\u80fd\u8bc6\u522b\u51fa\u9884\u8bbe\u7684\u201c\u6700\u4f18\u201d\u6216\u201c\u6700\u7b26\u5408\u903b\u8f91\u201d\u7684\u9009\u9879\uff0c\u4f46\u5bf9\u4e8e\u8bc4\u4f30\u771f\u6b63\u7684\u521b\u9020\u529b\u2014\u2014\u751f\u6210\u65b0\u9896\u3001\u72ec\u7279\u5185\u5bb9\u7684\u80fd\u529b\u2014\u2014\u5219\u663e\u5f97\u529b\u4e0d\u4ece\u5fc3\u3002<\/p>\n<p>\u4f8b\u5982\uff0c\u8003\u8651\u56fe 2 \u4e2d\u7684\u4efb\u52a1\uff1a\u6839\u636e\u56fe\u7247\u548c\u5df2\u6709\u6587\u5b57\uff0c\u586b\u8865\u201c\uff1f\u201d\u5904\uff0c\u8981\u6c42\u5185\u5bb9\u5bcc\u6709\u521b\u9020\u6027\u548c\u5e7d\u9ed8\u611f\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/04\/bc0ee9c4c987213.png\" alt=\"\u56fe 2\" \/><\/p>\n<p>\u5982\u679c\u8fd9\u662f\u4e00\u4e2a\u9009\u62e9\u9898\uff0c\u63d0\u4f9b\u9009\u9879\u201cA. \u53ef\u4ee5\u5e2e\u5fd9\u6276\u4e00\u4e0b\u6211\u5417\uff1f\u201d\u548c\u201cB. \u53ef\u4ee5\u5e2e\u6211\u89e3\u5f00\u624b\u94d0\u5417\uff1f\u201d\u3002\u00a0<code>LLM<\/code>\u00a0\u5f88\u53ef\u80fd\u9009\u62e9 B\uff0c\u5e76\u975e\u56e0\u4e3a\u5b83\u5c55\u73b0\u4e86\u521b\u9020\u529b\uff0c\u800c\u4ec5\u4ec5\u662f\u56e0\u4e3a B \u9009\u9879\u76f8\u5bf9\u4e8e A \u9009\u9879\u66f4\u201c\u7279\u522b\u201d\u6216\u201c\u4e0d\u5bfb\u5e38\u201d\uff0c\u6a21\u578b\u80fd\u591f\u901a\u8fc7\u6a21\u5f0f\u8bc6\u522b\u800c\u975e\u521b\u9020\u6027\u601d\u8003\u6765\u5b8c\u6210\u9009\u62e9\u3002<\/p>\n<p>\u8bc4\u4f30\u00a0<code>LLM<\/code>\u00a0\u7684\u521b\u9020\u529b\uff0c\u6838\u5fc3\u5e94\u8be5\u662f\u8003\u5bdf\u5176<strong>\u751f\u6210<\/strong>\u521b\u65b0\u5185\u5bb9\u7684\u80fd\u529b\uff0c\u800c\u4e0d\u662f<strong>\u5224\u5b9a<\/strong>\u5185\u5bb9\u662f\u5426\u521b\u65b0\u7684\u80fd\u529b\u3002\u4f20\u7edf\u7684\u8bc4\u4f30\u65b9\u6cd5\uff0c\u5982\u591a\u9879\u9009\u62e9\uff0c\u66f4\u4fa7\u91cd\u4e8e\u540e\u8005\uff0c\u56e0\u6b64\u5b58\u5728\u5c40\u9650\u6027\u3002\u5f53\u524d\uff0c\u80fd\u591f\u76f4\u63a5\u8bc4\u4f30\u751f\u6210\u80fd\u529b\u7684\u65b9\u6cd5\u4e3b\u8981\u662f\u4eba\u5de5\u8bc4\u4f30\u548c\u00a0<code>LLM-as-a-judge<\/code>\u00a0\uff08\u4f7f\u7528\u00a0<code>LLM<\/code>\u00a0\u4f5c\u4e3a\u8bc4\u5ba1\uff09\u3002\u4eba\u5de5\u8bc4\u4f30\u867d\u7136\u51c6\u786e\u4e14\u7b26\u5408\u4eba\u7c7b\u4ef7\u503c\u89c2\uff0c\u4f46\u6210\u672c\u9ad8\u6602\u4e14\u96be\u4ee5\u89c4\u6a21\u5316\u3002\u800c\u00a0<code>LLM-as-a-judge<\/code>\u00a0\u65b9\u6cd5\u5728\u521b\u9020\u529b\u8bc4\u4f30\u4efb\u52a1\u4e0a\u7684\u8868\u73b0\u5c1a\u4e0d\u6210\u719f\uff0c\u7ed3\u679c\u7a33\u5b9a\u6027\u4e5f\u6709\u5f85\u63d0\u9ad8\u3002<\/p>\n<p>\u9762\u5bf9\u8fd9\u4e9b\u6311\u6218\uff0c\u6765\u81ea\u4e2d\u5c71\u5927\u5b66\u3001\u54c8\u4f5b\u5927\u5b66\u3001\u9e4f\u57ce\u5b9e\u9a8c\u5ba4\u548c\u65b0\u52a0\u5761\u7ba1\u7406\u5927\u5b66\u7684\u7814\u7a76\u8005\u63d0\u51fa\u4e86\u4e00\u79cd\u65b0\u7684\u601d\u8def\u3002\u4ed6\u4eec\u4e0d\u518d\u76f4\u63a5\u8bc4\u5224\u751f\u6210\u5185\u5bb9\u7684\u201c\u597d\u574f\u201d\uff0c\u800c\u662f\u901a\u8fc7\u7814\u7a76\u00a0<strong><code>LLM<\/code>\u00a0\u4ea7\u751f\u4e0e\u4eba\u7c7b\u9ad8\u8d28\u91cf\u521b\u65b0\u5185\u5bb9\u76f8\u5f53\u7684\u54cd\u5e94\u6240\u9700\u8981\u7684\u201c\u4ee3\u4ef7\u201d<\/strong>\uff08\u53ef\u4ee5\u7406\u89e3\u4e3a\u6240\u9700\u4ed8\u51fa\u7684\u52aa\u529b\u6216\u4ea4\u4e92\u6210\u672c\uff09\uff0c\u6784\u5efa\u4e86\u4e00\u4e2a\u540d\u4e3a\u00a0<code>LoTbench<\/code>\u00a0\u7684\u591a\u8f6e\u4ea4\u4e92\u5f0f\u81ea\u52a8\u5316\u521b\u9020\u529b\u8bc4\u4f30\u8303\u5f0f\u3002\u8be5\u65b9\u6cd5\u65e8\u5728\u63d0\u4f9b\u4e00\u4e2a\u66f4\u53ef\u4fe1\u3001\u53ef\u6269\u5c55\u7684\u521b\u9020\u529b\u8861\u91cf\u6807\u51c6\u3002\u76f8\u5173\u7814\u7a76\u6210\u679c\u5df2\u53d1\u8868\u4e8e\u00a0<code>IEEE TPAMI<\/code>\u00a0\u671f\u520a\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/04\/a31a5aa5fa3b2dd.png\" alt=\"\u56fe 3\" \/><\/p>\n<ul>\n<li><strong>\u8bba\u6587\u9898\u76ee\uff1a<\/strong>\u00a0A Causality-aware Paradigm for Evaluating Creativity of Multimodal Large Language Models<\/li>\n<li><strong>\u8bba\u6587\u94fe\u63a5\uff1a<\/strong>\u00a0https:\/\/arxiv.org\/abs\/2501.15147<\/li>\n<li><strong>\u9879\u76ee\u4e3b\u9875\uff1a<\/strong>\u00a0https:\/\/lotbench.github.io<\/li>\n<\/ul>\n<p>&nbsp;<\/p>\n<h2>\u4efb\u52a1\u573a\u666f\uff1a\u65e5\u5f0f\u51b7\u5410\u69fd<\/h2>\n<p><code>LoTbench<\/code>\u00a0\u7684\u7814\u7a76\u57fa\u4e8e\u00a0<code>CVPR'24<\/code>\u00a0\u4f1a\u8bae\u4e0a\u63d0\u51fa\u7684\u201c\u6897\u738b\u201d\u5927\u6a21\u578b\uff08Let&#8217;s Think Outside the Box: Exploring Leap-of-Thought in Large Language Models with Creative Humor Generation\uff09\u5de5\u4f5c\u7684\u671f\u520a\u6269\u5c55\u3002\u7814\u7a76\u8005\u9009\u62e9\u4e86\u4e00\u79cd\u6e90\u81ea\u65e5\u672c\u4f20\u7edf\u6e38\u620f\u201c\u5927\u559c\u5229\u201d\uff08Oogiri\uff09\u3001\u5728\u4e2d\u6587\u4e92\u8054\u7f51\u4e0a\u88ab\u79f0\u4e3a\u201c\u65e5\u5f0f\u51b7\u5410\u69fd\u201d\u7684\u4efb\u52a1\u5f62\u5f0f\uff0c\u5982\u56fe 2 \u6240\u793a\u3002<\/p>\n<p>\u8fd9\u79cd\u4efb\u52a1\u8981\u6c42\u53c2\u4e0e\u8005\u89c2\u770b\u56fe\u7247\uff0c\u5e76\u8865\u5168\u6587\u5b57\uff0c\u4f7f\u5f97\u56fe\u6587\u7ed3\u5408\u540e\u4ea7\u751f\u521b\u65b0\u4e14\u5e7d\u9ed8\u7684\u6548\u679c\u3002\u9009\u62e9\u8be5\u4efb\u52a1\u4f5c\u4e3a\u8bc4\u4f30\u57fa\u7840\uff0c\u4e3b\u8981\u57fa\u4e8e\u4ee5\u4e0b\u51e0\u70b9\u8003\u8651\uff1a<\/p>\n<ol>\n<li><strong>\u9ad8\u521b\u9020\u529b\u8981\u6c42\uff1a<\/strong>\u00a0\u8be5\u4efb\u52a1\u76f4\u63a5\u8981\u6c42\u751f\u6210\u521b\u610f\u5e7d\u9ed8\u5185\u5bb9\uff0c\u662f\u5178\u578b\u7684\u521b\u9020\u529b\u6311\u6218\u3002<\/li>\n<li><strong>\u5951\u5408\u591a\u6a21\u6001\u6a21\u578b\uff1a<\/strong>\u00a0\u8f93\u5165\u4e3a\u56fe\u6587\uff0c\u8f93\u51fa\u4e3a\u6587\u5b57\u8865\u5168\uff0c\u5b8c\u5168\u7b26\u5408\u5f53\u524d\u591a\u6a21\u6001\u00a0<code>LLM<\/code>\u00a0\u7684\u80fd\u529b\u8303\u7574\u3002<\/li>\n<li><strong>\u4e30\u5bcc\u7684\u6570\u636e\u8d44\u6e90\uff1a<\/strong>\u00a0\u201c\u65e5\u5f0f\u51b7\u5410\u69fd\u201d\u5728\u7f51\u7edc\u793e\u533a\u6d41\u884c\u5ea6\u9ad8\uff0c\u79ef\u7d2f\u4e86\u5927\u91cf\u9ad8\u8d28\u91cf\u7684\u4eba\u7c7b\u521b\u4f5c\u5b9e\u4f8b\u53ca\u5e26\u6709\u8bc4\u4ef7\u4fe1\u606f\u7684\u6570\u636e\uff0c\u4fbf\u4e8e\u6784\u5efa\u8bc4\u6d4b\u6570\u636e\u96c6\u3002<\/li>\n<\/ol>\n<p>\u56e0\u6b64\uff0c\u201c\u65e5\u5f0f\u51b7\u5410\u69fd\u201d\u4e3a\u8bc4\u4f30\u591a\u6a21\u6001\u00a0<code>LLM<\/code>\u00a0\u7684\u521b\u9020\u529b\u63d0\u4f9b\u4e86\u4e00\u4e2a\u7406\u60f3\u4e14\u72ec\u7279\u7684\u5e73\u53f0\u3002<\/p>\n<p>&nbsp;<\/p>\n<h2>LoTbench \u8bc4\u4f30\u65b9\u6cd5<\/h2>\n<p><img decoding=\"async\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/04\/5c0b69cb3348394.png\" alt=\"\u56fe 4\" \/><\/p>\n<p>\u4e0e\u4f20\u7edf\u8bc4\u4f30\u8303\u5f0f\uff08\u5982\u9009\u62e9\u3001\u6392\u5e8f\uff09\u4e0d\u540c\uff0c\u00a0<code>LoTbench<\/code>\u00a0\u7684\u6838\u5fc3\u601d\u60f3\u662f\uff1a<strong>\u6d4b\u91cf\u4e00\u4e2a\u00a0<code>LLM<\/code>\u00a0\u9700\u8981\u7ecf\u8fc7\u591a\u5c11\u8f6e\u4ea4\u4e92\u624d\u80fd\u751f\u6210\u4e00\u4e2a\u4e0e\u9884\u8bbe\u7684\u4eba\u7c7b\u9ad8\u8d28\u91cf\u521b\u65b0\u54cd\u5e94\uff08\u00a0<code>HHCR<\/code>\u00a0\uff09\u201c\u5f02\u66f2\u540c\u5de5\u201d\u7684\u7b54\u6848\u3002<\/strong>\u00a0\u8fd9\u4e2a\u6240\u9700\u7684\u201c\u8f6e\u6570\u201d\u53cd\u6620\u4e86\u00a0<code>LLM<\/code>\u00a0\u8fbe\u5230\u7279\u5b9a\u521b\u610f\u76ee\u6807\u7684\u201c\u8ddd\u79bb\u201d\u6216\u201c\u6210\u672c\u201d\u3002<\/p>\n<p>\u5982\u56fe 3 \u53f3\u4fa7\u6240\u793a\uff0c\u5bf9\u4e8e\u4e00\u4e2a\u7ed9\u5b9a\u7684\u00a0<code>HHCR<\/code>\u00a0\uff0c\u00a0<code>LoTbench<\/code>\u00a0\u5e76\u975e\u8981\u6c42\u00a0<code>LLM<\/code>\u00a0\u7cbe\u786e\u590d\u5236\u5b83\uff0c\u800c\u662f\u770b\u00a0<code>LLM<\/code>\u00a0\u80fd\u5426\u5728\u591a\u8f6e\u5c1d\u8bd5\u4e2d\uff0c\u751f\u6210\u4e00\u4e2a\u867d\u7136\u8868\u8fbe\u65b9\u5f0f\u4e0d\u540c\u3001\u4f46\u521b\u610f\u6838\u5fc3\u548c\u6548\u679c\u76f8\u4f3c\uff08\u5373\u00a0<code>DAESO<\/code>\u00a0&#8211; Different Approach but Equally Satisfactory Outcome\uff09\u7684\u54cd\u5e94\u3002<\/p>\n<p><code>LoTbench<\/code>\u00a0\u7684\u5177\u4f53\u6d41\u7a0b\u5982\u56fe 4 \u6240\u793a\uff1a<\/p>\n<ol>\n<li><strong>\u4efb\u52a1\u6784\u5efa\uff1a<\/strong>\u00a0\u4ece\u201c\u65e5\u5f0f\u51b7\u5410\u69fd\u201d\u6570\u636e\u4e2d\u7cbe\u9009\u00a0<code>HHCR<\/code>\u00a0\u6837\u672c\u3002\u6bcf\u4e00\u8f6e\uff0c\u8981\u6c42\u5f85\u6d4b\u00a0<code>LLM<\/code>\u00a0\u6839\u636e\u56fe\u6587\u4fe1\u606f\uff0c\u751f\u6210\u4e00\u4e2a\u54cd\u5e94\u00a0<code>Rt<\/code>\u00a0\u6765\u8865\u5168\u6587\u5b57\u7a7a\u7f3a\u3002<\/li>\n<li><strong>DAESO \u5224\u65ad\uff1a<\/strong>\u00a0\u5224\u5b9a\u751f\u6210\u7684\u00a0<code>Rt<\/code>\u00a0\u662f\u5426\u4e0e\u76ee\u6807\u00a0<code>HHCR<\/code>\u00a0\uff08\u8bb0\u4e3a\u00a0<code>R<\/code>\u00a0\uff09\u8fbe\u5230\u4e86\u00a0<code>DAESO<\/code>\u00a0\u3002\u5982\u679c\u662f\uff0c\u5219\u8bb0\u5f55\u5f53\u524d\u8f6e\u6570\uff0c\u7528\u4e8e\u540e\u7eed\u8ba1\u7b97\u5206\u6570\uff1b\u5982\u679c\u5426\uff0c\u5219\u8fdb\u5165\u6b65\u9aa4 3\u3002<\/li>\n<li><strong>\u4ea4\u4e92\u5f0f\u63d0\u95ee\uff1a<\/strong>\u00a0\u82e5\u672a\u8fbe\u5230\u00a0<code>DAESO<\/code>\u00a0\uff0c\u5219\u8981\u6c42\u5f85\u6d4b\u00a0<code>LLM<\/code>\u00a0\u6839\u636e\u5f53\u524d\u7684\u4ea4\u4e92\u5386\u53f2\uff0c\u63d0\u51fa\u4e00\u4e2a\u4e00\u822c\u7591\u95ee\u53e5\u00a0<code>Qt<\/code>\u00a0\uff08\u4f8b\u5982\uff0c\u8be2\u95ee\u5173\u4e8e\u76ee\u6807\u521b\u610f\u65b9\u5411\u7684\u7ebf\u7d22\uff09\u3002<\/li>\n<li><strong>\u7cfb\u7edf\u53cd\u9988\uff1a<\/strong>\u00a0\u8bc4\u6d4b\u7cfb\u7edf\u6839\u636e\u00a0<code>HHCR<\/code>\u00a0\u7684\u5185\u5728\u903b\u8f91\uff0c\u5bf9\u00a0<code>LLM<\/code>\u00a0\u63d0\u51fa\u7684\u95ee\u9898\u00a0<code>Qt<\/code>\u00a0\u56de\u7b54\u201c\u662f\u201d\u6216\u201c\u5426\u201d\u3002<\/li>\n<li><strong>\u4fe1\u606f\u6574\u5408\u4e0e\u8fed\u4ee3\uff1a<\/strong>\u00a0\u5c06\u672c\u8f6e\u7684\u6240\u6709\u4ea4\u4e92\u4fe1\u606f\uff08\u5305\u62ec\u00a0<code>LLM<\/code>\u00a0\u7684\u751f\u6210\u3001\u63d0\u95ee\u3001\u7cfb\u7edf\u7684\u53cd\u9988\uff09\u4ee5\u53ca\u7cfb\u7edf\u63d0\u4f9b\u7684\u63d0\u793a\u6574\u5408\uff0c\u5f62\u6210\u4e0b\u4e00\u8f6e\u7684\u00a0<code>history prompt<\/code>\u00a0\uff0c\u8fd4\u56de\u6b65\u9aa4 1\uff0c\u5f00\u59cb\u65b0\u4e00\u8f6e\u7684\u5c1d\u8bd5\u3002<\/li>\n<\/ol>\n<p>\u8fd9\u4e2a\u8fc7\u7a0b\u6301\u7eed\u8fdb\u884c\uff0c\u76f4\u5230\u00a0<code>LLM<\/code>\u00a0\u751f\u6210\u4e86\u00a0<code>DAESO<\/code>\u00a0\u54cd\u5e94\uff0c\u6216\u8005\u8fbe\u5230\u4e86\u9884\u8bbe\u7684\u6700\u5927\u8f6e\u6570\u4e0a\u9650\u3002<\/p>\n<p>\u6700\u7ec8\u7684\u521b\u9020\u529b\u5206\u6570\u00a0<code>Sc<\/code>\u00a0\u57fa\u4e8e\u5bf9\u00a0<code>n<\/code>\u00a0\u4e2a\u00a0<code>HHCR<\/code>\u00a0\u6837\u672c\u3001\u8fdb\u884c\u00a0<code>m<\/code>\u00a0\u6b21\u91cd\u590d\u5b9e\u9a8c\u7684\u7ed3\u679c\u8ba1\u7b97\u5f97\u51fa\u3002\u5176\u8ba1\u7b97\u65b9\u5f0f\u5927\u81f4\u5982\u4e0b\uff08\u4ee5 HTML \u516c\u5f0f\u8868\u793a\uff09\uff1a<\/p>\n<p>S<sub>c<\/sub>\u00a0=\u00a0(\u00a01\u00a0\/\u00a0n\u00a0)\u00a0\u2211<sub>i=1<\/sub><sup>n<\/sup>\u00a0[\u00a0(\u00a01\u00a0\/\u00a0m\u00a0)\u00a0\u2211<sub>j=1<\/sub><sup>m<\/sup>\u00a0(\u00a01\u00a0\/\u00a0(\u00a01 +\u00a0k<sub>ij<\/sub>\u00a0)\u00a0)\u00a0]<\/p>\n<p>\u5176\u4e2d\uff0c<code>k_ij<\/code>\u00a0\u662f\u6a21\u578b\u5728\u7b2c\u00a0<code>j<\/code>\u00a0\u6b21\u91cd\u590d\u5b9e\u9a8c\u4e2d\uff0c\u9488\u5bf9\u7b2c\u00a0<code>i<\/code>\u00a0\u4e2a\u00a0<code>HHCR<\/code>\u00a0\u6837\u672c\uff0c\u6210\u529f\u751f\u6210\u00a0<code>DAESO<\/code>\u00a0\u54cd\u5e94\u6240\u7528\u7684\u8f6e\u6570\u3002<\/p>\n<p>\u8fd9\u4e2a\u521b\u9020\u529b\u5206\u6570\u00a0<code>Sc<\/code>\u00a0\u5177\u6709\u4ee5\u4e0b\u7279\u70b9\uff1a<\/p>\n<ol>\n<li><strong>\u53cd\u6bd4\u5173\u7cfb\uff1a<\/strong>\u00a0\u5206\u6570\u4e0e\u6240\u9700\u8f6e\u6570\u00a0<code>k<\/code>\u00a0\u6210\u53cd\u6bd4\u3002\u8f6e\u6570\u8d8a\u5c11\uff0c\u8868\u660e\u00a0<code>LLM<\/code>\u00a0\u8d8a\u5feb\u8fbe\u5230\u76ee\u6807\u521b\u610f\u6c34\u5e73\uff0c\u5f97\u5206\u8d8a\u9ad8\uff0c\u521b\u9020\u529b\u8d8a\u5f3a\u3002<\/li>\n<li><strong>\u96f6\u5206\u4e0b\u9650\uff1a<\/strong>\u00a0\u5982\u679c\u00a0<code>LLM<\/code>\u00a0\u5728\u6700\u5927\u8f6e\u6570\u9650\u5236\u5185\u59cb\u7ec8\u65e0\u6cd5\u751f\u6210\u00a0<code>DAESO<\/code>\u00a0\u54cd\u5e94\uff08\u76f8\u5f53\u4e8e\u8f6e\u6570\u8d8b\u4e8e\u65e0\u9650\uff09\uff0c\u5176\u5bf9\u5e94\u8be5\u6837\u672c\u7684\u5206\u6570\u8d8b\u8fd1\u4e8e 0\uff0c\u8868\u793a\u5728\u6b64\u4efb\u52a1\u4e0a\u521b\u9020\u529b\u4e0d\u8db3\u3002<\/li>\n<li><strong>\u9c81\u68d2\u6027\uff1a<\/strong>\u00a0\u901a\u8fc7\u5bf9\u591a\u4e2a\u00a0<code>HHCR<\/code>\u00a0\u6837\u672c\u8fdb\u884c\u591a\u6b21\u91cd\u590d\u5b9e\u9a8c\u53d6\u5e73\u5747\uff0c\u5206\u6570\u8003\u8651\u4e86\u521b\u610f\u7684\u591a\u6837\u6027\u548c\u96be\u5ea6\uff0c\u51cf\u5c11\u4e86\u5355\u6b21\u5b9e\u9a8c\u7684\u968f\u673a\u6027\u5f71\u54cd\u3002<\/li>\n<\/ol>\n<p>&nbsp;<\/p>\n<h2>\u5982\u4f55\u5224\u65ad\u201c\u5f02\u66f2\u540c\u5de5\u4e4b\u5999\u201d\uff08\u00a0<code>DAESO<\/code>\u00a0\uff09\uff1f<\/h2>\n<p><code>DAESO<\/code>\u00a0\u7684\u5224\u5b9a\u662f\u00a0<code>LoTbench<\/code>\u00a0\u65b9\u6cd5\u7684\u6838\u5fc3\u96be\u70b9\u4e4b\u4e00\u3002<\/p>\n<p><strong>\u4e3a\u4f55\u9700\u8981\u00a0<code>DAESO<\/code>\u00a0\u5224\u5b9a\uff1f<\/strong>\u00a0\u521b\u9020\u529b\u4efb\u52a1\u7684\u5173\u952e\u7279\u5f81\u4e4b\u4e00\u662f\u5176\u5f00\u653e\u6027\u548c\u591a\u6837\u6027\u3002\u5bf9\u4e8e\u540c\u4e00\u4e2a\u201c\u65e5\u5f0f\u51b7\u5410\u69fd\u201d\u573a\u666f\uff0c\u4eba\u7c7b\u53ef\u4ee5\u60f3\u51fa\u5f88\u591a\u79cd\u4e0d\u540c\u4f46\u540c\u6837\u5177\u6709\u521b\u610f\u548c\u5e7d\u9ed8\u611f\u7684\u7b54\u6848\u3002\u5982\u56fe 5 \u6240\u793a\uff0c\u201c\u6709\u6d3b\u529b\u7684\u95f9\u949f\u201d\u548c\u201c\u6709\u6d3b\u529b\u7684\u624b\u673a\u201d\u90fd\u56f4\u7ed5\u201c\u7269\u54c1\u56e0\u5145\u6ee1\u6d3b\u529b\u800c\u8df3\u52a8\u5e76\u53d1\u51fa\u58f0\u97f3\u201d\u8fd9\u4e00\u6838\u5fc3\u521b\u610f\u70b9\uff0c\u8fbe\u5230\u4e86\u76f8\u4f3c\u7684\u5e7d\u9ed8\u6548\u679c\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/04\/c212242c62391f6.png\" alt=\"\u56fe 5\" \/><\/p>\n<p>\u7b80\u5355\u5730\u901a\u8fc7\u6587\u5b57\u8868\u9762\u5339\u914d\u6216\u5e38\u89c4\u7684\u8bed\u4e49\u76f8\u4f3c\u5ea6\u8ba1\u7b97\uff0c\u65e0\u6cd5\u51c6\u786e\u6355\u6349\u8fd9\u79cd\u6df1\u5c42\u6b21\u7684\u521b\u610f\u76f8\u4f3c\u6027\u3002\u4f8b\u5982\uff0c\u201c\u6709\u6d3b\u529b\u7684\u8df3\u86a4\u201d\u867d\u7136\u4e5f\u6709\u201c\u6d3b\u529b\u201d\uff0c\u4f46\u7f3a\u5c11\u4e86\u201c\u95f9\u949f\u201d\u6216\u201c\u624b\u673a\u201d\u6240\u6697\u793a\u7684\u201c\u53d1\u51fa\u58f0\u97f3\u63d0\u9192\u201d\u7684\u529f\u80fd\u5173\u8054\u3002\u56e0\u6b64\uff0c\u5fc5\u987b\u5f15\u5165\u5bf9\u201c\u5f02\u66f2\u540c\u5de5\u4e4b\u5999\u201d\u7684\u5224\u5b9a\u673a\u5236\u3002<\/p>\n<p><strong>\u5982\u4f55\u5b9e\u73b0\u00a0<code>DAESO<\/code>\u00a0\u5224\u5b9a\uff1f<\/strong><\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/04\/f1f6b9a01f098e2.png\" alt=\"\u56fe 6\" \/><\/p>\n<p>\u7814\u7a76\u8005\u5728\u8bba\u6587\u4e2d\u63d0\u51fa\uff0c\u4e24\u4e2a\u54cd\u5e94\u82e5\u8981\u6ee1\u8db3\u00a0<code>DAESO<\/code>\u00a0\uff0c\u9700\u8981\u540c\u65f6\u6ee1\u8db3\u4e24\u4e2a\u6761\u4ef6\uff1a<\/p>\n<ol>\n<li><strong>\u76f8\u540c\u7684\u6838\u5fc3\u521b\u65b0\u89e3\u91ca\uff1a<\/strong>\u00a0\u4e24\u4e2a\u54cd\u5e94\u80cc\u540e\u7684\u521b\u610f\u903b\u8f91\u6216\u5e7d\u9ed8\u70b9\u662f\u57fa\u672c\u4e00\u81f4\u7684\u3002<\/li>\n<li><strong>\u76f8\u540c\u7684\u529f\u80fd\u76f8\u4f3c\u6027\uff1a<\/strong>\u00a0\u4e24\u4e2a\u54cd\u5e94\u5728\u5f15\u53d1\u5e7d\u9ed8\u7684\u201c\u529f\u80fd\u201d\u6216\u201c\u573a\u666f\u89d2\u8272\u201d\u4e0a\u662f\u76f8\u4f3c\u7684\u3002<\/li>\n<\/ol>\n<p>\u529f\u80fd\u76f8\u4f3c\u6027\u4e0d\u540c\u4e8e\u7eaf\u7cb9\u7684\u8bed\u4e49\u76f8\u4f3c\u6027\u3002\u5982\u56fe 6(a) \u7684\u4f8b\u5b50\u6240\u793a\uff0c\u5728\u201c\u7838\u6838\u6843\u201d\u8fd9\u4e2a\u7279\u5b9a\u529f\u80fd\u573a\u666f\u4e0b\uff0c\u201c\u8bfa\u57fa\u4e9a\u624b\u673a\u201d\u4e0e\u201c\u9524\u5b50\u201d\u7684\u529f\u80fd\u76f8\u4f3c\u6027\uff0c\u53ef\u80fd\u9ad8\u4e8e\u5b83\u4e0e\u201c\u4e09\u661f\u624b\u673a\u201d\u7684\u8bed\u4e49\u76f8\u4f3c\u6027\u3002<\/p>\n<p>\u4ec5\u6ee1\u8db3\u6838\u5fc3\u521b\u65b0\u89e3\u91ca\u76f8\u540c\uff0c\u53ef\u80fd\u5bfc\u81f4\u54cd\u5e94\u504f\u79bb\u4e3b\u9898\uff08\u5982\u56fe 5 \u4f8b\u5b50\u4e2d\u7684\u201c\u6709\u6d3b\u529b\u7684\u8df3\u86a4\u201d\uff0c\u7f3a\u5c11\u201c\u53d1\u58f0\u63d0\u9192\u201d\u529f\u80fd\uff09\uff1b\u4ec5\u6ee1\u8db3\u529f\u80fd\u76f8\u4f3c\u6027\u76f8\u540c\uff0c\u5219\u53ef\u80fd\u672a\u80fd\u6293\u4f4f\u521b\u610f\u7684\u6838\u5fc3\uff08\u5982\u56fe 5 \u4f8b\u5b50\u4e2d\u7684\u201c\u6709\u6d3b\u529b\u7684\u9f13\u201d\uff0c\u540c\u6837\u662f\u53d1\u58f0\u7269\u4f53\uff0c\u4f46\u7f3a\u5c11\u4e86\u56e0\u81ea\u8eab\u201c\u6d3b\u529b\u201d\u800c\u8df3\u52a8\u7684\u611f\u89c9\uff09\u3002<\/p>\n<p>\u5728\u5177\u4f53\u7684\u00a0<code>DAESO<\/code>\u00a0\u5224\u65ad\u5b9e\u73b0\u4e2d\uff0c\u7814\u7a76\u8005\u9996\u5148\u4e3a\u6bcf\u4e2a\u00a0<code>HHCR<\/code>\u00a0\u6837\u672c\u6807\u6ce8\u4e86\u8be6\u7ec6\u7684\u89e3\u91ca\uff0c\u8bf4\u660e\u5176\u5e7d\u9ed8\u548c\u521b\u610f\u7684\u6765\u6e90\u3002\u7136\u540e\uff0c\u7ed3\u5408\u56fe\u50cf\u7684\u6807\u9898\uff08caption\uff09\u4fe1\u606f\uff0c\u5229\u7528\u00a0<code>LLM<\/code>\u00a0\u672c\u8eab\u7684\u80fd\u529b\uff0c\u5728\u6587\u672c\u7a7a\u95f4\u4e2d\u4e3a\u00a0<code>HHCR<\/code>\u00a0\u6784\u5efa\u56e0\u679c\u94fe\u6761\uff08\u5982\u56fe 6(c) \u6240\u793a\uff09\uff0c\u89e3\u6790\u5176\u521b\u610f\u6784\u6210\u3002\u6700\u540e\uff0c\u8bbe\u8ba1\u7279\u5b9a\u7684\u6307\u4ee4\uff08instruction\uff09\uff0c\u8ba9\u53e6\u4e00\u4e2a\u00a0<code>LLM<\/code>\u00a0\uff08\u5982\u00a0<code>GPT-4o mini<\/code>\u00a0\uff09\u6839\u636e\u8fd9\u4e9b\u4fe1\u606f\uff0c\u5728\u6587\u672c\u7a7a\u95f4\u4e2d\u5224\u65ad\u5f85\u6d4b\u54cd\u5e94\u00a0<code>Rt<\/code>\u00a0\u4e0e\u76ee\u6807\u00a0<code>HHCR<\/code>\u00a0\u662f\u5426\u540c\u65f6\u6ee1\u8db3\u4e0a\u8ff0\u4e24\u4e2a\u00a0<code>DAESO<\/code>\u00a0\u6761\u4ef6\u3002<\/p>\n<p>\u7814\u7a76\u8868\u660e\uff0c\u4f7f\u7528\u00a0<code>GPT-4o mini<\/code>\u00a0\u8fdb\u884c\u00a0<code>DAESO<\/code>\u00a0\u5224\u65ad\uff0c\u53ef\u4ee5\u5728\u8f83\u4f4e\u7684\u8ba1\u7b97\u6210\u672c\u4e0b\u8fbe\u5230 80%-90% \u7684\u51c6\u786e\u7387\u3002\u8003\u8651\u5230\u00a0<code>LoTbench<\/code>\u00a0\u4f1a\u8fdb\u884c\u591a\u6b21\u91cd\u590d\u5b9e\u9a8c\uff0c\u5355\u6b21\u00a0<code>DAESO<\/code>\u00a0\u5224\u65ad\u7684\u5fae\u5c0f\u8bef\u5dee\u5bf9\u6700\u7ec8\u5e73\u5747\u5f97\u5206\u7684\u5f71\u54cd\u4f1a\u8fdb\u4e00\u6b65\u51cf\u5c0f\uff0c\u4ece\u800c\u4fdd\u8bc1\u4e86\u6574\u4f53\u8bc4\u4f30\u7684\u53ef\u9760\u6027\u3002<\/p>\n<p>&nbsp;<\/p>\n<h2>\u6d4b\u8bc4\u7ed3\u679c<\/h2>\n<p><img decoding=\"async\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/04\/927b49e7f282793.png\" alt=\"\u56fe 7\" \/><\/p>\n<p>\u7814\u7a76\u56e2\u961f\u4f7f\u7528\u00a0<code>LoTbench<\/code>\u00a0\u5bf9\u5f53\u524d\u4e00\u4e9b\u4e3b\u6d41\u7684\u591a\u6a21\u6001\u00a0<code>LLM<\/code>\u00a0\u8fdb\u884c\u4e86\u6d4b\u8bc4\u3002\u5982\u56fe 7 \u6240\u793a\uff0c\u7ed3\u679c\u663e\u793a\uff0c\u4ee5\u00a0<code>LoTbench<\/code>\u00a0\u7684\u6807\u51c6\u8861\u91cf\uff0c\u73b0\u6709\u00a0<code>LLM<\/code>\u00a0\u7684\u521b\u9020\u529b\u666e\u904d\u4e0d\u7b97\u5f3a\uff0c\u4e0e\u4eba\u7c7b\u9ad8\u8d28\u91cf\u521b\u610f\u54cd\u5e94\uff08\u00a0<code>HHCR<\/code>\u00a0\uff09\u76f8\u6bd4\u4ecd\u6709\u5dee\u8ddd\u3002\u7136\u800c\uff0c\u4e0e\u666e\u901a\u4eba\u7c7b\u6c34\u5e73\uff08\u56fe\u4e2d\u672a\u660e\u786e\u6807\u51fa\uff0c\u4f46\u53ef\u63a8\u65ad\uff09\u6216\u521d\u7ea7\u4eba\u7c7b\u6c34\u5e73\u76f8\u6bd4\uff0c\u90e8\u5206\u9876\u5c16\u00a0<code>LLM<\/code>\u00a0\uff08\u5982\u00a0<code><a href=\"https:\/\/www.kdjingpai.com\/geminibardgubu\/\">Gemini<\/a> 1.5 Pro<\/code>\u00a0\u548c\u00a0<code>Qwen-VL-max<\/code>\u00a0\uff09\u5df2\u7ecf\u5c55\u73b0\u51fa\u4e00\u5b9a\u7684\u7ade\u4e89\u529b\uff0c\u4e5f\u6697\u793a\u4e86\u00a0<code>LLM<\/code>\u00a0\u5728\u521b\u9020\u529b\u65b9\u9762\u5177\u5907\u8d85\u8d8a\u4eba\u7c7b\u7684\u6f5c\u529b\u3002<\/p>\n<p>\u56fe 8 \u53ef\u89c6\u5316\u4e86\u699c\u5355\u4e2d\u6392\u540d\u524d\u4e8c\u7684\u00a0<code>Gemini 1.5 Pro<\/code>\u00a0\u548c\u00a0<code>Qwen-VL-max<\/code>\u00a0\u6a21\u578b\u9488\u5bf9\u90e8\u5206\u00a0<code>HHCR<\/code>\u00a0\uff08\u7ea2\u8272\u6807\u6ce8\uff09\u751f\u6210\u7684\u00a0<code>DAESO<\/code>\u00a0\u54cd\u5e94\uff08\u84dd\u8272\u6807\u6ce8\uff09\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/04\/6670e24777c935b.png\" alt=\"\u56fe 8\" \/><\/p>\n<p>\u503c\u5f97\u4e00\u63d0\u7684\u662f\uff0c\u8fd1\u671f\u5907\u53d7\u5173\u6ce8\u7684\u00a0<code><a href=\"https:\/\/www.kdjingpai.com\/deepseek-vl2\/\">DeepSeek-VL2<\/a><\/code>\u00a0\u548c\u00a0<code>Janus-Pro-7B<\/code>\u00a0\u7cfb\u5217\u6a21\u578b\u4e5f\u63a5\u53d7\u4e86\u8bc4\u4f30\u3002\u7ed3\u679c\u8868\u660e\uff0c\u5b83\u4eec\u7684\u521b\u9020\u529b\u5728\u00a0<code>LoTbench<\/code>\u00a0\u6846\u67b6\u4e0b\uff0c\u5927\u81f4\u5904\u4e8e\u4eba\u7c7b\u521d\u7ea7\u9636\u6bb5\u7684\u6c34\u5e73\u3002\u8fd9\u8868\u660e\u5728\u63d0\u5347\u591a\u6a21\u6001\u00a0<code>LLM<\/code>\u00a0\u7684\u6df1\u5ea6\u521b\u9020\u529b\u65b9\u9762\uff0c\u4ecd\u6709\u76f8\u5f53\u5927\u7684\u63a2\u7d22\u7a7a\u95f4\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u5728\u5927\u8bed\u8a00\u6a21\u578b\uff08\u00a0LLM\u00a0\uff09\u7684\u7814\u7a76\u9886\u57df\uff0c\u6a21\u578b\u7684\u00a0Leap-of-Thought\u00a0\u80fd\u529b\uff0c\u5373\u521b\u9020\u529b\uff0c\u5176\u91cd\u8981\u6027\u4e0d\u4e9a\u4e8e\u4ee5\u00a0Chain-of-Thought\u00a0\u4e3a\u4ee3\u8868\u7684\u903b\u8f91\u63a8\u7406\u80fd\u529b\u3002\u7136\u800c\uff0c\u76ee\u524d\u9488\u5bf9\u00a0LLM\u00a0\u521b\u9020\u529b\u7684\u6df1\u5165\u8ba8\u8bba\u548c\u6709\u6548\u8bc4\u4f30\u65b9\u6cd5\u4ecd\u7136\u76f8\u5bf9\u532e\u4e4f\uff0c\u8fd9\u5728&#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[34],"tags":[],"class_list":["post-30582","post","type-post","status-publish","format-standard","hentry","category-knowledge"],"_links":{"self":[{"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/posts\/30582","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/comments?post=30582"}],"version-history":[{"count":0,"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/posts\/30582\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/media?parent=30582"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/categories?post=30582"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/ja\/wp-json\/wp\/v2\/tags?post=30582"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}