{"id":921,"date":"2026-01-15T15:00:07","date_gmt":"2026-01-15T07:00:07","guid":{"rendered":"https:\/\/1ow.top\/index.php\/2026\/01\/15\/%e5%a7%9a%e7%8f%ad%e4%bc%a0%e5%a5%87%e9%99%88%e7%ab%8b%e6%9d%b0%e5%85%a5%e8%81%8copenai%ef%bc%8116%e5%b2%81%e4%bf%9d%e9%80%81%e6%b8%85%e5%8d%8e%ef%bc%8c30%e5%b2%81%e6%8b%bf%e4%b8%8buc%e4%bc%af\/"},"modified":"2026-01-15T15:00:07","modified_gmt":"2026-01-15T07:00:07","slug":"%e5%a7%9a%e7%8f%ad%e4%bc%a0%e5%a5%87%e9%99%88%e7%ab%8b%e6%9d%b0%e5%85%a5%e8%81%8copenai%ef%bc%8116%e5%b2%81%e4%bf%9d%e9%80%81%e6%b8%85%e5%8d%8e%ef%bc%8c30%e5%b2%81%e6%8b%bf%e4%b8%8buc%e4%bc%af","status":"publish","type":"post","link":"https:\/\/1ow.top\/index.php\/2026\/01\/15\/%e5%a7%9a%e7%8f%ad%e4%bc%a0%e5%a5%87%e9%99%88%e7%ab%8b%e6%9d%b0%e5%85%a5%e8%81%8copenai%ef%bc%8116%e5%b2%81%e4%bf%9d%e9%80%81%e6%b8%85%e5%8d%8e%ef%bc%8c30%e5%b2%81%e6%8b%bf%e4%b8%8buc%e4%bc%af\/","title":{"rendered":"\u59da\u73ed\u4f20\u5947\u9648\u7acb\u6770\u5165\u804cOpenAI\uff0116\u5c81\u4fdd\u9001\u6e05\u534e\uff0c30\u5c81\u62ff\u4e0bUC\u4f2f\u514b\u5229\u52a9\u7406\u6559\u6388"},"content":{"rendered":"
\"\u59da\u73ed\u4f20\u5947\u9648\u7acb\u6770\u5165\u804cOpenAI\uff0116\u5c81\u4fdd\u9001\u6e05\u534e\uff0c30\u5c81\u62ff\u4e0bUC\u4f2f\u514b\u5229\u52a9\u7406\u6559\u6388\"\/<\/div><\/figure>\n
\"\u59da\u73ed\u4f20\u5947\u9648\u7acb\u6770\u5165\u804cOpenAI\uff0116\u5c81\u4fdd\u9001\u6e05\u534e\uff0c30\u5c81\u62ff\u4e0bUC\u4f2f\u514b\u5229\u52a9\u7406\u6559\u6388\"\/<\/div><\/figure>\n
\"\u59da\u73ed\u4f20\u5947\u9648\u7acb\u6770\u5165\u804cOpenAI\uff0116\u5c81\u4fdd\u9001\u6e05\u534e\uff0c30\u5c81\u62ff\u4e0bUC\u4f2f\u514b\u5229\u52a9\u7406\u6559\u6388\"\/<\/div><\/figure>\n
\"\u59da\u73ed\u4f20\u5947\u9648\u7acb\u6770\u5165\u804cOpenAI\uff0116\u5c81\u4fdd\u9001\u6e05\u534e\uff0c30\u5c81\u62ff\u4e0bUC\u4f2f\u514b\u5229\u52a9\u7406\u6559\u6388\"\/<\/div><\/figure>\n

henry \u53d1\u81ea \u51f9\u975e\u5bfa<\/p>\n

\u6700\u65b0\u6d88\u606f\uff1a\u59da\u73ed\u5927\u795e\u9648\u7acb\u6770\uff0c\u52a0\u76dfOpenAI\u4e86\u3002<\/p>\n

\u636e\u201cTop\u534e\u4eba\u793e\u6d88\u606f\u201d\uff0cOpenAI\u5185\u90e8\u786e\u8ba4\uff1a\u6e05\u534e\u59da\u73ed\u5929\u624d\u3001UC\u4f2f\u514b\u5229EECS\u52a9\u7406\u6559\u6388\u9648\u7acb\u6770\u5df2\u52a0\u76dfOpenAI\uff0c\u8d1f\u8d23\u6570\u5b66\u63a8\u7406\uff01<\/p>\n

\u503c\u5f97\u4e00\u63d0\u7684\u662f\uff0cOpenAI \u5728\u53bb\u5e74 9 \u6708\u53d1\u8868\u7684\u51fa\u5708\u8bba\u6587\u300aWhy Language Models Hallucinate\u300b\u4e2d\uff0c\u4e5f\u5f15\u7528\u4e86\u9648\u7acb\u6770\u53c2\u4e0e\u7684\u53e6\u4e00\u7bc7\u7814\u7a76\u300aWhy and How LLMs Hallucinate: Connecting the Dots with Subsequence Associations\u300b\u3002<\/p>\n

\u4e0e\u6b64\u540c\u65f6\uff0c\u9648\u7acb\u6770\u8fd1\u671f\u53c2\u4e0e\u7684\u6700\u65b0\u7814\u7a76\u65b9\u5411\u4e5f\u5341\u5206\u201c\u5f53\u4e0b\u201d\uff0c\u805a\u7126\u4e8e\u6269\u6563\u8bed\u8a00\u6a21\u578b\uff08Diffusion Language Models\uff09\uff0c\u7d27\u8ddf\u5f53\u524d\u751f\u6210\u6a21\u578b\u7684\u91cd\u8981\u6f14\u8fdb\u8def\u7ebf\u3002<\/p>\n

\u622a\u81f3\u76ee\u524d\uff0c\u9648\u7acb\u6770\u4e3b\u9875\u672a\u6709\u66f4\u65b0\u3002<\/p>\n

\u9648\u7acb\u6770\u51fa\u751f\u4e8e1995\u5e74\uff0c16\u5c81\u65f6\u83b7\u5f97\u5168\u56fd\u4fe1\u606f\u5b66\u5965\u8d5b\u91d1\u724c\uff08NOI\uff09\uff0c\u88ab\u4fdd\u9001\u8fdb\u5165\u6e05\u534e\u5927\u5b66\uff0c\u662f\u6e05\u534e\u5927\u5b66 \u201c\u59da\u73ed\u201d \u7684\u77e5\u540d\u6821\u53cb\uff0c\u957f\u671f\u4ece\u4e8b\u7406\u8bba\u8ba1\u7b97\u673a\u79d1\u5b66\u7814\u7a76\u3002<\/p>\n

2025\u5e74\uff0c\u9648\u7acb\u6770\u6b63\u5f0f\u5165\u804c\u52a0\u5dde\u5927\u5b66\u4f2f\u514b\u5229\u5206\u6821\uff08UC Berkeley\uff09\u7535\u6c14\u5de5\u7a0b\u4e0e\u8ba1\u7b97\u673a\u79d1\u5b66\u7cfb\uff08EECS\uff09\uff0c\u62c5\u4efb\u52a9\u7406\u6559\u6388\uff0c\u5e76\u6210\u4e3a\u4f2f\u514b\u5229\u7406\u8bba\u8ba1\u7b97\u673a\u79d1\u5b66\u56e2\u961f\uff08Berkeley Theory Group\uff09\u6210\u5458\uff0c\u4e3b\u8981\u4ece\u4e8b\u8ba1\u7b97\u590d\u6742\u6027\u7406\u8bba\u76f8\u5173\u7814\u7a76\u3002<\/p>\n

\u56de\u987e\u9648\u7acb\u6770\u7684\u6559\u80b2\u7ecf\u5386\u548c\u804c\u4e1a\u751f\u6daf\uff0c\u582a\u79f0\u5f00\u6302\u822c\u7684\u7ecf\u5386\u3002<\/p>\n

\u4ed6\u81ea\u521d\u4e2d\u8d77\u53c2\u52a0\u4fe1\u606f\u5b66\u7ade\u8d5b\uff0c\u662f\u4fe1\u606f\u5b66\u5965\u8d5b\uff08OI\uff09\u5708\u5185\u7684\u4f20\u5947\u9009\u624b\u4e4b\u4e00\uff1a<\/p>\n

2013 \u5e74\uff0c\u9648\u7acb\u6770\u4ece\u676d\u5dde\u5916\u56fd\u8bed\u5b66\u6821\u6bd5\u4e1a\uff0c\u9ad8\u4e09\u65f6\u66fe\u4ee5\u4e13\u6ce8\u5b66\u4e1a\u4e3a\u7531\u62d2\u7edd\u8c37\u6b4c\u5b9e\u4e60\u9080\u8bf7\u3002\u540c\u5e74\uff0c\u4ed6\u51ed\u501f\u7ade\u8d5b\u6210\u7ee9\u83b7\u5f97\u6e05\u534e\u5927\u5b66\u4fdd\u9001\u8d44\u683c\u3002<\/p>\n

\u8fdb\u5165\u6e05\u534e\u5927\u5b66\u59da\u73ed\u540e\uff0c\u9648\u7acb\u6770\u9010\u6b65\u5c06\u91cd\u5fc3\u4ece\u7ade\u8d5b\u8f6c\u5411\u79d1\u7814\u3002<\/p>\n

\u672c\u79d1\u671f\u95f4\uff0c\u4ed6\u5728AAAI\u3001AAMAS\u3001COLT\u3001CCC\u7b49\u8ba1\u7b97\u673a\u9886\u57df\u91cd\u8981\u4f1a\u8bae\u4e0a\u53d1\u8868\u591a\u7bc7\u8bba\u6587\uff0c\u5e76\u5f00\u59cb\u7cfb\u7edf\u6027\u5730\u6295\u5165 \u8ba1\u7b97\u590d\u6742\u6027\u7406\u8bba\u7814\u7a76\u3002<\/p>\n

\u5927\u4e09\u4e0b\u5b66\u671f\uff0c\u4ed6\u524d\u5f80MIT\u4ea4\u6d41\u5b66\u4e60\uff0c\u5e08\u4ece\u8457\u540d\u7406\u8bba\u8ba1\u7b97\u673a\u4e0e\u91cf\u5b50\u4fe1\u606f\u5b66\u8005Scott Aaronson\uff0c\u7814\u7a76\u91cf\u5b50\u590d\u6742\u6027\u3002<\/p>\n

\u5728MIT\u8bbf\u5b66\u671f\u95f4\uff0c\u4ed6\u89e3\u51b3\u4e86\u91cf\u5b50\u4fe1\u606f\u5b66\u8005John Watrous\u4e8e2002\u5e74\u63d0\u51fa\u7684\u4e00\u4e2aopen problem\u3002<\/p>\n

\u503c\u5f97\u4e00\u63d0\u7684\u662f\uff0cScott Aaronson\u6559\u6388\u540e\u6765\u4e8e2022\u5e74\u52a0\u5165OpenAI\uff0c\u4ece\u4e8bAI \u5b89\u5168\u7684\u7406\u8bba\u57fa\u7840\u7814\u7a76\u3002<\/p>\n

2017 \u5e74\uff0c\u9648\u7acb\u6770\u5728\u8ba1\u7b97\u673a\u79d1\u5b66\u57fa\u7840\u5e74\u5ea6\u7814\u8ba8\u4f1a\uff08FOCS\uff09 \u4e0a\u53d1\u8868\u8bba\u6587\uff0c\u89e3\u51b3\u4e86\u8ba1\u7b97\u590d\u6742\u6027\u9886\u57df\u7684\u91cd\u8981\u95ee\u9898\uff0c\u6210\u4e3a\u9996\u4f4d\u5728 FOCS \u4e0a\u53d1\u8868\u8bba\u6587\u7684\u4e2d\u56fd\u672c\u79d1\u751f\u3002<\/p>\n

\u540c\u5e74\uff0c\u4ed6\u4ece\u6e05\u534e\u59da\u73ed\u6bd5\u4e1a\uff0c\u8d74MIT\u653b\u8bfb\u8ba1\u7b97\u673a\u79d1\u5b66\u535a\u58eb\u5b66\u4f4d\u3002<\/p>\n

\u535a\u58eb\u671f\u95f4\uff0c\u9648\u7acb\u6770\u5e08\u4eceRyan Williams\uff0c\u7814\u7a76\u65b9\u5411\u96c6\u4e2d\u4e8e\u8ba1\u7b97\u590d\u6742\u6027\u7406\u8bba\u4e0e\u7ec6\u7c92\u5ea6\u590d\u6742\u5ea6\u7406\u8bba\u3002<\/p>\n

\u671f\u95f4\uff0c\u4ed6\u66fe\u591a\u6b21\u5728FOCS\u3001STOC\u7b49\u7406\u8bba\u8ba1\u7b97\u673a\u9876\u7ea7\u4f1a\u8bae\u53d1\u8868\u8bba\u6587\uff0c\u83b7\u5f97FOCS\u6700\u4f73\u5b66\u751f\u8bba\u6587\u5956\u7b49\u91cd\u8981\u5b66\u672f\u8363\u8a89\uff0c\u5305\u62ec\uff1a<\/p>\n

2022 \u5e74\uff0c\u9648\u7acb\u6770\u4eceMIT\u83b7\u5f97\u535a\u58eb\u5b66\u4f4d\uff0c\u968f\u540e\u52a0\u5165UC Berkeley Miller\u7814\u7a76\u6240\uff0c\u62c5\u4efbMiller Postdoctoral Fellow\uff08\u7c73\u52d2\u535a\u58eb\u540e\u7814\u7a76\u5458\uff09\u3002<\/p>\n

Miller Fellowship \u6bcf\u5e74\u4ec5\u6388\u4e88\u5c11\u6570\u6770\u51fa\u9752\u5e74\u5b66\u8005\uff0c\u4ed6\u5728\u4f2f\u514b\u5229\u671f\u95f4\u7684\u5408\u4f5c\u5bfc\u5e08\u5305\u62ecAvishay Tal\u4ee5\u53ca\u91cf\u5b50\u8ba1\u7b97\u5960\u57fa\u4ebaUmesh V. Vazirani\u3002<\/p>\n

2024\u5e74\uff0c\u9648\u7acb\u6770\u4e00\u7bc7\u540d\u4e3a\u300a\u590d\u6742\u6027\u4e0b\u754c\u7684\u9006\u5411\u6570\u5b66\u300b\u66f4\u662f\u7ed9\u56f0\u6270\u5b66\u754c\u8fd1 50 \u5e74\u7684\u4e00\u7c7b\u8ba1\u7b97\u590d\u6742\u6027\u96be\u9898\u5e26\u6765\u65b0\u601d\u8def\u3002<\/p>\n

2025\u5e74\uff0c\u4ed6\u6b63\u5f0f\u52a0\u5165UC Berkeley\uff0c\u6210\u4e3aEECS\u52a9\u7406\u6559\u6388\uff0c\u5e76\u5f00\u59cb\u4e3b\u8bb2\u7814\u7a76\u751f\u8bfe\u7a0b \u300aComputational Complexity Theory\u300b\u3002<\/p>\n

\u76ee\u524d\uff0c\u9648\u7acb\u6770\u7684\u4e3b\u8981\u7814\u7a76\u65b9\u5411\u5305\u62ecP\u4e0eNP\u3001\u7535\u8def\u590d\u6742\u6027\u3001\u7ec6\u7c92\u5ea6\u590d\u6742\u6027\u3001\u53bb\u968f\u673a\u5316\uff08Derandomization\uff09\u3001\u7b97\u6cd5\u4e0b\u754c\u7b49\u7406\u8bba\u8ba1\u7b97\u673a\u79d1\u5b66\u6838\u5fc3\u95ee\u9898\u3002<\/p>\n

\u4ed6\u5728\u53bb\u968f\u673a\u5316\u4e0e\u590d\u6742\u6027\u4e0b\u754c\u4e4b\u95f4\u7684\u8054\u7cfb\u3001\u590d\u6742\u6027\u96be\u5ea6\u653e\u5927\uff08Hardness Magnification\uff09\u7b49\u65b9\u5411\u505a\u51fa\u4e86\u7cfb\u7edf\u6027\u8d21\u732e\u3002<\/p>\n

\u6b64\u5916\uff0c\u4ed6\u4e5f\u5f00\u59cb\u5c06\u590d\u6742\u6027\u7406\u8bba\u7684\u65b9\u6cd5\u5f15\u5165\u91cf\u5b50\u7269\u7406\u4e0eAI\u5b89\u5168\u7b49\u524d\u6cbf\u9886\u57df\u3002<\/p>\n

\u73b0\u5728\uff0c\u5728OpenAI\u660e\u786e\u5f00\u542fAI4S\u7684\u63a2\u7d22\u65b9\u5411\u540e\uff0c\u9648\u7acb\u6770\u6210\u4e86OpenAI\u4e00\u5458\u3002<\/p>\n

\u4e0d\u8fc7\u9648\u7acb\u6770\u4e00\u5982\u65e2\u5f80\u4fdd\u6301\u7740\u4f4e\u8c03\uff0c\u5728\u4ed6\u4e2a\u4eba\u7684\u5404\u4e2a\u5e73\u53f0\uff0c\u4f9d\u7136\u8fd8\u662f\u6700\u65b0\u7684\u8bba\u6587\u6210\u679c\u6d88\u606f\u3002<\/p>\n

\u53c2\u8003\u8d44\u6599\uff1a[1]https:\/\/www.tsinghua.org.cn\/info\/1953\/13913.htm[2]https:\/\/chen-lijie.github.io\/documents\/CV.pdf[3]https:\/\/chen-lijie.github.io\/<\/p>\n

\n

\u7248\u6743\u58f0\u660e\uff1a<\/strong><\/p>\n

\u672c\u535a\u5ba2\u90e8\u5206\u5185\u5bb9\u4e3a\u8f6c\u8f7d\u6587\u7ae0\uff0c\u65e8\u5728\u5206\u4eab\u6709\u4ef7\u503c\u7684\u4fe1\u606f\uff0c\u7248\u6743\u5f52\u539f\u4f5c\u8005\u6240\u6709\u3002<\/p>\n

\u8f6c\u8f7d\u4ec5\u4e3a\u4e2a\u4eba\u5b66\u4e60\u4e0e\u4ea4\u6d41\u76ee\u7684\uff0c\u4e0d\u5bf9\u6587\u7ae0\u89c2\u70b9\u8d1f\u8d23\uff0c\u4ea6\u4e0d\u7528\u4e8e\u4efb\u4f55\u5546\u4e1a\u7528\u9014\u3002<\/p>\n

\u5982\u6d89\u53ca\u7248\u6743\u95ee\u9898\uff0c\u8bf7\u8054\u7cfb\u672c\u4eba\u5220\u9664\u3002<\/p>\n<\/blockquote>\n","protected":false},"excerpt":{"rendered":"

henry \u53d1\u81ea \u51f9\u975e\u5bfa<\/p>\n

\u6700\u65b0\u6d88\u606f\uff1a\u59da\u73ed\u5927\u795e\u9648\u7acb\u6770\uff0c\u52a0\u76dfOpenAI\u4e86\u3002<\/p>\n

\u636e\u201cTop\u534e\u4eba\u793e\u6d88\u606f\u201d\uff0cOpenAI\u5185\u90e8\u786e\u8ba4\uff1a\u6e05\u534e\u59da\u73ed\u5929\u624d\u3001UC\u4f2f\u514b\u5229EECS\u52a9\u7406\u6559\u6388\u9648\u7acb\u6770\u5df2\u52a0\u76dfOpenAI\uff0c\u8d1f\u8d23\u6570\u5b66\u63a8\u7406\uff01<\/p>\n

\u503c\u5f97\u4e00\u63d0\u7684\u662f\uff0cOpenAI \u5728\u53bb\u5e74 9 \u6708\u53d1\u8868\u7684\u51fa\u5708\u8bba\u6587\u300aWhy Language Models Hallucinate\u300b\u4e2d\uff0c\u4e5f\u5f15\u7528\u4e86\u9648\u7acb\u6770\u53c2\u4e0e\u7684\u53e6\u4e00\u7bc7\u7814\u7a76\u300aWhy and How LLMs <\/p>\n","protected":false},"author":1,"featured_media":917,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-921","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-xwzx"],"_links":{"self":[{"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/posts\/921","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/comments?post=921"}],"version-history":[{"count":0,"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/posts\/921\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/media\/917"}],"wp:attachment":[{"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/media?parent=921"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/categories?post=921"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/1ow.top\/index.php\/wp-json\/wp\/v2\/tags?post=921"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}