{"id":561,"date":"2025-11-15T21:51:06","date_gmt":"2025-11-15T13:51:06","guid":{"rendered":"https:\/\/www.ndnlab.com\/?p=561"},"modified":"2025-11-27T15:39:08","modified_gmt":"2025-11-27T07:39:08","slug":"neurips%e5%94%af%e4%b8%80%e6%bb%a1%e5%88%86%e8%ae%ba%e6%96%87%e6%9b%9d%e5%85%89%ef%bc%8c%e6%9d%a5%e8%87%aa%e6%b8%85%e5%8d%8e%e4%b8%8a%e4%ba%a4","status":"publish","type":"post","link":"https:\/\/www.ndnlab.com\/?p=561","title":{"rendered":"NeurIPS\u552f\u4e00\u6ee1\u5206\u8bba\u6587\u66dd\u5149\uff0c\u6765\u81ea\u6e05\u534e\u4e0a\u4ea4"},"content":{"rendered":"\n

\u56db\u4e2a\u5ba1\u7a3f\u4eba\u5168\u7ed96\u5206\uff0cNeurIPS\u552f\u4e00\u6ee1\u5206\u8bba\u6587\u70b8\u4e86\uff01<\/p>\n\n\n\n

\u4e4b\u6240\u4ee5\u8bf4\u5b83\u70b8\uff0c\u4e3b\u8981\u662f\u8bba\u6587\u7ed9\u51fa\u7684\u7ed3\u8bba\u5b9e\u5728\u592a\u51fa\u4eba\u610f\u6599\u4e86\u2014\u2014<\/p>\n\n\n\n

\u771f\u6b63\u51b3\u5b9a\u63a8\u7406\u4e0a\u9650\u7684\u662f\u57fa\u5ea7\u6a21\u578b\u672c\u8eab\u800c\u975e\u5f3a\u5316\u5b66\u4e60\uff0c\u4e14\u84b8\u998f\u6bd4\u5f3a\u5316\u5b66\u4e60\u66f4\u6709\u671b\u5b9e\u73b0\u5927\u6a21\u578b\u81ea\u6211\u8fdb\u5316\u3002<\/strong><\/p>\n\n\n\n

\u597d\u5bb6\u4f19\uff0c\u8fd9\u65e0\u5f02\u4e8e\u7ed9\u6b63\u7099\u624b\u53ef\u70ed\u7684RLVR\uff08\u53ef\u9a8c\u8bc1\u5956\u52b1\u7684\u5f3a\u5316\u5b66\u4e60\uff09\u8fce\u9762\u6cfc\u4e0b\u4e00\u76c6\u51b7\u6c34~<\/p>\n\n\n\n

RLVR\uff0c\u81ea\u5927\u6a21\u578b\u63a8\u7406\u8303\u5f0f\u5f00\u542f\u540e\u5c31\u6210\u4e3a\u4e00\u4f17\u4e3b\u6d41\u6a21\u578b\uff08\u5982OpenAI-o1\u3001DeepSeek-R1\uff09\u7684\u6838\u5fc3\u9a71\u52a8\u529b\u3002<\/p>\n\n\n\n

\u7531\u4e8e\u65e0\u9700\u4eba\u5de5\u6807\u6ce8\uff0c\u901a\u8fc7\u81ea\u52a8\u9a8c\u8bc1\u5956\u52b1\u4f18\u5316\u6a21\u578b\uff0c\u5b83\u4e00\u5ea6\u88ab\u89c6\u4e3a\u5b9e\u73b0\u6a21\u578b\u81ea\u6211\u8fdb\u5316\u3001\u903c\u8fd1\u66f4\u9ad8\u63a8\u7406\u80fd\u529b\u7684\u7ec8\u6781\u8def\u5f84\u3002<\/p>\n\n\n\n

\u4f46\u6765\u81ea\u6e05\u534e\u4e0a\u4ea4\u7684\u8fd9\u7bc7\u8bba\u6587<\/strong>\uff0c\u5374\u8ba9\u98ce\u5411\u9661\u7136\u751f\u53d8\u2014\u2014<\/p>\n\n\n\n

\u5982\u679c\u8fdb\u5316\u7684\u94a5\u5319\u4e0d\u5728\u5f3a\u5316\u5b66\u4e60\uff0c\u90a3\u5f53\u524d\u56f4\u7ed5RLVR\u7684\u5de8\u989d\u6295\u5165\u4e0e\u63a2\u7d22\uff0c\u610f\u4e49\u4f55\u5728\uff1f<\/p>\n\n\n

\n
\"NeurIPS\u552f\u4e00\u6ee1\u5206\u8bba\u6587\u66dd\u5149\uff0c\u6765\u81ea\u6e05\u534e\u4e0a\u4ea4\u63d2\u56fe\"<\/figure>\n<\/div>\n\n\n

\u771f\u6b63\u80fd\u7a81\u7834\u63a8\u7406\u4e0a\u9650\uff1a\u84b8\u998f\u800c\u975e\u5f3a\u5316\u5b66\u4e60<\/h1>\n\n\n\n

\u8fd9\u7bc7\u8bba\u6587\u9898\u76ee\u4e3a\u300aDoes Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model? \u300b\uff0c\u201c\u83b7NeurIPS\u552f\u4e00\u6ee1\u5206\u201d\u7684\u7ed3\u8bba\u7531PaperCopilot\uff08\u975e\u5b98\u65b9\u8bba\u6587\u5206\u6790\u5e73\u53f0\uff09\u7edf\u8ba1\u5f97\u51fa\u3002<\/p>\n\n\n\n

\u540c\u65f6\u5b83\u8fd8\u8363\u83b7ICML 2025 AI4Math Workshop\u6700\u4f73\u8bba\u6587\u5956\uff0c\u5e76\u5165\u9009NeurIPS 2025\u5927\u4f1a\u53e3\u5934\u62a5\u544a\u3002<\/p>\n\n\n

\n
\"NeurIPS\u552f\u4e00\u6ee1\u5206\u8bba\u6587\u66dd\u5149\uff0c\u6765\u81ea\u6e05\u534e\u4e0a\u4ea4\u63d2\u56fe1\"<\/figure>\n<\/div>\n\n\n

\u4e4b\u6240\u4ee5\u63d0\u51fa\u8fd9\u9879\u7814\u7a76\uff0c\u4e3b\u8981\u662f\u8fd1\u5e74\u6765RLVR\u5728\u5927\u8bed\u8a00\u6a21\u578b\u4e2d\u88ab\u5e7f\u6cdb\u5e94\u7528\u4e8e\u63d0\u5347\u6570\u5b66\u3001\u7f16\u7a0b\u3001\u89c6\u89c9\u63a8\u7406\u7b49\u4efb\u52a1\u7684\u8868\u73b0\u3002<\/p>\n\n\n\n

\u968f\u4e4b\u800c\u6765\u7684\u662f\uff0cAI\u5708\u666e\u904d\u5047\u8bbe\u2014\u2014<\/p>\n\n\n\n

RLVR\u4e0d\u4f46\u80fd\u63d0\u5347\u63a8\u7406\u6548\u7387\uff0c\u8fd8\u53ef\u80fd\u6269\u5c55\u6a21\u578b\u7684\u63a8\u7406\u80fd\u529b\uff0c\u5373\u8ba9\u6a21\u578b\u5b66\u4f1a\u5e95\u5c42\u57fa\u7840\u6a21\u578b\u672c\u6765\u4e0d\u4f1a\u7684\u65b0\u63a8\u7406\u8def\u5f84\u3002<\/p>\n\n\n\n

\u4f46\u95ee\u9898\u662f\uff0c\u8fd9\u4e00\u7ed3\u8bba\u771f\u7684\u6210\u7acb\u5417\uff1f<\/p>\n\n\n\n

\u4e8e\u662f\u5e26\u7740\u7591\u95ee\uff0c\u6765\u81ea\u6e05\u534e\u4e0a\u4ea4\u7684\u7814\u7a76\u56e2\u961f\u6838\u5fc3\u60f3\u8981\u5f04\u6e05\u4e00\u4e2a\u95ee\u9898\uff1a<\/p>\n\n\n\n

\n

RLVR\u662f\u5426\u771f\u7684\u8ba9\u5927\u8bed\u8a00\u6a21\u578b\u8d85\u8d8a\u5176\u201c\u5e95\u6a21\u201d\u63a8\u7406\u80fd\u529b\u8fb9\u754c\uff0c\u8fd8\u662f\u53ea\u662f\u4f18\u5316\u5df2\u6709\u80fd\u529b\uff1f<\/strong><\/p>\n<\/blockquote>\n\n\n\n

\u800c\u901a\u8fc7\u4e00\u7cfb\u5217\u5b9e\u9a8c\uff0c\u56e2\u961f\u5f97\u51fa\u4ee5\u4e0b\u6700\u65b0\u7ed3\u8bba\uff1a<\/p>\n\n\n\n

    \n
  • RLVR\u4e3b\u8981\u662f\u5728\u201c\u5f3a\u5316\u201d\u5e95\u6a21\u5df2\u6709\u7684\u8def\u5f84\uff0c\u800c\u4e0d\u662f\u201c\u53d1\u73b0\u201d\u5e95\u6a21\u6ca1\u6709\u7684\u8def\u5f84\u3002<\/li>\n\n\n\n
  • RL\u8bad\u7ec3\u540e\u7684\u6a21\u578b\u5728\u4f4e\u91c7\u6837\u6b21\u6570\uff08\u5982pass@1\uff09\u8868\u73b0\u66f4\u597d\uff0c\u4f46\u968f\u7740\u91c7\u6837\u6b21\u6570\u589e\u52a0\uff08pass@64\u3001pass@256\u2026\uff09\uff0c\u5e95\u6a21\u53cd\u800c\u80fd\u8d85\u8fc7RL\u6a21\u578b\uff0c\u8fd9\u8bf4\u660e\u5e95\u6a21\u9690\u85cf\u7684\u63a8\u7406\u80fd\u529b\u88ab\u4f4e\u4f30\u4e86\u3002<\/li>\n\n\n\n
  • \u591a\u79cdRL\u7b97\u6cd5\uff08\u5982PPO\u3001GRPO\u3001Reinforce++\u7b49\uff09\u5728\u63d0\u5347\u91c7\u6837\u6548\u7387\u65b9\u9762\u5dee\u5f02\u4e0d\u5927\uff0c\u4e14\u4e0e\u201c\u7406\u8bba\u4e0a\u5e95\u6a21\u6700\u5927\u80fd\u529b\u201d\u76f8\u6bd4\uff0c\u4ecd\u6709\u660e\u663e\u5dee\u8ddd\uff0c\u8fd9\u8bf4\u660e\u60f3\u9760RL\u7a81\u7834\u5e95\u6a21\u4e0a\u9650\u8fd8\u4e0d\u591f\u3002<\/li>\n\n\n\n
  • \u84b8\u998f\u65b9\u6cd5\u66f4\u6709\u53ef\u80fd\u201c\u6269\u5c55\u201d\u6a21\u578b\u7684\u63a8\u7406\u80fd\u529b\u8303\u56f4\uff0c\u56e0\u4e3a\u5176\u63a5\u6536\u6765\u81ea\u6559\u5e08\u6a21\u578b\u7684\u65b0\u63a8\u7406\u6a21\u5f0f\uff0c\u800cRLVR\u66f4\u53d7\u9650\u4e8e\u5e95\u6a21\u3002<\/li>\n<\/ul>\n\n\n
    \n
    \"NeurIPS\u552f\u4e00\u6ee1\u5206\u8bba\u6587\u66dd\u5149\uff0c\u6765\u81ea\u6e05\u534e\u4e0a\u4ea4\u63d2\u56fe2\"<\/figure>\n<\/div>\n\n\n

    \u6362\u53e5\u8bdd\u8bf4\uff0c\u4e0e\u666e\u904d\u8ba4\u77e5\u76f8\u53cd\uff0cRLVR\u7684\u5b9e\u9645\u4f5c\u7528\u5f88\u53ef\u80fd\u88ab\u4e25\u91cd\u9ad8\u4f30\u4e86\u3002<\/p>\n\n\n

    \n
    \"NeurIPS\u552f\u4e00\u6ee1\u5206\u8bba\u6587\u66dd\u5149\uff0c\u6765\u81ea\u6e05\u534e\u4e0a\u4ea4\u63d2\u56fe3\"<\/figure>\n<\/div>\n\n\n

    \u5173\u952e\u8bc4\u4f30\u6307\u6807\uff1apass@k<\/h1>\n\n\n\n

    \u800c\u4e3a\u4e86\u5f97\u51fa\u4e0a\u8ff0\u7ed3\u8bba\uff0c\u4ed6\u4eec\u91c7\u7528\u4e86pass@k<\/strong>\u8fd9\u4e00\u5173\u952e\u8bc4\u4f30\u6307\u6807\u3002<\/p>\n\n\n\n

    \u6240\u8c13pass@k\uff0c\u662f\u6307\u8861\u91cf\u4e00\u4e2a\u6a21\u578b\u5728\u591a\u6b21\u5c1d\u8bd5\u4e2d\uff0c\u81f3\u5c11\u6210\u529f\u4e00\u6b21\u7684\u51e0\u7387\u3002<\/p>\n\n\n\n

    \u76f8\u6bd4\u4e00\u4e9b\u4f20\u7edf\u6307\u6807\uff08\u5982greedy decoding\u51c6\u786e\u7387\uff09\u4ec5\u53cd\u6620\u5e73\u5747\u8868\u73b0\uff0c\u5b83\u901a\u8fc7\u591a\u8f6e\u91c7\u6837\u63ed\u793a\u6a21\u578b\u7684\u63a8\u7406\u8fb9\u754c\uff0c\u80fd\u66f4\u7cbe\u51c6\u5224\u65ad\u6a21\u578b\u662f\u5426\u201c\u6709\u80fd\u529b\u201d\u89e3\u51b3\u95ee\u9898\uff0c\u800c\u975e\u201c\u5927\u6982\u7387\u201d\u89e3\u51b3\u95ee\u9898\u3002<\/p>\n\n\n\n

    \u5177\u4f53\u6765\u8bf4\uff0c\u4ed6\u4eec\u4e3b\u8981\u628a\u5e95\u6a21\u3001RL\u6a21\u578b\u653e\u5728\u540c\u4e00\u6279\u9898\u76ee\u4e0a\u53cd\u590d\u6d4b\u8bd5\uff0c\u6765\u770b\u6a21\u578b\u662f\u201c\u771f\u7684\u53d8\u806a\u660e\u201d\u8fd8\u662f\u53ea\u662f\u201c\u66f4\u4f1a\u6311\u7b54\u6848\u201d\u3002<\/p>\n\n\n\n

    \u4e3a\u907f\u514d\u5b9e\u9a8c\u7ed3\u679c\u7684\u5c40\u9650\u6027\uff0c\u56e2\u961f\u9009\u53d6\u4e86\u5927\u8bed\u8a00\u6a21\u578b\u63a8\u7406\u80fd\u529b\u7684\u4e09\u5927\u5178\u578b\u5e94\u7528\u9886\u57df\uff0c\u5e76\u642d\u914d\u6743\u5a01\u57fa\u51c6\u6570\u636e\u96c6\uff0c\u786e\u4fdd\u6d4b\u8bd5\u7684\u5168\u9762\u6027\u548c\u4ee3\u8868\u6027\u3002<\/p>\n\n\n\n

      \n
    • \u6570\u5b66\u63a8\u7406\uff08GSM8K\u3001MATH500\u7b496\u4e2a\u57fa\u51c6\uff09<\/li>\n\n\n\n
    • \u4ee3\u7801\u751f\u6210\uff08LiveCodeBench\u7b493\u4e2a\u57fa\u51c6\uff09<\/li>\n\n\n\n
    • \u89c6\u89c9\u63a8\u7406\uff08MathVista\u7b492\u4e2a\u57fa\u51c6\uff09<\/li>\n<\/ul>\n\n\n\n

      \u6a21\u578b\u5219\u4ee5\u4e3b\u6d41\u5927\u8bed\u8a00\u6a21\u578b\u5bb6\u65cf\u4e3a\u57fa\u7840\uff0c\u5305\u62ecQwen2.5\u7cfb\u5217\uff0870\u4ebf\u3001140\u4ebf\u3001320\u4ebf\u53c2\u6570\uff09\u548cLLaMA-3.1-80\u4ebf\u53c2\u6570\u6a21\u578b\u7b49\uff0c\u5e76\u6784\u5efa\u201c\u57fa\u7840\u6a21\u578b vs RLVR\u8bad\u7ec3\u6a21\u578b\u201d\u7684\u5bf9\u7167\u7ec4\u5408\u3002<\/p>\n\n\n\n

      \u5176\u4e2dRLVR\u8bad\u7ec3\u6a21\u578b\u662f\u6307\uff0c\u5206\u522b\u7528PPO\u3001GRPO\u3001Reinforce++\u7b496\u79cd\u4e3b\u6d41RLVR\u7b97\u6cd5\u8bad\u7ec3\u540e\u7684\u7248\u672c\uff0c\u5f62\u6210\u591a\u7ec4\u5e73\u884c\u5bf9\u7167\u3002\u8fd9\u6837\u65e2\u80fd\u5bf9\u6bd4RLVR\u4e0e\u57fa\u7840\u6a21\u578b\u7684\u5dee\u5f02\uff0c\u4e5f\u80fd\u6a2a\u5411\u6bd4\u8f83\u4e0d\u540cRLVR\u7b97\u6cd5\u7684\u6548\u679c\u3002<\/p>\n\n\n

      \n
      \"NeurIPS\u552f\u4e00\u6ee1\u5206\u8bba\u6587\u66dd\u5149\uff0c\u6765\u81ea\u6e05\u534e\u4e0a\u4ea4\u63d2\u56fe4\"<\/figure>\n<\/div>\n\n\n

      \u7136\u540e\u5c31\u662f\u5bf9\u4e0d\u540c\u6a21\u578b\u5728\u5404\u57fa\u51c6\u4efb\u52a1\u4e0a\u7684pass@k\u6307\u6807<\/strong>\u8fdb\u884c\u591a\u7ef4\u5ea6\u91c7\u96c6\u4e0e\u5206\u6790\u3002<\/p>\n\n\n\n

      \u9488\u5bf9\u6bcf\u4e2a\u6d4b\u8bd5\u6837\u672c\uff0c\u5206\u522b\u8ba9\u57fa\u7840\u6a21\u578b\u548cRLVR\u6a21\u578b\u8fdb\u884c\u4e0d\u540c\u6b21\u6570\u7684\u91c7\u6837\uff08k\u503c\u4ece1\u9010\u6b65\u63d0\u5347\u81f31024\uff09\uff0c\u8bb0\u5f55\u6bcf\u6b21\u91c7\u6837\u4e2d\u201c\u81f3\u5c11\u51fa\u73b0\u4e00\u4e2a\u6b63\u786e\u7ed3\u679c\u201d\u7684\u6982\u7387\u3002<\/p>\n\n\n\n

      \u968f\u540e\u56e2\u961f\u91cd\u70b9\u5206\u6790\u4e24\u4e2a\u5173\u952e\u89c4\u5f8b\uff1a<\/p>\n\n\n\n

      \u4e00\u662f\u540c\u4e00k\u503c\u4e0b\uff0cRLVR\u6a21\u578b\u4e0e\u57fa\u7840\u6a21\u578b\u7684pass@k\u5dee\u5f02\uff1b\u4e8c\u662f\u968f\u7740k\u503c\u589e\u5927\uff0c\u4e24\u7c7b\u6a21\u578bpass@k\u66f2\u7ebf\u7684\u53d8\u5316\u8d8b\u52bf\u3002<\/strong><\/p>\n\n\n\n

      \u540c\u65f6\uff0c\u7ed3\u5408\u6a21\u578b\u8f93\u51fa\u7684\u63a8\u7406\u8def\u5f84\u56f0\u60d1\u5ea6\u5206\u6790\uff08perplexity\uff09\u3001\u53ef\u89e3\u95ee\u9898\u5b50\u96c6\u6bd4\u5bf9\u7b49\u8f85\u52a9\u624b\u6bb5\uff0c\u6700\u7ec8\u5f62\u6210\u5bf9RLVR\u80fd\u529b\u7684\u5168\u9762\u5224\u65ad\u3002<\/p>\n\n\n\n

      \u8bba\u6587\u4f5c\u8005\u4ecb\u7ecd<\/h1>\n\n\n\n

      \u503c\u5f97\u4e00\u63d0\u7684\u662f\uff0c\u8fd9\u9879\u7814\u7a76\u8fd8\u662f\u51fa\u81ea\u54b1\u4eec\u56fd\u5185\u7814\u7a76\u4eba\u5458\u4e4b\u624b\u3002<\/p>\n\n\n\n

      \u4e00\u51718\u4f4d\uff0c7\u4f4d\u6765\u81ea\u6e05\u534e\u5927\u5b66LeapLab\uff0c1\u4f4d\u6765\u81ea\u4e0a\u6d77\u4ea4\u901a\u5927\u5b66\u3002<\/p>\n\n\n\n

      \u9879\u76ee\u8d1f\u8d23\u4ebaYang Yue (\u4e50\u6d0b)<\/strong>\uff0c\u6e05\u534e\u5927\u5b66\u81ea\u52a8\u5316\u7cfb\u56db\u5e74\u7ea7\u535a\u58eb\u751f\u3002<\/p>\n\n\n\n

      \u7814\u7a76\u65b9\u5411\u4e3a\u5f3a\u5316\u5b66\u4e60\u3001\u4e16\u754c\u6a21\u578b\u3001\u591a\u6a21\u6001\u5927\u6a21\u578b\u548c\u5177\u8eab\u667a\u80fd\uff0c\u4e4b\u524d\u66fe\u5728\u989c\u6c34\u6210\u521b\u529e\u7684\u65b0\u52a0\u5761Sea AI Lab\u548c\u5b57\u8282\u8df3\u52a8 Seed\u56e2\u961f\u5b9e\u4e60\u8fc7\u3002<\/p>\n\n\n\n

      \u867d\u7136\u8fd8\u662f\u5b66\u751f\uff0c\u4f46\u53d1\u8868\u6216\u53c2\u4e0e\u53d1\u8868\u7684\u591a\u7bc7\u8bba\u6587\u5747\u5165\u9009\u9876\u4f1a\u3002\u8fd9\u5f53\u4e2d\uff0c\u4ed6\u4ee5\u6838\u5fc3\u4f5c\u8005\u8eab\u4efd\u53d1\u8868\u7684\u8bba\u6587\u300aHow Far is Video Generation from World Model: A Physical Law Perspective\u300b\uff0c\u56e0\u63a2\u7d22\u89c6\u9891\u6a21\u578b\u80fd\u5426\u5b66\u4f1a\u7269\u7406\u89c4\u5f8b\uff0c\u8fd8\u88ab\u56fd\u5185\u5916\u4f17\u591a\u5927\u4f6cYan Lecun\uff0cxie saining\uff0cKevin Murphy\u7b49\u8f6c\u53d1\u3002<\/p>\n\n\n

      \n
      \"NeurIPS\u552f\u4e00\u6ee1\u5206\u8bba\u6587\u66dd\u5149\uff0c\u6765\u81ea\u6e05\u534e\u4e0a\u4ea4\u63d2\u56fe5\"<\/figure>\n<\/div>\n\n\n

      \u53e6\u4e00\u4f4d\u548c\u4ed6\u8d21\u732e\u76f8\u540c\u7684\u4f5c\u8005Zhiqi Chen<\/strong>\uff0c\u76ee\u524d\u4e3a\u6e05\u534e\u5927\u5b66\u81ea\u52a8\u5316\u5de5\u7a0b\u7cfb\u5927\u4e09\u5b66\u751f\u3002<\/p>\n\n\n\n

      \u7814\u7a76\u65b9\u5411\u4e3a\u63a8\u7406\u5bc6\u96c6\u578b\u5927\u8bed\u8a00\u6a21\u578b\u7684\u5f3a\u5316\u5b66\u4e60\uff0c\u5728\u6821\u671f\u95f4\u591a\u6b21\u83b7\u5f97\u56fd\u5bb6\u5956\u5b66\u91d1\u3002<\/p>\n\n\n

      \n
      \"NeurIPS\u552f\u4e00\u6ee1\u5206\u8bba\u6587\u66dd\u5149\uff0c\u6765\u81ea\u6e05\u534e\u4e0a\u4ea4\u63d2\u56fe6\"<\/figure>\n<\/div>\n\n\n

      \u901a\u8baf\u4f5c\u8005Gao Huang(\u9ec4\u9ad8)<\/strong>\uff0c\u6e05\u534e\u5927\u5b66\u81ea\u52a8\u5316\u7cfb\u526f\u6559\u6388\u3001\u535a\u58eb\u751f\u5bfc\u5e08\uff0c LeapLab\u8d1f\u8d23\u4eba\u3002<\/p>\n\n\n\n

      \u4ed6\u6700\u77e5\u540d\u7684\u5de5\u4f5c\u4e4b\u4e00\u5c31\u662f\u53d1\u8868\u4e86\u8bba\u6587\u300aDensely Connected Convolutional Networks\u300b\uff0c\u5176\u4e2d\u63d0\u51fa\u4e86\u7ecf\u5178\u5377\u79ef\u67b6\u6784\u6a21\u578bDenseNet\u3002<\/p>\n\n\n\n

      \u8be5\u8bba\u6587\u4e0d\u4ec5\u8363\u83b7CVPR2017\u6700\u4f73\u8bba\u6587\uff0c\u800c\u4e14\u88ab\u7f16\u5165\u591a\u672c\u6df1\u5ea6\u5b66\u4e60\u8457\u4f5c\uff0c\u5355\u7bc7\u5f15\u7528\u91cf\u63a5\u8fd16\u4e07\u6b21\u3002<\/p>\n\n\n

      \n
      \"NeurIPS\u552f\u4e00\u6ee1\u5206\u8bba\u6587\u66dd\u5149\uff0c\u6765\u81ea\u6e05\u534e\u4e0a\u4ea4\u63d2\u56fe7\"<\/figure>\n<\/div>\n\n\n

      \u5176\u4ed6\u4f5c\u8005\u4e2d\uff0c\u6765\u81ea\u6e05\u534e\u7684\u8fd8\u6709\uff1a<\/p>\n\n\n\n