{"id":59139,"date":"2025-05-10T18:44:01","date_gmt":"2025-05-10T10:44:01","guid":{"rendered":"https:\/\/swarma.org\/?p=59139"},"modified":"2025-05-10T18:44:01","modified_gmt":"2025-05-10T10:44:01","slug":"%e3%80%8aaixscience%e5%8d%81%e5%a4%a7%e5%89%8d%e6%b2%bf%e8%a7%82%e5%af%9f%e3%80%8b9%ef%bc%9a%e5%90%88%e6%88%90%e6%95%b0%e6%8d%ae%e5%92%8c%e6%95%b0%e6%8d%ae%e5%9f%ba%e7%a1%80%e8%ae%be%e6%96%bd","status":"publish","type":"post","link":"https:\/\/swarma.org\/?p=59139","title":{"rendered":"\u300aAI\u00d7SCIENCE\u5341\u5927\u524d\u6cbf\u89c2\u5bdf\u300b9\uff1a\u5408\u6210\u6570\u636e\u548c\u6570\u636e\u57fa\u7840\u8bbe\u65bd"},"content":{"rendered":"
\n
\n
\n
\n
\n
\n
\n
\"\u56fe\u7247\"<\/span><\/section>\n
\n
\n
\n
\n
\n

<\/span><\/section>\n<\/section>\n
\n
\n

\u5bfc\u8bed<\/span><\/strong><\/p>\n<\/section>\n<\/section>\n

\n

<\/span><\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\n
\n

\u4e0a\u6d77\u79d1\u5b66\u667a\u80fd\u7814\u7a76\u9662\u3001\u96c6\u667a\u79d1\u5b66\u7814\u7a76\u4e2d\u5fc3\u548c\u963f\u91cc\u4e91\u8054\u5408\u53d1\u5e03\u4e86\u300aAI \u00d7 Science\u5341\u5927\u524d\u6cbf\u89c2\u5bdf\u300b\uff0c\u68b3\u7406\u51fa35\u4e2a\u7814\u7a76\u524d\u6cbf\uff0c\u6765\u63a8\u52a8\u79d1\u5b66\u53d1\u5c55\u7684\u9ec4\u91d1\u65f6\u4ee3\u5230\u6765\u3002\u672c\u7bc7\u4e3a\u524d\u6cbf\u89c2\u5bdf9\uff0c\u626b\u63cf\u4e0b\u65b9\u4e8c\u7ef4\u7801\uff0c\u53ef\u83b7\u5f97\u5b8c\u6574\u7248\u4e0b\u8f7d\u3001\u5feb\u901f\u94fe\u63a5\u8bba\u6587\u539f\u6587\u3002<\/span><\/span><\/strong><\/p>\n


<\/span><\/strong><\/section>\n
<\/span><\/strong><\/section>\n
\u626b\u63cf\u4e8c\u7ef4\u7801 \u4e0b\u8f7d\u5b8c\u6574\u62a5\u544a<\/span><\/strong><\/section>\n<\/section>\n
<\/strong><\/strong><\/section>\n<\/section>\n
\n
<\/strong><\/strong><\/section>\n<\/section>\n
\n
<\/strong><\/strong><\/section>\n
<\/strong><\/strong><\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\n
\n
<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n
\n
\"\u56fe\u7247\"<\/section>\n<\/section>\n<\/section>\n<\/section>\n


<\/span><\/p>\n<\/section>\n

\n
\n
\n
\n
\n

<\/span><\/section>\n

<\/span><\/section>\n
\n
<\/section>\n<\/section>\n<\/section>\n
\n

\u5408\u6210\u6570\u636e\u548c\u6570\u636e\u57fa\u7840\u8bbe\u65bd<\/span><\/strong><\/span><\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/span><\/section>\n

<\/span><\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


<\/span><\/p>\n

\u80cc\u666f\u4ecb\u7ecd<\/span><\/span><\/p>\n

\u5728AI\u5927\u6a21\u578b\u65f6\u4ee3\uff0c\u6570\u636e\u4f5c\u4e3a\u5173\u952e\u6218\u7565\u8d44\u6e90\uff0c\u65e2\u662f\u6a21\u578b\u8bad\u7ec3\u7684\u57fa\u7840\u8981\u7d20\uff0c\u4e5f\u662f\u5236\u7ea6\u6027\u80fd\u63d0\u5347\u7684\u6838\u5fc3\u74f6\u9888\uff0c\u5176\u8d28\u91cf\u4e0e\u89c4\u6a21\u76f4\u63a5\u51b3\u5b9a\u667a\u80fd\u7cfb\u7edf\u7684\u80fd\u529b\u8fb9\u754c\u3002<\/span><\/span><\/span>\u79d1<\/span><\/span>\u5b66\u667a\u80fd\u7684\u53d1\u5c55\u540c\u6837\u9762\u4e34\u7740\u8bf8\u591a\u6570\u636e\u6311\u6218\uff0c\u5176\u4e2d\u6700\u4e3a\u5173\u952e\u7684\u662f\u6570\u636e\u7a00\u7f3a\u95ee\u9898[1]\u3002\u4e00\u65b9\u9762\uff0c\u5728\u67d0\u4e9b\u79d1\u5b66\u9886\u57df\uff0c\u5f80\u5f80\u96be\u4ee5\u83b7\u53d6\u5927\u89c4\u6a21\u3001\u9ad8\u8d28\u91cf\u7684\u6807\u6ce8\u6570\u636e[2]\uff1b\u53e6\u4e00\u65b9\u9762\uff0c\u67d0\u4e9b\u6570\u636e\u8d44\u6e90\u4e30\u5bcc\u7684\u9886\u57df<\/span>\uff08\u5982\u533b\u7597\uff09<\/span>\u5f80\u5f80\u6d89\u53ca\u9690\u79c1\u6216\u6cd5\u5f8b\u9650\u5236[3]\u3002\u6b64\u5916\uff0c\u8de8\u5b66\u79d1\u591a\u6e90\u6570\u636e\u6574\u5408\u4e5f\u9762\u4e34\u5de8\u5927\u969c\u788d\uff0c\u6570\u636e\u683c\u5f0f\u3001\u6807\u51c6\u548c\u7ed3\u6784\u7684\u5dee\u5f02\uff0c\u5bfc\u81f4AI\u6a21\u578b\u96be\u4ee5\u8de8\u8d8a\u591a\u4e2a\u9886\u57df\u8fdb\u884c\u901a\u7528\u5b66\u4e60\uff0c\u8fdb\u4e00\u6b65\u52a0\u5267\u4e86\u6570\u636e\u7a00\u7f3a\u95ee\u9898\u3002<\/span><\/span><\/p>\n

\u5408\u6210\u6570\u636e<\/span><\/span><\/strong>\uff08Synthetic Data\uff09<\/span>\u4e0e<\/span><\/span>\u6570\u636e\u57fa\u7840\u8bbe\u65bd<\/span><\/span><\/strong>\u5efa\u8bbe\u662f\u89e3\u51b3\u4ee5\u4e0a\u6311\u6218\u7684\u4e24\u4e2a\u91cd\u8981\u624b\u6bb5\u3002\u5728\u5408\u6210\u6570\u636e\u4e2d\uff0c\u6700\u57fa\u7840\u4e5f\u6700\u5173\u952e\u7684\u662f<\/span><\/span>\u6570\u636e\u751f\u6210<\/span><\/span><\/strong>\uff08Data Generation\uff09<\/span>\u95ee\u9898\u3002\u4e0e\u4f20\u7edf\u63d0\u793a\u5de5\u7a0b\u4e0d\u540c[13][14]\uff0c\u5373\u901a\u8fc7\u63d0\u793a\u5bf9\u90e8\u7f72\u7684LLMs\u8f93\u5165\u6587\u672c\u6570\u636e X \u8fdb\u884c\u9884\u6d4b\u4ee5\u751f\u6210\u6807\u7b7e Y\uff0c\u5408\u6210\u6570\u636e\u751f\u6210\u5219\u8981\u6c42LLMs\u6839\u636e\u6761\u4ef6\u5316\u7684\u6807\u7b7e Y \u63d0\u793a\u751f\u6210\u6587\u672c\u6570\u636e X\uff0c\u4ee5\u6b64\u9002\u914d\u591a\u6837\u5316\u7684\u4e0b\u6e38\u4efb\u52a1\u7684\u6570\u636e\u9700\u6c42\u3002<\/span><\/span><\/p>\n

<\/section>\n

LLMs\u5728\u751f\u6210\u7279\u5b9a\u6807\u7b7e\u7684\u5408\u6210\u6570\u636e\uff08a\uff09\u4e0e\u63d0\u793a\u8bcd\u9884\u6d4b\u751f\u6210\u6807\u7b7e\uff08b\uff09\u4e4b\u95f4\u7684\u6bd4\u8f83 [13]\uff5c\u56fe\u7247\u6765\u6e90\uff1aXu Guo&Yiqiang Chen.\uff082024\uff09 “Generative AI for Synthetic Data Generation: Methods\uff0cChallenges and the Future”<\/span><\/span><\/em><\/span><\/p>\n

\u7136\u800c\uff0c\u82e5\u76f4\u63a5\u5c06\u6709\u9650\u6807\u7b7e\u548c\u4efb\u52a1\u4fe1\u606f\u5d4c\u5165\u5230\u63d0\u793a\u4e2d\uff0cLLMs\u751f\u6210\u7684\u6570\u636e\u53ef\u80fd\u4e0e\u4efb\u52a1\u65e0\u5173\uff0c\u7f3a\u4e4f\u591a\u6837\u6027\uff0c\u89c4\u6a21\u4e5f\u5f88\u6709\u9650\u3002\u56e0\u6b64\u9700\u8981\u66f4\u5148<\/span><\/span>\u8fdb\u7684<\/span><\/span>\u63d0\u793a\u6280\u672f<\/span><\/span><\/span>\uff1a<\/span><\/span><\/span>\u5982<\/span><\/span><\/span>\u5c5e\u6027\u63a7\u5236\u63d0\u793a<\/span><\/span><\/strong>\uff08Attribute-controlled prompt\uff09<\/span>\uff0c\u901a\u8fc7\u6307\u5b9a\u4e00\u7ec4\u5c5e\u6027\u3001\u5e76\u5728\u63d0\u793a\u6a21\u677f\u4e2d\u8fdb\u884c\u5c5e\u6027\u6df7\u5408\uff0c\u4ee5\u6b64\u5b9a\u4e49\u4efb\u52a1\u83b7\u53d6\u5408\u6210\u6570\u636e\u7684\u6df7\u5408\u8f6f\u63d0\u793a\u65b9\u6cd5 [15]\uff0c\u6216\u4eceLLMs\u4e2d\u76f4\u63a5\u63d0\u53d6\u7279\u5b9a\u5c5e\u6027\u7684\u63d0\u793a\uff0c\u5e76\u67e5\u8be2\u751f\u6210\u7279\u5b9a\u6570\u636e\u7684<\/span><\/span>AttrPrompt [16]<\/span><\/span><\/span>\uff1b\u5982<\/span><\/span><\/span>\u8bcd\u6c47\u5316\u6280\u672f<\/span><\/span><\/strong><\/span>\uff08<\/span><\/span>Verbalizer\uff09<\/span>\uff0c\u901a\u8fc7\u5c06\u4f5c\u4e3a\u6761\u4ef6\u7684\u5c5e\u6027\u7c7b\u6269\u5c55\u4e3a\u4e00\u7ec4\u8bed\u4e49\u76f8\u4f3c\u7684\u63d0\u793a\u6765\u4fc3\u8fdb\u591a\u6837\u5316\u6570\u636e\u751f\u6210\uff0c\u4f8b\u5982 MetaPrompt [17]\u9996\u5148\u4eceLLMs\u83b7\u53d6\u6269\u5c55\u63d0\u793a\uff0c\u7136\u540e\u5229\u7528\u4e30\u5bcc\u7684\u63d0\u793a\u8fdb\u4e00\u6b65\u63d0\u793aLLMs\u751f\u6210\u6570\u636e\u3002<\/span><\/span><\/p>\n

<\/section>\n

LLMs\u751f\u6210\u7279\u5b9a\u5c5e\u6027\u4efb\u52a1\u8bad\u7ec3\u6570\u636e\u7684\u65b9\u6cd5\u3002\u540c\u4e0a[13]\u3002<\/span><\/span><\/em><\/span><\/p>\n

\u5f53\u7136\uff0c\u4ee5\u4e0a\u5408\u6210\u6570\u636e\u751f\u6210\u65b9\u6cd5\uff0c\u4e3b\u8981\u76f4\u63a5\u9488\u5bf9\u4e8e\u6587\u672c\u6570\u636e\u7b49\u7ed3\u6784\u5316\u7684\u3001\u5e8f\u5217\u5316\u7684\u6570\u636e\uff0c\u5bf9\u4e8e\u8981\u4fdd\u6301\u5168\u5c40\u4e00\u81f4\u6027\u7684\u590d\u6742\u6570\u636e<\/span>\uff08\u5982\u5408\u6210\u533b\u5b66\u5f71\u50cf\u548c3D\u56fe\u5f62\uff09<\/span>\uff0c\u8fd9\u4e9b\u6280\u672f\u9700\u8981\u8fdb\u884c\u8fc1\u79fb\u548c\u6269\u5c55\u3002<\/span><\/span><\/p>\n

\u9664\u4e86\u6570\u636e\u751f\u6210\uff0c\u5408\u6210\u6570\u636e\u53e6\u5916\u4e24\u4e2a\u91cd\u8981\u5b50\u9886\u57df\u662f<\/span><\/span>\u79d1\u5b66\u6570\u636e\u8868\u793a<\/span><\/span><\/strong>\uff08Scientific Data Representation\uff09<\/span>\u548c<\/span><\/span>\u6a21\u578b\u81ea\u6211\u6539\u8fdb\u65b9\u5411\uff08Model Self-Improvement\uff09<\/span><\/span><\/strong>\u7684\u63a2\u7d22\u3002\u524d\u8005\u91cd\u70b9\u7814\u7a76\u5982\u4f55\u6709\u6548\u5730\u7f16\u7801\u548c\u7ec4\u7ec7\u5408\u6210\u6570\u636e\uff0c\u540e\u8005\u63a2\u7d22\u5982\u4f55\u8ba9\u6a21\u578b\u901a\u8fc7\u5408\u6210\u6570\u636e\u6765\u63d0\u5347\u81ea\u8eab\u6027\u80fd<\/span><\/span>\u3002<\/span><\/span>\u9ad8\u6548\u7684\u6570\u636e\u8868\u5f81\u65b9\u6cd5\u901a\u8fc7\u63d0\u53d6\u5173\u952e\u7279\u5f81\u63d0\u5347\u751f\u6210\u6570\u636e\u7684\u4fdd\u771f\u5ea6\uff0c\u800c\u6a21\u578b\u7684\u81ea\u8fed\u4ee3\u4f18\u5316\u673a\u5236\u5219\u80fd\u6301\u7eed\u6269\u5c55\u6570\u636e\u8986\u76d6\u8303\u56f4\uff0c\u4e8c\u8005\u7684\u534f\u540c\u4f5c\u7528\u5b9e\u73b0\u4e86\u9ad8\u8d28\u91cf\u4e0e\u89c4\u6a21\u5316\u6570\u636e\u751f\u6210\u7684\u540c\u6b65\u7a81\u7834\u3002<\/span><\/span><\/span><\/p>\n

\u4e0b\u6587\u4e2d\u5c06\u91cd\u70b9\u4ecb\u7ecd\u5173\u4e8e\u5408\u6210\u6570\u636e\u7684\u6570\u636e\u751f\u6210\u3001\u79d1\u5b66\u6570\u636e\u8868\u793a\u3001\u6a21\u578b\u81ea\u6211\u6539\u8fdb\u4e09\u4e2a\u9886\u57df\u7684\u6700\u65b0\u8fdb<\/span><\/span>\u5c55\u548c<\/span><\/span>\u4ee3\u8868\u6027\u5de5\u4f5c<\/span><\/span><\/span>\uff0c<\/span><\/span>\u4ee5\u53ca\u79d1\u5b66\u6570\u636e\u57fa\u7840\u8bbe\u65bd\u5efa\u8bbe\u7684\u03c0-HuB\u9879\u76ee\u3002<\/span><\/span><\/p>\n


<\/span><\/h2>\n

\n
\n
\n
\n
\n

<\/span><\/section>\n

<\/span><\/section>\n
\n
<\/section>\n<\/section>\n<\/section>\n
\n

\u7814\u7a76\u8fdb\u5c55<\/span><\/strong><\/span><\/strong><\/strong><\/p>\n<\/section>\n

\n

<\/span><\/section>\n

<\/span><\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


<\/span><\/p>\n

\n

\u8fdb\u5c55\u76ee\u5f55<\/span><\/span><\/p>\n

    \n
  1. \n

    \u5408\u6210\u6570\u636e\u751f\u6210\uff1a\u4ece\u5e8f\u5217\u6570\u636e\u5230\u590d\u6742\u6570\u636e<\/span><\/span><\/p>\n<\/li>\n

  2. \n

    \u79d1\u5b66\u6570\u636e\u8868\u793a\uff1a\u4ece\u56fe\u795e\u7ecf\u7f51\u7edc\u5230\u591a\u6a21\u6001\u6570\u636e\u8868\u5f81<\/span><\/span><\/p>\n<\/li>\n

  3. \n

    \u6a21\u578b\u81ea\u6211\u6539\u8fdb\uff1a\u901a\u8fc7\u5408\u6210\u6570\u636e\u63d0\u5347AI\u7cfb\u7edf\u6027\u80fd<\/span><\/span><\/p>\n<\/li>\n

  4. \n

    \u6570\u636e\u57fa\u7840\u8bbe\u65bd\uff1a\u4eba\u4f53\u86cb\u767d\u8d28\u7ec4\u8ba1\u5212<\/span><\/span><\/p>\n<\/li>\n<\/ol>\n<\/blockquote>\n

    \u5408\u6210\u6570\u636e\u751f\u6210\uff1a\u4ece\u5e8f\u5217\u6570\u636e\u5230\u590d\u6742\u6570\u636e<\/span><\/span><\/h3>\n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong>\u5408\u6210\u6570\u636e\u7684\u6838\u5fc3\u4f18\u52bf\u4e0d\u4ec5\u5728\u4e8e\u53ef\u5927\u89c4\u6a21\u751f\u6210\u6570\u636e\uff0c\u800c\u4e14\u53ef\u4ee5\u6839\u636e\u7279\u5b9a\u9700\u6c42\u8fdb\u884c\u5b9a\u5236\u3002\u901a\u8fc7\u5f15\u5165\u53ef\u63a7\u7684\u53d8\u5316\u786e\u4fdd\u4e0d\u540c\u7c7b\u522b\u6570\u636e\u7684\u5e73\u8861\u8868\u793a\uff0c\u589e\u5f3a\u6a21\u578b\u7684\u9c81\u68d2\u6027\u3002\u5728\u8fd9\u4e2a\u9886\u57df\u4e2d\uff0c\u4ece\u7b80\u5355\u7684\u5e8f\u5217\u6570\u636e\u5230\u590d\u6742\u7684\u591a\u6a21\u6001\u6570\u636e\uff0c\u90fd\u51fa\u73b0\u4e86\u4e00\u4e9b\u7a81\u7834\u6027\u7684\u7814\u7a76\u6210\u679c\u3002<\/span><\/span><\/p>\n

    \u5728\u4f17\u591a\u521b\u65b0\u5de5\u4f5c\u4e2d\uff0c\u82f1\u4f1f\u8fbe\u4e8e2023\u5e74\u5f00\u53d1\u7684 MimicGen\u7cfb\u7edf\u5c55\u73b0\u51fa\u4e86\u7279\u522b\u7684\u4ef7\u503c[4]\u3002\u8fd9\u4e2a\u7cfb\u7edf\u5c55\u793a\u4e86\u5982\u4f55\u4ece\u6709\u9650\u7684\u4eba\u7c7b\u793a\u8303\u4e2d\u6784\u5efa\u5927\u89c4\u6a21\u8bad\u7ec3\u6570\u636e\u96c6\u7684\u6709\u6548\u65b9\u6cd5\u3002\u901a\u8fc7\u4ec5\u4f7f\u7528\u7ea6200\u4e2a\u4eba\u7c7b\u6f14\u793a\u6837\u672c\uff0cMimicGen\u6210\u529f\u751f\u6210\u4e86\u5305\u542b\u8d85\u8fc75\u4e07\u4e2a\u793a\u8303\u7684\u5927\u89c4\u6a21\u6570\u636e\u96c6\u3002\u8fd9\u4e9b\u6570\u636e\u6db5\u76d6\u4e8618\u79cd\u4e0d\u540c\u4efb\u52a1\uff0c\u5e76\u5728\u591a\u79cd\u573a\u666f\u914d\u7f6e\u3001\u5bf9\u8c61\u5b9e\u4f8b\u548c\u673a\u5668\u4eba\u624b\u81c2\u64cd\u4f5c\u4e2d\u5c55\u73b0\u51fa\u826f\u597d\u7684\u591a\u6837\u6027\u3002<\/span><\/span><\/p>\n

    <\/section>\n

    MimicGen\u5408\u6210\u6570\u636e\u751f\u6210\u793a\u610f\u56fe\uff5c\u56fe\u7247\u6765\u6e90\uff1aMandlekar et al.(2023) “Mimicgen: A data generation system for scalable robot learning using human demonstrations.”<\/span><\/span><\/em><\/span><\/p>\n

    MimicGen\u7684\u5de5\u4f5c\u6d41\u7a0b\u4e2d\uff0c\u7cfb\u7edf\u9996\u5148\u5bf9\u6e90\u6570\u636e\u96c6\u4e2d\u7684\u6f14\u793a\u8fdb\u884c\u7cbe\u7ec6\u5316\u89e3\u6790\uff0c\u5c06\u6bcf\u4e2a\u6f14\u793a\u5206\u89e3\u4e3a\u591a\u4e2a\u4ee5\u5bf9\u8c61\u4e3a\u4e2d\u5fc3\u7684\u5b50\u4efb\u52a1\u5355\u5143<\/span>\uff08\u56fe2\u5de6\uff09<\/span>\u3002\u5728\u751f\u6210\u65b0\u6570\u636e\u65f6\uff0c\u7cfb\u7edf\u4f1a\u9009\u62e9\u5408\u9002\u7684\u53c2\u8003\u6bb5\uff0c\u5e76\u5c06\u5176\u667a\u80fd\u5730\u8f6c\u6362\u5230\u65b0\u573a\u666f\u4e2d\uff0c\u901a\u8fc7\u8c03\u6574\u5bf9\u8c61\u4f4d\u59ff\u6765\u9002\u5e94\u4e0d\u540c\u73af\u5883\u3002\u6700\u7ec8\uff0c\u7cfb\u7edf\u4f7f\u7528\u672b\u7aef\u6267\u884c\u5668\u63a7\u5236\u5668\u6765\u5b9e\u73b0\u8fd9\u4e9b\u8f6c\u6362\u540e\u7684\u76ee\u6807\u4f4d\u59ff\u5e8f\u5217<\/span>\uff08\u56fe2\u53f3\uff09<\/span>\uff0c\u786e\u4fdd\u751f\u6210\u6570\u636e\u7684\u5b9e\u7528\u6027\u548c\u53ef\u9760\u6027\u3002<\/span><\/span><\/p>\n

    <\/section>\n

    MimicGen \u7cfb\u7edf\u6d41\u7a0b\uff5c\u56fe\u7247\u6765\u6e90\uff1aMandlekar et al.(2023) “Mimicgen: A data generation system for scalable robot learning using human demonstrations.”<\/span><\/span><\/em><\/span><\/p>\n

    \u5728\u5b9e\u9645\u5e94\u7528\u4e2d\uff0c\u8fd9\u4e9b\u6280\u672f\u53ef\u4ee5\u8fdb\u884c\u8de8\u6a21\u6001\u6574\u5408\uff0c\u5c06\u6587\u672c\u9886\u57df\u7684\u63a7\u5236\u673a\u5236\u4e0e\u89c6\u89c9\u751f\u6210\u76f8\u7ed3\u5408\uff0c\u901a\u8fc7\u5c5e\u6027\u63a7\u5236\u63d0\u793a\u5b9a\u4e49\u89c6\u89c9\u76ee\u6807\uff0c\u518d\u5229\u7528\u8bcd\u6c47\u5316\u6280\u672f\u6269\u5c55\u89c6\u89c9\u8868\u8fbe\uff0c\u6700\u7ec8\u7531\u89c6\u89c9\u751f\u6210\u6a21\u578b\u5b8c\u6210\u5177\u4f53\u751f\u6210\u4efb\u52a1\u3002\u8fd9\u79cd\u6280\u672f\u8fc1\u79fb\u4e0d\u4ec5\u6269\u5c55\u4e86\u539f\u6709\u6280\u672f\u7684\u5e94\u7528\u8303\u56f4\uff0c\u4e5f\u4e3a\u590d\u6742\u6570\u636e\u7684\u751f\u6210\u63d0\u4f9b\u4e86\u66f4\u7cbe\u786e\u7684\u63a7\u5236\u673a\u5236\u3002<\/span><\/span><\/p>\n

    \u8fd9\u79cd\u65b9\u6cd5\u4e0d\u4ec5\u663e\u8457\u63d0\u9ad8\u4e86\u6570\u636e\u751f\u6210\u7684\u6548\u7387\uff0c\u66f4\u91cd\u8981\u7684\u662f\u4fdd\u8bc1\u4e86\u751f\u6210\u6570\u636e\u7684\u8d28\u91cf\u548c\u591a\u6837\u6027\u3002\u901a\u8fc7\u8fd9\u79cd\u65b9\u5f0f\uff0cMimicGen\u4e3a\u89e3\u51b3\u4eba\u5de5\u667a\u80fd\u9886\u57df\u4e2d\u7684\u6570\u636e\u74f6\u9888\u95ee\u9898\u63d0\u4f9b\u4e86\u4e00\u4e2a\u53ef\u884c\u7684\u89e3\u51b3\u65b9\u6848\uff0c\u540c\u65f6\u4e5f\u4e3a\u672a\u6765\u5408\u6210\u6570\u636e\u751f\u6210\u6280\u672f\u7684\u53d1\u5c55\u6307\u660e\u4e86\u65b9\u5411\u3002<\/span><\/span><\/p>\n

    \u79d1\u5b66\u6570\u636e\u8868\u793a\uff1a\u4ece\u56fe\u795e\u7ecf\u7f51\u7edc\u5230\u591a\u6a21\u6001\u6570\u636e\u8868\u5f81<\/span><\/span><\/h3>\n

    \u63a8\u8350\u7406\u7531\uff1a\u79d1\u5b66\u6570\u636e\u8868\u793a<\/span><\/span><\/strong>\u662f\u4e00\u4e2a\u81f4\u529b\u4e8e\u5f00\u53d1\u548c\u4f18\u5316\u7528\u4e8e\u8868\u793a\u3001\u5904\u7406\u548c\u5206\u6790\u79d1\u5b66\u6570\u636e\u7684\u65b9\u6cd5\u548c\u6280\u672f\u7684\u7814\u7a76\u9886\u57df\u3002\u8fd9\u4e00\u9886\u57df\u7684\u6838\u5fc3\u76ee<\/span><\/span>\u6807\u662f<\/span><\/span>\u627e\u5230\u6709\u6548\u65b9\u5f0f<\/span><\/span><\/span>\u6765\u6355<\/span><\/span>\u6349\u548c\u8868\u8fbe\u590d\u6742\u79d1\u5b66\u6570\u636e\u4e2d\u7684\u672c\u8d28\u7279\u5f81\u548c\u5173\u7cfb\uff0c\u4f7f\u8fd9\u4e9b\u6570\u636e\u80fd\u591f\u88ab\u673a\u5668\u5b66\u4e60\u6a21\u578b\u66f4\u597d\u5730\u7406\u89e3\u548c\u5229\u7528\u3002<\/span><\/span><\/p>\n

    \u79d1\u5b66\u6570\u636e\u8868\u793a\u65e9\u671f\u7684\u7814\u7a76\u4e3b\u8981\u96c6\u4e2d\u5728\u4f20\u7edf\u7684\u6570\u636e\u7ed3\u6784\u548c\u7edf\u8ba1\u65b9\u6cd5\u4e0a\uff0c\u968f\u7740\u6df1\u5ea6\u5b66\u4e60\u548c\u5927\u8bed\u8a00\u6a21\u578b\u5174\u8d77\uff0c\u8fd9\u4e2a\u9886\u57df\u5f00\u59cb\u5411\u66f4\u590d\u6742\u548c\u5f3a\u5927\u7684\u8868\u793a\u65b9\u6cd5\u53d1\u5c55\u3002\u56fe\u795e\u7ecf\u7f51\u7edc\u3001\u56fe\u5377\u79ef\u7f51\u7edc\u7b49\u5f00\u521b\u6027\u5730\u5c06\u6df1\u5ea6\u5b66\u4e60\u4e0e\u56fe\u7ed3\u6784\u6570\u636e\u5904\u7406\u7ed3\u5408\uff0c\u53d8\u5206\u81ea\u7f16\u7801\u5668\u5728\u751f\u7269\u5b66\u548c\u7269\u7406\u5b66\u9886\u57df\u5c55\u73b0\u51fa\u5f3a\u5927\u7684\u5efa\u6a21\u80fd\u529b\uff0c\u81ea\u7136\u8bed\u8a00\u5904\u7406\u6280\u672f\u7684\u8fdb\u6b65\u4e5f\u4e3a\u79d1\u5b66\u6587\u732e\u7684\u8868\u793a\u548c\u7406\u89e3\u5e26\u6765\u4e86\u65b0\u7684\u53ef\u80fd\u6027\u3002\u7279\u522b\u662f2023\u5e74\u63d0\u51fa\u7684\u56fe\u6587\u672c\u8054\u5408\u8868\u793a\u65b9\u6cd5\uff0c\u6210\u529f\u5730\u5c06\u6587\u672c\u8bed\u4e49\u548c\u5f15\u7528\u5173\u7cfb\u7ed3\u6784\u7edf\u4e00\u8d77\u6765\uff0c\u4e3a\u79d1\u5b66\u77e5\u8bc6\u7684\u8868\u793a\u5f00\u8f9f\u4e86\u65b0\u9014\u5f84\u3002<\/span><\/span><\/p>\n

    \u6b64\u5916\uff0c\u79d1\u5b66\u6570\u636e\u8868\u793a\u9886\u57df\u9762\u4e34\u7684\u4e3b\u8981\u6311\u6218\u5305\u62ec\u5982\u4f55\u5904\u7406\u591a\u6a21\u6001\u79d1\u5b66\u6570\u636e\u3001\u5982\u4f55\u786e\u4fdd\u8868\u793a\u7684\u53ef\u89e3\u91ca\u6027\uff0c\u4ee5\u53ca\u5982\u4f55\u5728\u4fdd\u6301\u6570\u636e\u5b8c\u6574\u6027\u7684\u540c\u65f6\u5b9e\u73b0\u9ad8\u6548\u7684\u538b\u7f29\u8868\u793a\u7b49\u3002\u5728\u8fd9\u65b9\u9762<\/span><\/span>\uff0c<\/span><\/span>\u5357\u6d0b\u7406\u5de5<\/span><\/span><\/span>\u5927\u5b66<\/span><\/span><\/span>\u3001\u5317\u4eac\u90ae\u7535\u5927\u5b66\u548c\u5b57\u8282\u8df3\u52a8\u5408\u4f5c\u8bad\u7ec3\u7684 LLaVA-Video\u6a21\u578b[5]\uff0c\u901a\u8fc7\u751f\u6210\u8de8\u6a21\u6001\u7684\u5408\u6210\u6570\u636e\uff0c\u5e2e\u52a9\u6a21\u578b\u5b66\u4e60\u4e0d\u540c\u6a21\u6001\u95f4\u7684\u5173\u8054\u5173\u7cfb\uff0c<\/span><\/span>\u5927\u5e45\u63d0\u5347<\/span><\/span><\/span>\u4e86<\/span><\/span><\/span>\u591a\u6a21\u6001\u8868\u5f81\u80fd\u529b\u3002<\/span><\/span><\/span><\/p>\n

    LLaVA-Video\u7814\u7a76\u56e2\u961f\u5f00\u53d1\u4e86\u4e00\u4e2a\u521b\u65b0\u7684\u89c6\u9891\u7406\u89e3\u7cfb\u7edf\uff0c\u5176\u6838\u5fc3\u662f\u57fa\u4e8e\u5927\u89c4\u6a21\u8de8\u6a21\u6001\u5408\u6210\u6570\u636e\u96c6LLaVA-Video-178K[5]\u3002\u8fd9\u4e2a\u6570\u636e\u96c6\u5305\u542b\u4e86178K\u4e2a\u89c6\u9891\u6837\u672c\u548c1.3M\u4e2a\u6307\u4ee4\u8ddf\u968f\u6837\u672c\uff0c\u6db5\u76d6\u4e86\u89c6\u9891\u63cf\u8ff0\u3001\u5f00\u653e\u5f0f\u95ee\u7b54\u548c\u591a\u9879\u9009\u62e9\u95ee\u7b54\u7b49\u591a\u79cd\u4efb\u52a1\u5f62\u5f0f\u3002<\/span><\/span><\/p>\n

    \u7814\u7a76\u56e2\u961f\u91c7\u7528\u4e86\u4e00\u79cd\u57fa\u4e8eGPT-4\u7684\u4e09\u5c42\u7ea7\u9012\u5f52\u751f\u6210\u67b6\u6784\u6765\u6784\u5efa\u89c6\u9891\u5185\u5bb9\u7684\u5c42\u7ea7\u8868\u793a<\/span>\uff08\u5982\u56fe 3 \u6240\u793a\uff09<\/span>\u3002\u8fd9\u79cd\u67b6\u6784\u901a\u8fc7\u65f6\u5e8f\u5d4c\u5165<\/span>\uff08Temporal Embedding\uff09<\/span>\u548c\u8de8\u6a21\u6001\u6ce8\u610f\u529b\u673a\u5236<\/span>\uff08Cross-modal Attention\uff09<\/span>\u6765\u5b9e\u73b0\u89c6\u89c9\u548c\u6587\u672c\u4fe1\u606f\u7684\u6709\u6548\u6574\u5408\u3002\u5177\u4f53\u6765\u8bf4\uff0c\u5bf9\u5f53\u524d\u5c42\u7ea7\u7684\u65f6\u95f4\u70b9 t\uff0c\u4ee5\u53ca\u6700\u540e\u7684\u65f6\u95f4\u70b9 T\u3002\u7cfb\u7edf\u7684\u4e09\u4e2a\u5c42\u7ea7\u5206\u522b\u627f\u62c5\u4e0d\u540c\u7684\u8868\u793a\u4efb\u52a1\uff1a<\/span><\/span><\/p>\n

    \uff08a\uff09 \u5728\u7b2c\u4e00\u7ea7\uff0c\u4e3a\u65f6\u95f4\u70b9 t \u751f\u6210\u5b57\u5e55\u65f6\uff0c\u53c2\u8003\u4e86\u5f53\u524d\u65f6\u95f4\u70b9\u7684\u5e27\u753b\u9762\u3001\u524d\u4e00\u4e2a\u65f6\u95f4\u70b9\u7684\u5b57\u5e55\uff0c\u4ee5\u53ca<\/span>\uff08\u5982\u679c\u9002\u7528\uff09<\/span>\u6700\u8fd1\u7684\u7b2c\u4e8c\u7ea7\u603b\u7ed3\u63cf\u8ff0\u3002\uff08b\uff09 \u5728\u7b2c\u4e8c\u7ea7\uff0c\u4e3a\u65f6\u95f4\u70b9 t \u751f\u6210\u5b57\u5e55\u65f6\uff0c\u57fa\u4e8e\u524d\u4e00\u4e2a\u7b2c\u4e8c\u533a\u95f4\u7684\u5b57\u5e55\u548c\u6700\u8fd1\u4e09\u4e2a\u7b2c\u4e00\u533a\u95f4\u7684\u5b57\u5e55\u3002\uff08c\uff09 \u5728\u7b2c\u4e09\u7ea7\uff0c\u4e3a\u6700\u7ec8\u65f6\u95f4\u70b9 T \u751f\u6210\u6574\u4f53\u5b57\u5e55\u65f6\uff0c\u53c2\u8003\u4e86\u6700\u8fd1\u7684\u7b2c\u4e8c\u533a\u95f4\u5b57\u5e55\u548c\u5f53\u524d\u7684\u7b2c\u4e00\u533a\u95f4\u5b57\u5e55\uff0c\u751f\u6210\u5bf9\u6574\u4e2a\u89c6\u9891\u5185\u5bb9\u7684\u5168\u5c40\u8bed\u4e49\u8868\u793a\u3002<\/span><\/span><\/p>\n

    <\/section>\n

    LLaVA-Video-178K\u89c6\u9891\u8be6\u7ec6\u63cf\u8ff0\u751f\u6210\u6d41\u7a0b\uff5c\u56fe\u7247\u6765\u6e90\uff1aZhang et al. (2024) “Video Instruction Tuning With Synthetic Data.” <\/span><\/span><\/em><\/span><\/p>\n

    \u8fd9\u79cd\u591a\u5c42\u7ea7\u8868\u793a\u65b9\u6cd5\u7684\u521b\u65b0\u4e4b\u5904\u5728\u4e8e\u5b83\u4e0d\u4ec5\u89e3\u51b3\u4e86\u89c6\u9891\u5185\u5bb9\u7684\u65f6\u5e8f\u4f9d\u8d56\u95ee\u9898\uff0c\u8fd8\u5b9e\u73b0\u4e86\u89c6\u89c9\u548c\u8bed\u8a00\u6a21\u6001\u7684\u6df1\u5ea6\u878d\u5408\u3002\u901a\u8fc7\u9012\u8fdb\u5f0f\u7684\u591a\u5c42\u7ea7\u8868\u793a\u6765\u6355\u6349\u89c6\u9891\u5185\u5bb9\u7684\u4e0d\u540c\u7c92\u5ea6\u7279\u5f81\uff0c\u4ece\u800c\u5b9e\u73b0\u4ece\u5c40\u90e8\u7ec6\u8282\u5230\u5168\u5c40\u8bed\u4e49\u7684\u5b8c\u6574\u8868\u793a\u6846\u67b6\u3002\u8fd9\u79cd\u65b9\u6cd5\u4e0d\u4ec5\u63d0\u9ad8\u4e86\u6a21\u578b\u7684\u6027\u80fd\uff0c\u4e5f\u4e3a\u672a\u6765\u7684\u591a\u6a21\u6001\u5185\u5bb9\u7406\u89e3\u7814\u7a76\u63d0\u4f9b\u4e86\u65b0\u7684\u601d\u8def\u3002<\/span><\/span><\/p>\n

    \u968f\u7740\u65b0\u6280\u672f\u7684\u4e0d\u65ad\u6d8c\u73b0\u548c\u4ea4\u53c9\u5b66\u79d1\u7684\u6df1\u5165\u878d\u5408\uff0c\u79d1\u5b66\u6570\u636e\u8868\u793a\u9886\u57df\u6709\u671b\u5728\u63a8\u52a8\u79d1\u5b66\u53d1\u73b0\u548c\u4fc3\u8fdb\u5b66\u79d1\u53d1\u5c55\u65b9\u9762\u53d1\u6325\u66f4\u52a0\u91cd\u8981\u7684\u4f5c\u7528\u3002<\/span><\/span><\/p>\n

    \u6a21\u578b\u81ea\u6211\u6539\u8fdb\uff1a\u901a\u8fc7\u5408\u6210\u6570\u636e\u63d0\u5347AI\u7cfb\u7edf\u6027\u80fd<\/span><\/span><\/h3>\n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong>\u5728\u4eba\u5de5\u667a\u80fd\u7814\u7a76\u9886\u57df\u4e2d\uff0c\u6a21\u578b\u81ea\u6211\u6539\u8fdb\u4ee3\u8868\u4e86\u4e00\u4e2a\u6781\u5177\u524d\u666f\u7684\u53d1\u5c55\u65b9\u5411\u3002\u8fd9\u79cd\u65b9\u6cd5\u8ba9AI\u7cfb\u7edf\u80fd\u591f\u901a\u8fc7\u751f\u6210\u548c\u5229\u7528\u5408\u6210\u6570\u636e\u6765\u589e\u5f3a\u81ea\u8eab\u80fd\u529b\uff0c\u65e0\u9700\u8fc7\u5ea6\u4f9d\u8d56\u5916\u90e8\u6570\u636e\u6e90\u6216\u66f4\u5f3a\u5927\u7684\u6559\u5e08\u6a21\u578b\u3002\u8fd9\u4e00\u6280\u672f\u4e0d\u4ec5\u964d\u4f4e\u4e86\u5bf9\u9ad8\u8d28\u91cf\u8bad\u7ec3\u6570\u636e\u7684\u4f9d\u8d56\uff0c\u66f4\u5f00\u521b\u4e86AI\u7cfb\u7edf\u81ea\u4e3b\u5b66\u4e60\u548c\u8fdb\u5316\u7684\u65b0\u8303\u5f0f\u3002\u968f\u7740\u5927\u8bed\u8a00\u6a21\u578b\u7684\u53d1\u5c55\uff0c\u8fd9\u79cd\u81ea\u6211\u6539\u8fdb\u673a\u5236\u5c55\u73b0\u51fa\u8d8a\u6765\u8d8a\u91cd\u8981\u7684\u4ef7\u503c\u3002<\/span><\/span><\/p>\n

    \u6a21\u578b\u81ea\u6211\u6539\u8fdb\u7684\u7814\u7a76\u805a\u7126\u4e8e\u901a\u8fc7\u5408\u6210\u6570\u636e\u589e\u5f3a\u6a21\u578b\u6027\u80fd\u3002<\/span><\/span><\/span>\u8fd9<\/span><\/span>\u4e2a\u9886\u57df\u7684\u4e00\u4e2a\u4ee3\u8868\u6027\u5de5\u4f5c\u662f\u81ea\u8bad\u7ec3<\/span>\uff08Self-training\uff09<\/span>\u65b9\u6cd5\uff0c\u6a21\u578b\u9996\u5148\u5728\u6709\u9650\u7684\u6807\u6ce8\u6570\u636e\u4e0a\u8bad\u7ec3\uff0c\u7136\u540e\u751f\u6210\u65b0\u7684\u8bad\u7ec3\u6837\u672c\u6765\u6539\u8fdb\u81ea\u8eab\u3002\u53e6\u4e00\u4e2a\u521b\u65b0\u6027\u7684\u7814\u7a76\u662f\u6a21\u578b\u84b8\u998f<\/span>\uff08Model Distillation\uff09<\/span>\u4e0e\u5408\u6210\u6570\u636e\u7684\u7ed3\u5408\uff0c\u901a\u8fc7\u751f\u6210\u7279\u5b9a\u7684\u8bad\u7ec3\u6837\u672c\u6765\u4f18\u5316\u77e5\u8bc6\u8fc1\u79fb\u8fc7\u7a0b\u3002\u8fd9\u4e9b\u65b9\u6cd5\u5c55\u793a\u4e86\u5982\u4f55\u901a\u8fc7\u5408\u6210\u6570\u636e\u6765\u5b9e\u73b0\u6a21\u578b\u80fd\u529b\u7684\u6301\u7eed\u63d0\u5347\u3002<\/span><\/span><\/p>\n

    \u5728\u8fd9\u4e00\u9886\u57df\u7684\u6700\u65b0\u7a81\u7834\u4e2d\uff0c\u5361\u5185\u57fa\u6885\u9686\u5927\u5b66\u548c\u6e05\u534e\u5927\u5b66\u7814\u7a76\u56e2\u961f\u5f00\u53d1\u7684SELF-GUIDE\u65b9\u6cd5[6]\u5c55\u793a\u4e86\u663e\u8457\u6210\u679c\u3002\u8fd9\u79cd\u521b\u65b0\u65b9\u6cd5\u901a\u8fc7\u8bbe\u8ba1\u9ad8\u6548\u7684\u591a\u9636\u6bb5\u751f\u6210\u673a\u5236\uff0c\u4f7f\u8bed\u8a00\u6a21\u578b\u80fd\u591f\u81ea\u4e3b\u751f\u6210\u4efb\u52a1\u7279\u5b9a\u7684\u8bad\u7ec3\u6570\u636e\u3002\u7cfb\u7edf\u901a\u8fc7\u9010\u6b65\u751f\u6210\u201c\u8f93\u5165-\u8f93\u51fa\u201d\u6570\u636e\u5bf9\uff0c\u5e76\u7ecf\u8fc7\u4e25\u683c\u7684\u7b5b\u9009\u8fc7\u7a0b\uff0c\u786e\u4fdd\u751f\u6210\u6570\u636e\u7684\u8d28\u91cf\u548c\u76f8\u5173\u6027\u3002\u8fd9\u4e9b\u81ea\u751f\u6210\u7684\u6570\u636e\u968f\u540e\u88ab\u7528\u4e8e\u6a21\u578b\u7684\u8fdb\u4e00\u6b65\u5fae\u8c03\uff0c\u5f62\u6210\u4e00\u4e2a\u826f\u6027\u7684\u81ea\u6211\u63d0\u5347\u5faa\u73af\u3002<\/span><\/span><\/p>\n

    <\/section>\n

    SELF-GUIDE \u9488\u5bf9\u751f\u6210\u4efb\u52a1\u7684\u6d41\u7a0b\uff5c\u56fe\u7247\u6765\u6e90\uff1aZhao et al.  (2024) “Self-guide: Better task-specific instruction following via self-synthetic finetuning.” <\/span><\/span><\/em><\/span><\/p>\n

    \u7814\u7a76\u7ed3\u679c\u4ee4\u4eba\u632f\u594b\uff1aSELF-GUIDE\u5728\u591a\u4e2a\u4efb\u52a1\u4e0a\u90fd\u53d6\u5f97\u4e86\u663e\u8457\u7684\u6027\u80fd\u63d0\u5347\u3002\u5728\u5206\u7c7b\u4efb\u52a1\u4e2d\uff0c\u6a21\u578b\u6027\u80fd\u83b7\u5f97\u4e86\u7ea615%\u7684\u7edd\u5bf9\u63d0\u5347\uff1b\u5728\u751f\u6210\u4efb\u52a1\u4e2d\uff0c\u63d0\u5347\u5e45\u5ea6\u66f4\u662f\u8fbe\u5230\u4e86\u7ea618%\u3002\u8fd9\u4e9b\u6570\u636e\u5145\u5206\u8bc1\u660e\u4e86\u8be5\u65b9\u6cd5\u5728\u63d0\u5347\u6a21\u578b\u80fd\u529b\u65b9\u9762\u7684\u6709\u6548\u6027\u3002<\/span><\/span><\/p>\n

    SELF-GUIDE\u7684\u6210\u529f\u4e0d\u4ec5\u8bc1\u660e\u4e86\u6a21\u578b\u81ea\u6211\u6539\u8fdb\u7b56\u7565\u7684\u53ef\u884c\u6027\uff0c\u4e5f\u4e3a\u672a\u6765AI\u7cfb\u7edf\u7684\u53d1\u5c55\u63d0\u4f9b\u4e86\u65b0\u7684\u601d\u8def\u3002\u8fd9\u79cd\u81ea\u4e3b\u5b66\u4e60\u548c\u6539\u8fdb\u7684\u80fd\u529b\uff0c\u53ef\u80fd\u4f1a\u5e26\u6765\u66f4\u667a\u80fd\u3001<\/span><\/span>\u9002\u5e94\u6027<\/span><\/span><\/span>\u66f4<\/span><\/span><\/span>\u5f3a\u7684AI\u7cfb\u7edf<\/span><\/span><\/span>\uff0c\u63a8\u52a8\u6574\u4e2a\u9886\u57df\u5411\u7740\u66f4\u9ad8\u6c34\u5e73\u53d1\u5c55\u3002\u901a\u8fc7\u6301\u7eed\u7684\u81ea\u6211\u6539\u8fdb\uff0cA<\/span><\/span>I\u7cfb\u7edf\u6709\u671b\u5728\u66f4\u591a\u590d\u6742\u4efb\u52a1\u4e2d\u5c55\u73b0\u51fa\u66f4\u5f3a\u7684\u6027\u80fd\u548c\u9002\u5e94\u80fd\u529b<\/span><\/span><\/p>\n

    \u6570\u636e\u57fa\u7840\u8bbe\u65bd\uff1a\u4eba\u4f53\u86cb\u767d\u8d28\u7ec4\u8ba1\u5212<\/span><\/span><\/h3>\n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong> \u9996\u6b21\u63d0\u51fa\u6784\u5efa\u4eba\u4f53\u86cb\u767d\u8d28\u7ec4\u7684\u7cbe\u786e\u201c\u5bfc\u822a\u7cfb\u7edf\u201d[18]\uff0c\u4e0d\u4ec5\u5c06\u5f7b\u5e95\u6539\u53d8\u6211\u4eec\u5bf9\u4eba\u4f53\u751f\u547d\u6d3b\u52a8\u7684\u8ba4\u8bc6\uff0c\u968f\u7740\u5355\u7ec6\u80de\u86cb\u767d\u8d28\u7ec4\u5b66\u7b49\u6280\u672f[19]\u7684\u5feb\u901f\u53d1\u5c55\uff0c\u66f4\u6709\u671b\u63a8\u52a8\u533b\u5b66\u8303\u5f0f\u4ece\u88ab\u52a8\u6cbb\u7597\u5411\u4e3b\u52a8\u9884\u9632\u548c\u7cbe\u786e\u533b\u7597\u8f6c\u53d8\u3002<\/span><\/span><\/p>\n

    \u5728\u751f\u547d\u79d1\u5b66\u7814\u7a76\u4e2d\uff0c\u57fa\u56e0\u7ec4\u544a\u8bc9\u6211\u4eec\u751f\u547d\u7684\u53ef\u80fd\u6027\uff0c\u86cb\u767d\u8d28\u7ec4\u5219\u5c55\u793a\u4e86\u751f\u547d\u7684\u73b0\u5b9e\u72b6\u6001\u3002\u4eba\u4f53\u5185\u7ea637\u4e07\u4ebf\u4e2a\u7ec6\u80de\u867d\u7136\u5171\u4eab\u76f8\u540c\u7684\u57fa\u56e0\u7ec4\uff0c\u5374\u80fd\u5c55\u73b0\u51fa\u4e30\u5bcc\u591a\u6837\u7684\u5f62\u6001\u548c\u529f\u80fd\u3002\u8fd9\u79cd\u795e\u5947\u7684\u5206\u5316\u548c\u8c03\u63a7\u8fc7\u7a0b\uff0c\u6b63\u662f\u7531\u86cb\u767d\u8d28\u7684\u7cbe\u786e\u8868\u8fbe\u548c\u8c03\u63a7\u7f51\u7edc\u6240\u4e3b\u5bfc\u3002\u7136\u800c\uff0c\u76ee\u524d\u6211\u4eec\u5bf9\u4eba\u4f53\u86cb\u767d\u8d28\u7ec4\u7684\u8ba4\u8bc6\u4ecd\u7136\u5341\u5206\u6709\u9650\uff0c\u8fd9\u6781\u5927\u5730\u5236\u7ea6\u4e86\u7cbe\u51c6\u533b\u7597\u7684\u53d1\u5c55\u3002<\/span><\/span><\/p>\n

    <\/section>\n

    \u03c0-HuB\u9879\u76ee\u7684\u603b\u4f53\u76ee\u6807\uff5c\u56fe\u7247\u6765\u6e90\uff1aHe, F., Aebersold, R., Baker, M.S. et al. \u03c0-HuB: the proteomic navigator of the human body. Nature <\/span><\/span><\/em><\/span>636<\/span><\/span><\/strong><\/em><\/span>, 322\u2013331 (2024).” <\/span><\/span><\/em><\/span><\/span><\/p>\n

    \u03c0-HuB\u9879\u76ee\u63d0\u51fa\u4e86\u4e09\u4e2a\u7a81\u7834\u6027\u7684\u7814\u7a76\u76ee\u6807\uff1a\u9996\u5148\u662f\u63ed\u793a\u4eba\u4f53\u7684\u6784\u5efa\u539f\u7406\uff0c\u901a\u8fc7\u6700\u65b0\u7684\u5355\u7ec6\u80de\u86cb\u767d\u8d28\u7ec4\u5b66\u6280\u672f\uff0c\u7ed8\u5236\u4e0d\u540c\u7c7b\u578b\u7ec6\u80de\u4e2d\u86cb\u767d\u8d28\u7684\u7cbe\u786e\u7ec4\u6210\u53ca\u5176\u8c03\u63a7\u7f51\u7edc\uff1b\u5176\u6b21\u662f\u5efa\u7acb\u201c\u5143\u4eba\u7c7b\u201d<\/span> (Meta Homo Sapiens) <\/span>\u8ba1\u7b97\u6a21\u578b\uff0c\u8ffd\u8e2a\u8bb0\u5f55\u4eba\u4f53\u86cb\u767d\u8d28\u7ec4\u5728\u4e0d\u540c\u751f\u547d\u9636\u6bb5\u7684\u52a8\u6001\u53d8\u5316\u89c4\u5f8b\uff1b\u6700\u540e\u662f\u5f00\u53d1\u03c0-HuB\u5bfc\u822a\u7cfb\u7edf\uff0c\u5c06\u86cb\u767d\u8d28\u7ec4\u5b66\u6570\u636e\u8f6c\u5316\u4e3a\u75be\u75c5\u9884\u9632\u3001\u8bca\u65ad\u548c\u6cbb\u7597\u7684\u7cbe\u786e\u6307\u5bfc\u3002<\/span><\/span><\/p>\n

    \u5728\u6280\u672f\u8def\u7ebf\u4e0a\uff0c\u8be5\u9879\u76ee\u91c7\u7528\u4e86\u591a\u5c42\u6b21\u3001\u591a\u7ef4\u5ea6\u7684\u7814\u7a76\u7b56\u7565\u3002\u5728\u7ec6\u80de\u6c34\u5e73\uff0c\u8fd0\u7528\u6700\u65b0\u7684\u5355\u7ec6\u80de\u86cb\u767d\u8d28\u7ec4\u5b66\u6280\u672f\uff0c\u5b9e\u73b0\u5bf9\u4eba\u4f53\u4e3b\u8981\u5668\u5b98\u7ec6\u80de\u7c7b\u578b\u7684\u7cbe\u786e\u5206\u6790\uff1b\u5728\u7fa4\u4f53\u6c34\u5e73\uff0c\u901a\u8fc7\u5927\u89c4\u6a21\u961f\u5217\u7814\u7a76[20]\uff0c\u63ed\u793a\u751f\u6d3b\u65b9\u5f0f\u3001\u73af\u5883\u56e0\u7d20\u7b49\u5bf9\u86cb\u767d\u8d28\u7ec4\u7684\u5f71\u54cd\uff1b\u5728\u4e34\u5e8a\u5e94\u7528\u5c42\u9762\uff0c\u901a\u8fc7\u5efa\u7acb\u6807\u51c6\u5316\u7684\u751f\u7269\u6837\u672c\u5e93\u548c\u6570\u636e\u5206\u6790\u5e73\u53f0\uff0c\u63a8\u52a8\u86cb\u767d\u8d28\u7ec4\u5b66\u5728\u7cbe\u51c6\u533b\u7597\u4e2d\u7684\u8f6c\u5316\u5e94\u7528\u3002<\/span><\/span><\/p>\n

    \u7814\u7a76\u6210\u679c\u7684\u5e94\u7528\u524d\u666f\u4ee4\u4eba\u632f\u594b\u3002\u5728\u7b2c\u4e00\u9636\u6bb5<\/span>\uff082024-2033\u5e74\uff09<\/span>\uff0c\u9879\u76ee\u5c06\u91cd\u70b9\u5b9e\u73b0\u4e09\u4e2a\u7a81\u7834\uff1a\u5b8c\u6210\u4eba\u4f53\u4e3b\u8981\u5668\u5b98\u7684\u7ec6\u80de\u7c7b\u578b\u86cb\u767d\u8d28\u56fe\u8c31\uff1b\u5efa\u7acb\u57fa\u4e8e\u86cb\u767d\u8d28\u7ec4\u5b66\u7684\u5065\u5eb7\u8bc4\u4f30\u4f53\u7cfb\uff1b\u5f00\u53d1\u65b0\u7684\u75be\u75c5\u65e9\u671f\u8bca\u65ad\u6807\u5fd7\u7269\u548c\u6cbb\u7597\u9776\u70b9\u3002\u8fd9\u4e9b\u6210\u679c\u5c06\u4e3a\u5b9e\u73b0\u66f4\u5177\u6709\u5b9e\u8df5\u667a\u6167\u7684\u7cbe\u51c6\u533b\u5b66\u5960\u5b9a\u575a\u5b9e\u57fa\u7840\uff0c\u63a8\u52a8\u533b\u5b66\u6a21\u5f0f\u4ece\u88ab\u52a8\u6cbb\u7597\u5411\u4e3b\u52a8\u9884\u9632\u8f6c\u53d8\u3002<\/span><\/span><\/p>\n

    \u03c0-HuB\u9879\u76ee\u7684\u521b\u65b0\u6027\u4e0d\u4ec5\u4f53\u73b0\u5728\u5176\u79d1\u5b66\u76ee\u6807\u4e0a\uff0c\u4e5f\u4f53\u73b0\u5728\u5176\u7ec4\u7ec7\u6a21\u5f0f\u4e0a\u3002\u9879\u76ee\u91c7\u7528\u5f00\u653e\u79d1\u5b66\u7684\u7406\u5ff5\uff0c\u5efa\u7acb\u4e86\u56fd\u9645\u5316\u7684\u7814\u7a76\u56e2\u961f\u7f51\u7edc\uff0c\u5e76\u627f\u8bfa\u5c06\u7814\u7a76\u6570\u636e\u548c\u5206\u6790\u5de5\u5177\u5411\u5168\u7403\u79d1\u7814\u754c\u5f00\u653e\u5171\u4eab\u3002\u8fd9\u79cd\u534f\u4f5c\u6a21\u5f0f\u5c06\u5927\u5927\u52a0\u901f\u86cb\u767d\u8d28\u7ec4\u5b66\u7814\u7a76\u7684\u8fdb\u5c55\u3002\u03c0-HuB\u9879\u76ee\u7684\u542f\u52a8\uff0c\u6807\u5fd7\u7740\u4eba\u7c7b\u63a2\u7d22\u751f\u547d\u5965\u79d8\u548c\u8ffd\u6c42\u5065\u5eb7\u7684\u5f81\u7a0b\u53c8\u8fc8\u51fa\u4e86\u91cd\u8981\u4e00\u6b65\u3002<\/span><\/span><\/p>\n


    <\/span><\/h2>\n

    \n
    \n
    \n
    \n
    \n

    <\/span><\/section>\n

    <\/span><\/section>\n
    \n
    <\/section>\n<\/section>\n<\/section>\n
    \n

    \u6311\u6218\u4e0e\u5c55\u671b<\/span><\/strong><\/span><\/strong><\/strong><\/p>\n<\/section>\n

    \n

    <\/span><\/section>\n

    <\/span><\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n<\/h3>\n


    <\/span><\/p>\n

    \u867d\u7136 OpenAI\u7684CEO Sam Altman \u9884\u89c1AI\u6700\u7ec8\u5c06\u80fd\u4ea7\u751f\u8db3\u591f\u4f18\u8d28\u7684\u5408\u6210\u6570\u636e\u6765\u8bad\u7ec3\u81ea\u8eab\uff0c\u4f46\u6709\u7814\u7a76\u8868\u660e[21]\uff0c\u52a3\u8d28\u4fe1\u606f\u548c\u4e0d\u5f53\u7684\u8bad\u7ec3\u65b9\u6cd5\u4ecd\u53ef\u80fd\u5bfc\u81f4LLMs\u201c\u6a21\u578b\u5d29\u6e83\u201d\u3002\u56e0\u6b64\u5408\u6210\u6570\u636e\u867d\u7136\u6f5c\u529b\u5de8\u5927\uff0c<\/span><\/span>\u4f46\u5728\u5b9e\u9645\u5e94\u7528\u4e2d\u4ecd\u9762\u4e34\u7740<\/span><\/span><\/span>\u8bf8\u591a<\/span><\/span><\/span>\u6311\u6218\u3002<\/span><\/span><\/span><\/p>\n

    \u6570\u636e\u8d28\u91cf\u4e0e\u4fdd\u771f\u5ea6\u95ee\u9898<\/span><\/span><\/strong>\u81f3\u5173\u91cd\u8981\u3002\u5408\u6210\u6570\u636e\u7684\u8d28\u91cf\u76f4\u63a5\u5f71\u54cd\u6a21\u578b\u6027\u80fd\u3002\u57fa\u4e8e\u865a\u5047\u3001\u5e7b\u89c9\u6216\u6709\u504f\u5dee\u7684\u6570\u636e\u8bad\u7ec3\u7684\u6a21\u578b\u4e0d\u4ec5\u53ef\u80fd\u5728\u73b0\u5b9e\u573a\u666f\u4e2d\u8868\u73b0\u4e0d\u4f73\uff0c\u7f3a\u4e4f\u6cdb\u5316\u80fd\u529b[22]\uff0c\u5982\u679c\u8bbe\u8ba1\u548c\u9a8c\u8bc1\u4e0d\u5f53\uff0c\u8fd8\u53ef\u80fd\u4f1a\u653e\u5927\u5df2\u6709\u504f\u5dee\u6216\u5f15\u5165\u65b0\u7684\u504f\u5dee[23]\u3002<\/span><\/span>\u53bb\u6c61\u67d3\u8bc4\u4f30\uff08decontamination evaluation\uff09\u96be\u5ea6\u95ee\u9898<\/span><\/span><\/strong>\u4ecd\u9700\u8fdb\u4e00\u6b65\u63a2\u7d22\u3002\u7531\u4e8e\u5408\u6210\u6570\u636e\u53ef\u80fd\u5305\u542b\u91cd\u8ff0\u7684\u57fa\u51c6\u6570\u636e\u7248\u672c\uff0c\u4f20\u7edf\u7684\u8bcd\u7ea7\u522b\u53bb\u6c61\u67d3\u65b9\u6cd5\u53ef\u80fd\u4f1a\u5931\u6548[24]\uff0c\u5bfc\u81f4\u65e0\u6cd5\u6709\u6548\u533a\u5206\u6a21\u578b\u662f\u771f\u6b63\u7406\u89e3\u548c\u5b66\u4e60\u4e86\u65b0\u77e5\u8bc6\uff0c\u8fd8\u662f\u4ec5\u4ec5\u5728\u8bb0\u5fc6\u548c\u91cd\u590d\u8bad\u7ec3\u6570\u636e\u4e2d\u7684\u5185\u5bb9\u3002<\/span><\/span>\u9690\u79c1\u4e0e\u4f26\u7406\u95ee\u9898<\/span><\/span><\/strong>\u4e5f\u4f9d\u7136\u4e25\u5cfb\u3002\u5c3d\u7ba1\u5408\u6210\u6570\u636e\u63d0\u4f9b\u4e86\u4e00\u4e9b\u4e0d\u635f\u5bb3\u4e2a\u4eba\u9690\u79c1\u7684\u9014\u5f84[25]\uff0c\u4f46\u5728\u654f\u611f\u9886\u57df\u4e2d<\/span><\/span>\u4f7f\u7528\u5408\u6210\u6570\u636e\u65f6\u4ecd\u5b58\u5728\u4f26\u7406\u95ee\u9898\u3002\u6709\u7814\u7a76[26]\u8868\u660e\u53ef\u4ee5\u4ece\u8bad\u7ec3\u6570\u636e\u96c6\u4e2d\u63d0\u53d6\u7279\u5b9a\u4fe1\u606f\uff0c<\/span><\/span>\u8fd9\u610f\u5473\u7740\u5408\u6210\u6570\u636e\u53ef\u80fd<\/span><\/span><\/span>\u4f1a\u5728<\/span><\/span><\/span>\u65e0\u610f\u4e2d\u66b4\u9732\u57fa\u7840\u8bad\u7ec3\u6570\u636e<\/span><\/span><\/span>\u7684<\/span><\/span><\/span>\u67d0\u4e9b\u654f\u611f\u4fe1\u606f\u548c\u9690\u79c1\u5185\u5bb9[27]\u3002<\/span><\/span><\/span><\/p>\n

    \u5f53\u7136\uff0c\u5728\u9762\u5bf9\u6311\u6218\u7684\u540c\u65f6\uff0c\u5408\u6210\u6570\u636e\u672a\u6765\u540c\u6837\u6709\u4e0b\u9762\u51e0\u4e2a\u5177\u6709\u524d\u666f\u7684\u7814\u7a76\u53d1\u5c55\u65b9\u5411\u503c\u5f97\u671f\u5f85\u3002<\/span><\/span><\/p>\n

    \u6570\u636e\u8d28\u91cf\u4e0e\u591a\u6837\u6027\u63d0\u5347<\/span><\/span><\/strong>\u3002\u5408\u6210\u6570\u636e\u672a\u6765\u7814\u7a76\u5e94\u4e13\u6ce8\u4e8e\u5f00\u53d1\u57fa\u4e8eGANs\u548c\u6269\u6563\u6a21\u578b\u7b49\u7684\u65b0\u6280\u672f\uff0c\u5e76\u7ed3\u5408\u9886\u57df\u7279\u5b9a\u77e5\u8bc6\uff0c\u901a\u8fc7RAG\u7b49\u65b9\u6cd5\u786e\u4fdd\u6570\u636e\u8d28\u91cf\u548c\u591a\u6837\u6027\uff0c\u62d3\u5c55\u5230\u5305\u62ec\u533b\u7597\u3001\u91d1\u878d\u548c\u793e\u4f1a\u79d1\u5b66\u7b49\u5e94\u7528\u9886\u57df\u3002<\/span><\/span><\/p>\n

    \u6570\u636e\u8d28\u91cf\u76d1\u7ba1\u673a\u5236\u7814\u7a76<\/span><\/span><\/strong>\u3002\u968f\u7740AI\u590d\u6742\u6027\u63d0\u5347\uff0c\u4f20\u7edf\u6570\u636e\u8bc4\u4f30\u65b9\u6cd5\u5df2\u4e0d\u8db3\u4ee5\u5e94\u5bf9\u6311\u6218\u3002\u9700\u8981\u5efa\u7acb\u66f4\u7cfb\u7edf\u5316\u7684\u76d1\u7ba1\u6846\u67b6\uff0c\u5b9e\u73b0\u5408\u6210\u6570\u636e\u7684\u5168\u9762\u8d28\u91cf\u8bc4\u4f30\u3001\u81ea\u52a8\u5316\u7b5b\u67e5\u548c\u591a\u573a\u666f\u9a8c\u8bc1\u3002<\/span><\/span><\/p>\n

    \u63a2\u7d22\u5408\u6210\u6570\u636e\u7684\u89c4\u6a21\u6548\u5e94<\/span><\/span><\/strong>\u3002\u9274\u4e8e\u4e00\u4e9b\u7ecf\u8fc7\u7cbe\u5fc3\u8bad\u7ec3\u7684\u5c0f\u578b\u8bed\u8a00\u6a21\u578b\uff0c\u80fd\u8d85\u51faChinchilla\u5b9a\u5f8b<\/span>\uff08\u6a21\u578b\u6027\u80fd\u4e0e\u8bad\u7ec3\u6570\u636e\u89c4\u6a21\u548c\u6a21\u578b\u53c2\u6570\u91cf\u5448\u6b63\u6bd4\uff09<\/span>\u7684\u9884\u6d4b\uff0c\u672a\u6765\u7814\u7a76\u9700\u8981\u63a2\u7d22\u5408\u6210\u6570\u636e\u89c4\u6a21\u7684\u201c\u8d28\u91cf-\u6570\u91cf\u201d\u6743\u8861\u673a\u5236\uff0c\u4ee5\u627e\u5230\u63d0\u5347\u6a21\u578b\u6027\u80fd\u7684\u6700\u4f18\u6570\u636e\u7b56\u7565\u3002<\/span><\/span><\/p>\n

    \u81ea\u6211\u6539\u8fdb\u80fd\u529b\u7684\u6d8c\u73b0\u7814\u7a76\u3002<\/span><\/span><\/strong>\u6700\u65b0\u7814\u7a76\u663e\u793a\u51fa\u79ef\u6781\u8fdb\u5c55\uff0c\u4f46\u4ecd\u9700\u6df1\u5165\u63a2\u7d22\u5176\u7406\u8bba\u57fa\u7840\u3001\u5c40\u9650\u6027\u548c\u6f5c\u5728\u98ce\u9669\uff0c\u4ece\u800c\u63a8\u52a8\u66f4\u5177\u9002\u5e94\u6027\u548c\u81ea\u4e3b\u6027\u7684AI\u5b66\u4e60\u8fc7\u7a0b\u3002<\/span><\/span><\/p>\n

    \u591a\u6e90\u6570\u636e\u878d\u5408\u57fa\u7840\u8bbe\u65bd\u5efa\u8bbe<\/span><\/span><\/strong>\u3002\u9700\u8981\u89e3\u51b3\u6570\u636e\u6807\u51c6\u5316\u548c\u8bed\u4e49\u5bf9\u9f50\u95ee\u9898\uff0c\u7279\u522b\u662f\u5728\u5904\u7406\u8de8\u7ec4\u7ec7\u3001\u8de8\u9886\u57df\u7684\u654f\u611f\u6570\u636e\u65f6\uff0c\u5982\u4f55\u5728\u4fc3\u8fdb\u6570\u636e\u5171\u4eab\u7684\u540c\u65f6\u786e\u4fdd\u6570\u636e\u5b89\u5168\u4e0e\u9690\u79c1\u95ee\u9898\u3002\u8fd1\u671f\uff0cAnthropic \u5f00\u6e90\u7684\u300c\u6a21\u578b\u4e0a\u4e0b\u6587\u534f\u8bae\u300dMCP<\/span>\uff08Model Context Protocol\uff09<\/span>[28] \u5df2\u7ecf\u8fc8\u51fa\u4e86\u91cd\u8981\u4e00\u6b65\u3002<\/span><\/span><\/p>\n

    \u603b\u4e4b\uff0c\u867d\u7136\u76ee\u524d\u4ecd\u9762\u4e34\u8bf8\u591a\u6311\u6218\uff0c\u4f46\u6b63\u5982\u8bb8\u591a\u7814\u7a76\u8005\u6240\u9884\u89c1\u7684\uff0c\u5408\u6210\u6570\u636e\u4e0d\u4ec5\u662f\u89e3\u51b3\u5f53\u524dAI\u53d1\u5c55\u74f6\u9888\u7684\u5de5\u5177\uff0c\u66f4\u53ef\u80fd\u6210\u4e3a\u5f00\u542f\u4e0b\u4e00\u4ee3\u4eba\u5de5\u667a\u80fd\u9769\u547d\u7684\u94a5\u5319\u3002\u7279\u522b\u662f\u5728\u81ea\u6211\u6539\u8fdb\u80fd\u529b\u65b9\u9762\u7684\u7814\u7a76\uff0c\u80fd\u5e2e\u52a9\u6211\u4eec\u5b9e\u73b0\u4ece\u72ed\u4e49AI\u5230\u901a\u7528\u4eba\u5de5\u667a\u80fd\u7684\u8de8\u8d8a\uff0c\u63a8\u52a8\u4eba\u7c7b\u548cAI\u5411\u66f4\u667a\u80fd\u3001\u66f4\u52a0\u7f8e\u597d\u7684\u672a\u6765\u8fc8\u8fdb\u3002<\/span><\/span><\/p>\n

    \u53c2\u8003\u6587\u732e<\/span><\/span><\/h2>\n

    [1] Villalobos\uff0cPablo\uff0cet al. “Will we run out of data? an analysis of the limits of scaling datasets in machine learning.” <\/span><\/span>arXiv preprint arXiv:2211.04325<\/span><\/span><\/em> \uff082022\uff09<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u6df1\u5165\u5206\u6790\u673a\u5668\u5b66\u4e60\u6570\u636e\u96c6\u6269\u5c55\u7684\u6781\u9650\u95ee\u9898\uff0c\u5bf9\u5f53\u524d\u5927\u6a21\u578b\u53d1\u5c55\u4e2d\u7684\u6570\u636e\u74f6\u9888\u63d0\u4f9b\u4e86\u91cd\u8981\u89c1\u89e3\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [2] Gilardi\uff0cFabrizio\uff0cMeysam Alizadeh\uff0cand Ma\u00ebl Kubli. “ChatGPT outperforms crowd workers for text-annotation tasks.” <\/span><\/span>Proceedings of the National Academy of Sciences<\/span><\/span><\/em> 120.30 \uff082023\uff09: e2305016120.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u9996\u6b21\u7cfb\u7edf\u6027\u5730\u8bc1\u660eChatGPT\u5728\u6587\u672c\u6807\u6ce8\u4efb\u52a1\u4e0a\u4f18\u4e8e\u4eba\u5de5\u4f17\u5305\uff0c\u4e3aAI\u8f85\u52a9\u6570\u636e\u6807\u6ce8\u63d0\u4f9b\u4e86\u5b9e\u8bc1\u7814\u7a76\u652f\u6301\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [3] Abay\uff0cNazmiye Ceren\uff0cet al. “Privacy preserving synthetic data release using deep learning.” <\/span><\/span>Machine Learning and Knowledge Discovery in Databases: European Conference\uff0cECML PKDD 2018\uff0cDublin\uff0cIreland\uff0cSeptember 10\u201314\uff0c2018\uff0cProceedings\uff0cPart I 18<\/span><\/span><\/em>. Springer International Publishing\uff0c2019.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u63d0\u51fa\u4e86\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u9690\u79c1\u4fdd\u62a4\u5408\u6210\u6570\u636e\u751f\u6210\u65b9\u6cd5\uff0c\u5728\u6570\u636e\u9690\u79c1\u548c\u6548\u7528\u4e4b\u95f4\u53d6\u5f97\u4e86\u826f\u597d\u5e73\u8861\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [4] Mandlekar\uff0cAjay\uff0cet al. “Mimicgen: A data generation system for scalable robot learning using human demonstrations.” <\/span><\/span>arXiv preprint arXiv:2310.17596<\/span><\/span><\/em> \uff082023\uff09.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u521b\u65b0\u6027\u5730\u63d0\u51fa\u4e86\u57fa\u4e8e\u4eba\u7c7b\u793a\u8303\u7684\u673a\u5668\u4eba\u5b66\u4e60\u6570\u636e\u751f\u6210\u7cfb\u7edf\uff0c\u4e3a\u89e3\u51b3\u673a\u5668\u4eba\u5b66\u4e60\u4e2d\u7684\u6570\u636e\u74f6\u9888\u63d0\u4f9b\u4e86\u65b0\u601d\u8def\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [5] Zhang\uff0cYuanhan\uff0cet al. “Video Instruction Tuning With Synthetic Data.” <\/span><\/span>arXiv preprint arXiv:2410.02713<\/span><\/span><\/em> \uff082024\uff09.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u63a2\u7d22\u4e86\u89c6\u9891\u6307\u4ee4\u5fae\u8c03\u7684\u5408\u6210\u6570\u636e\u751f\u6210\u65b9\u6cd5\uff0c\u4e3a\u591a\u6a21\u6001\u5927\u6a21\u578b\u7684\u8bad\u7ec3\u63d0\u4f9b\u4e86\u65b0\u7684\u89c6\u89d2\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [6] Zhao\uff0cChenyang\uff0cet al. “Self-guide: Better task-specific instruction following via self-synthetic finetuning.” <\/span><\/span>arXiv preprint arXiv:2407.12874<\/span><\/span><\/em> \uff082024\uff09.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u63d0\u51fa\u4e86\u81ea\u751f\u6210\u5f0f\u6307\u4ee4\u5fae\u8c03\u65b9\u6cd5\uff0c\u663e\u8457\u63d0\u5347\u4e86\u6a21\u578b\u5bf9\u7279\u5b9a\u4efb\u52a1\u7684\u7406\u89e3\u80fd\u529b\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [7] Chen\uff0cZixiang\uff0cet al. “Self-play fine-tuning converts weak language models to strong language models.” <\/span><\/span>arXiv preprint arXiv:2401.01335<\/span><\/span><\/em> \uff082024\uff09.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u521b\u65b0\u6027\u5730\u63d0\u51fa\u4e86\u81ea\u5bf9\u5f08\u5f0f\u5fae\u8c03\u65b9\u6cd5\uff0c\u8bc1\u660e\u4e86\u5f31\u6a21\u578b\u53ef\u4ee5\u901a\u8fc7\u81ea\u6211\u63d0\u5347\u53d8\u5f3a\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [8] Wu\uff0cTianhao\uff0cet al. “Meta-rewarding language models: Self-improving alignment with llm-as-a-meta-judge.” <\/span><\/span>arXiv preprint arXiv:2407.19594<\/span><\/span><\/em> \uff082024\uff09.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u63d0\u51fa\u4e86\u521b\u65b0\u7684\u5143\u5956\u52b1\u673a\u5236\u6765\u6539\u8fdb\u8bed\u8a00\u6a21\u578b\u5bf9\u9f50\uff0c\u4e3a\u6a21\u578b\u81ea\u6211\u63d0\u5347\u63d0\u4f9b\u4e86\u65b0\u6846\u67b6\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [9] Liang\uff0cYiming\uff0cet al. “I-SHEEP: Self-Alignment of LLM from Scratch through an Iterative Self-Enhancement Paradigm.” <\/span><\/span>arXiv preprint arXiv:2408.08072<\/span><\/span><\/em> \uff082024\uff09.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u63d0\u51fa\u4e86\u4ece\u96f6\u5f00\u59cb\u7684LLM\u81ea\u5bf9\u9f50\u8fed\u4ee3\u589e\u5f3a\u8303\u5f0f\uff0c\u4e3a\u6a21\u578b\u5bf9\u9f50\u63d0\u4f9b\u4e86\u5168\u65b0\u601d\u8def\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [10] Van Breugel\uff0cBoris\uff0cZhaozhi Qian\uff0cand Mihaela Van Der Schaar. “Synthetic data\uff0creal errors: how \uff08not\uff09 to publish and use synthetic data.” <\/span><\/span>International Conference on Machine Learning<\/span><\/span><\/em>. PMLR\uff0c2023.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u6df1\u5165\u5206\u6790\u4e86\u5408\u6210\u6570\u636e\u4f7f\u7528\u4e2d\u7684\u5e38\u89c1\u9519\u8bef\uff0c\u63d0\u4f9b\u4e86\u5b9e\u7528\u7684\u5408\u6210\u6570\u636e\u53d1\u5e03\u548c\u4f7f\u7528\u6307\u5357\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [11] Barbierato\uff0cEnrico\uff0cet al. “A methodology for controlling bias and fairness in synthetic data generation.” <\/span><\/span>Applied Sciences<\/span><\/span><\/em> 12.9 \uff082022\uff09: 4619.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u63d0\u51fa\u4e86\u4e00\u79cd\u5728\u5408\u6210\u6570\u636e\u751f\u6210\u4e2d\u63a7\u5236\u504f\u5dee\u548c\u516c\u5e73\u6027\u7684\u65b9\u6cd5\uff0c\u65e8\u5728\u63d0\u9ad8\u6570\u636e\u7684\u516c\u6b63\u6027\uff0c\u8be5\u65b9\u6cd5\u5728\u51cf\u5c11\u7b97\u6cd5\u504f\u5dee\u548c\u4fc3\u8fdb\u516c\u5e73\u6027\u65b9\u9762\u5177\u6709\u91cd\u8981\u610f\u4e49\uff0c\u7279\u522b\u662f\u5728\u654f\u611f\u5e94\u7528\u9886\u57df\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [12] Mattern\uff0cJustus\uff0cet al. “Membership inference attacks against language models via neighbourhood comparison.” <\/span><\/span>arXiv preprint arXiv:2305.18462<\/span><\/span><\/em> \uff082023\uff09.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u8fd9\u9879\u7814\u7a76\u63d0\u51fa\u4e86\u57fa\u4e8e\u90bb\u57df\u6bd4\u8f83\u7684\u521b\u65b0\u6210\u5458\u63a8\u7406\u653b\u51fb\u65b9\u6cd5\uff0c\u6709\u6548\u63ed\u793a\u4e86\u8bed\u8a00\u6a21\u578b\u5728\u8bad\u7ec3\u6570\u636e\u9690\u79c1\u4fdd\u62a4\u65b9\u9762\u7684\u6f5c\u5728\u6f0f\u6d1e\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [13] Xu Guo&Yiqiang Chen.\uff082024\uff09 “Generative AI for Synthetic Data Generation: Methods\uff0cChallenges and the Future”<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u5168\u9762\u68b3\u7406\u4e86\u751f\u6210\u5f0fAI\u5728\u5408\u6210\u6570\u636e\u751f\u6210\u9886\u57df\u7684\u6700\u65b0\u8fdb\u5c55\u548c\u6280\u672f\u6311\u6218\uff0c\u4e3a\u7814\u7a76\u8005\u63d0\u4f9b\u4e86\u5b8c\u6574\u7684\u9886\u57df\u53d1\u5c55\u5168\u8c8c\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [14] I. J. Goodfellow\uff0c J. Pouget-Abadie\uff0c M. Mirza\uff0c B. Xu\uff0c D. Warde-Farley\uff0c S. Ozair\uff0c A. Courville\uff0c and Y. Bengio\uff0c \u201cGenerative adversarial networks\uff0c\u201d2014<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u8fd9\u662fGAN\u9886\u57df\u7684\u5f00\u5c71\u4e4b\u4f5c\uff0c\u5f00\u521b\u4e86\u751f\u6210\u5bf9\u6297\u7f51\u7edc\u7684\u7814\u7a76\u65b9\u5411\u5e76\u5960\u5b9a\u4e86\u7406\u8bba\u57fa\u7840\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [15] D. Chen\uff0c C. Lee\uff0c Y. Lu\uff0c D. Rosati\uff0c and Z. Yu\uff0c \u201cMixture of soft prompts for controllable data generation\uff0c\u201d in Findings of the Association for Computational Linguistics: EMNLP 2023\uff0c H. Bouamor\uff0c J. Pino\uff0c and K. Bali\uff0c Eds.<\/span><\/span><\/p>\n

    Singapore: Association for Computational Linguistics\uff0c Dec. 2023\uff0c pp. 14 815\u201314 833. [Online]. Available: https:\/\/aclanthology.org\/2023.findings-emnlp.988<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u5728EMNLP\u53d1\u8868\u7684\u7814\u7a76\u521b\u65b0\u6027\u5730\u63d0\u51fa\u4e86\u6df7\u5408\u8f6f\u63d0\u793a\u65b9\u6cd5\u6765\u5b9e\u73b0\u53ef\u63a7\u7684\u6570\u636e\u751f\u6210\uff0c\u4e3a\u63d0\u793a\u5de5\u7a0b\u63d0\u4f9b\u4e86\u65b0\u601d\u8def\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [16] Y. Yu\uff0c Y. Zhuang\uff0c J. Zhang\uff0c Y. Meng\uff0c A. Ratner\uff0c R. Krishna\uff0c J. Shen\uff0c and C. Zhang\uff0c \u201cLarge language model as attributed training data generator: A tale of diversity and bias\uff0c\u201d in Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track\uff0c 2023.<\/span><\/span><\/p>\n

    [Online]. Available: https:\/\/openreview.net\/forum?id=6hZIfAY9GD<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u6df1\u5165\u63a2\u8ba8\u4e86\u5927\u8bed\u8a00\u6a21\u578b\u5728\u8bad\u7ec3\u6570\u636e\u751f\u6210\u8fc7\u7a0b\u4e2d\u7684\u591a\u6837\u6027\u548c\u504f\u89c1\u95ee\u9898\uff0c\u4e3a\u63d0\u5347AI\u7cfb\u7edf\u7684\u516c\u5e73\u6027\u63d0\u4f9b\u4e86\u91cd\u8981\u53c2\u8003\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [17] L. Reynolds and K. McDonell\uff0c \u201cPrompt programming for large language models: Beyond the few-shot paradigm\uff0c\u201d 2021.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u7a81\u7834\u6027\u5730\u63a2\u7d22\u4e86\u8d85\u8d8a\u5c11\u6837\u672c\u8303\u5f0f\u7684\u63d0\u793a\u7f16\u7a0b\u65b9\u6cd5\uff0c\u4e3a\u5927\u8bed\u8a00\u6a21\u578b\u7684\u63d0\u793a\u5de5\u7a0b\u5f00\u8f9f\u4e86\u65b0\u65b9\u5411\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [18] He, F., Aebersold, R., Baker, M.S. <\/span><\/span>et al.<\/span><\/span><\/em> \u03c0-HuB: the proteomic navigator of the human body. <\/span><\/span>Nature<\/span><\/span><\/em>636<\/span><\/span><\/strong>, 322\u2013331 (2024). https:\/\/doi.org\/10.1038\/s41586-024-08280-5<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u8fd9\u7bc7Nature\u6587\u7ae0\u4ecb\u7ecd\u4e86\u7a81\u7834\u6027\u7684\u4eba\u4f53\u86cb\u767d\u8d28\u7ec4\u5bfc\u822a\u5668 \u03c0-HuB\u5e73\u53f0\uff0c\u4e3a\u7406\u89e3\u4eba\u4f53\u751f\u7406\u673a\u5236\u548c\u75be\u75c5\u53d1\u5c55\u63d0\u4f9b\u4e86\u65b0\u89c6\u89d2\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [19] Kelly, R.T. Single-cell proteomics: progress and prospects. Mol. Cell. Proteomics 19, 1739\u20131748 (2020).<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u7cfb\u7edf\u603b\u7ed3\u4e86\u5355\u7ec6\u80de\u86cb\u767d\u8d28\u7ec4\u5b66\u7684\u6280\u672f\u8fdb\u5c55\u548c\u5e94\u7528\u524d\u666f\uff0c\u4e3a\u8be5\u9886\u57df\u7814\u7a76\u8005\u63d0\u4f9b\u4e86\u91cd\u8981\u53c2\u8003\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [20] Eldjarn, G.H. et al. Large-scale plasma proteomics comparisons through genetics and disease associations. Nature 622, 348\u2013358 (2023).<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u63d0 \u5728Nature\u53d1\u8868\u7684\u5927\u89c4\u6a21\u8840\u6d46\u86cb\u767d\u8d28\u7ec4\u5b66\u7814\u7a76\uff0c\u901a\u8fc7\u4e0e\u9057\u4f20\u5b66\u548c\u75be\u75c5\u5173\u8054\u5206\u6790\u63ed\u793a\u4e86\u91cd\u8981\u7684\u751f\u7269\u5b66\u673a\u5236\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [21] Shumailov\uff0cI.\uff0cShumaylov\uff0cZ.\uff0cZhao\uff0cY. <\/span><\/span>et al.<\/span><\/span><\/em> AI models collapse when trained on recursively generated data. <\/span><\/span>Nature<\/span><\/span><\/em> 631\uff0c755\u2013759 \uff082024\uff09. https:\/\/doi.org\/10.1038\/s41586-024-07566-y<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>Nature\u4e0a\u53d1\u8868\u7684\u91cd\u8981\u7814\u7a76\uff0c\u63ed\u793a\u4e86AI\u6a21\u578b\u5728\u9012\u5f52\u751f\u6210\u6570\u636e\u8bad\u7ec3\u65f6\u7684\u5d29\u6e83\u73b0\u8c61\uff0c\u5bf9\u5f53\u524d\u751f\u6210\u5f0fAI\u7684\u5c40\u9650\u6027\u63d0\u4f9b\u4e86\u5173\u952e\u6d1e\u5bdf\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [22] Y. Zhu\uff0c R. Kiros\uff0c R. Zemel\uff0c R. Salakhutdinov\uff0c R. Urtasun\uff0c A. Torralba\uff0c and S. Fidler\uff0c \u201cAligning books and movies: Towards story-like visual explanations by watching movies and reading books\uff0c\u201d in Proceedings of the IEEE international conference on computer vision\uff0c 2015\uff0c pp. 19\u201327<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u5f00\u521b\u6027\u5730\u63a2\u7d22\u4e86\u7535\u5f71\u548c\u56fe\u4e66\u5185\u5bb9\u7684\u8de8\u6a21\u6001\u5bf9\u9f50\u95ee\u9898\uff0c\u4e3a\u89c6\u89c9\u53d9\u4e8b\u7406\u89e3\u63d0\u4f9b\u4e86\u91cd\u8981\u7684\u7814\u7a76\u57fa\u7840\uff0c\u662f\u591a\u6a21\u6001\u5b66\u4e60\u9886\u57df\u7684\u7ecf\u5178\u6587\u732e\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [23] C. Peng\uff0c X. Yang\uff0c A. Chen\uff0c K. E. Smith\uff0c N. PourNejatian\uff0c A. B.<\/span><\/span><\/p>\n

    Costa\uff0c C. Martin\uff0c M. G. Flores\uff0c Y. Zhang\uff0c T. Magoc et al.\uff0c \u201cA study of generative large language model for medical research and healthcare\uff0c\u201darXiv preprint arXiv:2305.13523\uff0c 2023.<\/span><\/span><\/p>\n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong>\u7cfb\u7edf\u7814\u7a76\u4e86\u5927\u8bed\u8a00\u6a21\u578b\u5728\u533b\u7597\u7814\u7a76\u548c\u5065\u5eb7\u62a4\u7406\u9886\u57df\u7684\u5e94\u7528\u524d\u666f\uff0c\u6df1\u5165\u5206\u6790\u4e86\u5176\u6f5c\u529b\u548c\u5c40\u9650\u6027\u3002<\/span><\/span><\/p>\n

    [24] S. Moore\uff0c R. Tong\uff0c A. Singh\uff0c Z. Liu\uff0c X. Hu\uff0c Y. Lu\uff0c J. Liang\uff0c C. Cao\uff0c H. Khosravi\uff0c P. Denny et al.\uff0c \u201cEmpowering education with llms-the next-gen interface and content generation\uff0c\u201d in International Conference on Artificial Intelligence in Education. Springer\uff0c 2023\uff0c pp. 32\u201337.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u63a2\u8ba8\u4e86\u5927\u8bed\u8a00\u6a21\u578b\u5728\u6559\u80b2\u9886\u57df\u7684\u521b\u65b0\u5e94\u7528\uff0c\u7279\u522b\u5173\u6ce8\u4e86\u63a5\u53e3\u8bbe\u8ba1\u548c\u5185\u5bb9\u751f\u6210\u65b9\u9762\u7684\u7a81\u7834\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [25] N. Rane\uff0c \u201cRole and challenges of chatgpt and similar generative artificial intelligence in business management\uff0c\u201d Available at SSRN 4603227\uff0c 2023.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>Rane (2023) \u8be6\u7ec6\u5206\u6790\u4e86ChatGPT\u7b49\u751f\u6210\u5f0fAI\u5728\u5546\u4e1a\u7ba1\u7406\u4e2d\u7684\u89d2\u8272\u5b9a\u4f4d\u548c\u9762\u4e34\u7684\u6311\u6218\uff0c\u4e3a\u4f01\u4e1a\u5e94\u7528\u63d0\u4f9b\u4e86\u5b9e\u7528\u7684\u53c2\u8003\u6846\u67b6\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [26] Y. Cao\uff0c S. Li\uff0c Y. Liu\uff0c Z. Yan\uff0c Y. Dai\uff0c P. S. Yu\uff0c and L. Sun\uff0c \u201cA comprehensive survey of ai-generated content (aigc): A history of generative ai from gan to chatgpt\uff0c\u201d arXiv preprint arXiv:2303.04226\uff0c 2023.<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u5168\u9762\u56de\u987e\u4e86AI\u751f\u6210\u5185\u5bb9(AIGC)\u7684\u53d1\u5c55\u5386\u7a0b\uff0c\u4eceGAN\u5230ChatGPT\u7684\u6280\u672f\u6f14\u8fdb\uff0c\u662f\u7406\u89e3\u751f\u6210\u5f0fAI\u53d1\u5c55\u7684\u91cd\u8981\u7efc\u8ff0\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [27] A. Bauer\uff0c S. Trapp\uff0c M. Stenger\uff0c R. Leppich\uff0c S. Kounev\uff0c M. Leznik\uff0c K. Chard\uff0c and I. Foster\uff0c \u201cComprehensive exploration of synthetic data generation: A survey\uff0c\u201d arXiv preprint arXiv:2401.02524\uff0c 2024<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u5168\u9762\u7efc\u8ff0\u4e86\u5408\u6210\u6570\u636e\u751f\u6210\u7684\u5404\u79cd\u65b9\u6cd5\u548c\u5e94\u7528\uff0c\u4e3a\u7814\u7a76\u8005\u63d0\u4f9b\u4e86\u7cfb\u7edf\u6027\u7684\u53c2\u8003\u6846\u67b6\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    [28] https:\/\/www.anthropic.com\/news\/model-context-protocol<\/span><\/span><\/p>\n\n\n\n
    \n

    \u63a8\u8350\u7406\u7531\uff1a<\/span><\/span><\/strong><\/span>\u901a\u8fc7MCP\u534f\u8bae\u4e3a\u5927\u8bed\u8a00\u6a21\u578b\u5728\u957f\u6587\u672c\u5904\u7406\u80fd\u529b\u4e0a\u7684\u7a81\u7834\u6027\u8fdb\u5c55\uff0c\u4e3a\u5927\u6a21\u578b\u67b6\u6784\u4f18\u5316\u63d0\u4f9b\u4e86\u65b0\u601d\u8def\u3002<\/span><\/span><\/span><\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n


    <\/span><\/strong><\/p>\n

    \n

    \u51fa\u54c1<\/span><\/span><\/strong>\uff1a\u6f06\u8fdc\u3001\u5434\u529b\u6ce2\u3001\u5f20\u6c5f<\/span><\/span><\/span><\/p>\n

    \u8fd0\u8425<\/span><\/span><\/strong>\uff1a\u5b5f\u664b\u5b87\u3001\u738b\u5a77<\/span><\/span><\/p>\n

    \u64b0\u7a3f<\/span><\/span><\/strong>\uff1a\u5f20\u6c5f\u3001\u6768\u71d5\u9752\u3001\u738b\u5a77\u3001\u738b\u671d\u4f1a\u3001\u5341\u4e09\u7ef4\u3001\u5468\u8389\u3001\u6881\u91d1\u3001\u8881\u51b0\u3001\u6c5f\u5343\u6708\u3001\u5218\u5fd7\u6bc5<\/span><\/span><\/p>\n

    \u9e23\u8c22<\/span><\/span><\/strong>\uff08\u6309\u59d3\u6c0f\u62fc\u97f3\u987a\u5e8f\uff0c\u6392\u540d\u4e0d\u5206\u5148\u540e\uff09\uff1a<\/span><\/span><\/p>\n

    \u66f9\u98ce\u96f7 \u3001\u9648\u5c0f\u6768 \u3001\u7a0b\u8fdc\u3001\u675c\u6c85\u5c82 \u3001\u6bb5\u90c1\u3001\u65b9\u69af\u6977 \u3001\u4ed8\u5f66\u4f1f\u3001 \u9ad8\u60a6\u3001\u9ec4\u67ef\u946b\u3001\u674e\u660a\u3001\u5218\u5723\u8d85\u3001\u8c2d\u4f1f\u654f\u3001\u5434\u6cf0\u9716\u3001\u5434\u8273\u73b2\u3001\u5411\u7ea2\u519b\u3001\u5f20\u9aa5\u3001\u5f20\u8273\u3001\u6731\u601d\u8bed<\/span><\/span><\/p>\n<\/blockquote>\n


    <\/span><\/p>\n


    <\/span><\/p>\n

    <\/section>\n


    <\/span><\/p>\n

    \n
    \n
    \n
    \n

    AI+Science \u8bfb\u4e66\u4f1a<\/span><\/strong><\/p>\n<\/section>\n<\/section>\n<\/section>\n<\/section>\n

    \n


    <\/span><\/span><\/p>\n<\/section>\n

    <\/span><\/section>\n

    \"\u56fe\u7247\"<\/span><\/a><\/p>\n

    <\/span><\/p>\n

    <\/span><\/section>\n
    \u8be6\u60c5\u8bf7\u89c1\uff1a<\/span><\/span><\/span><\/section>\n
    \u4eba\u5de5\u667a\u80fd\u548c\u79d1\u5b66\u53d1\u73b0\u76f8\u4e92\u8d4b\u80fd\u7684\u65b0\u8303\u5f0f\uff1aAI+Science \u8bfb\u4e66\u4f1a\u542f\u52a8<\/a><\/span><\/span><\/span><\/section>\n

    <\/span><\/section>\n

    <\/span><\/section>\n
    \u63a8\u8350\u9605\u8bfb<\/span><\/span><\/strong><\/strong><\/strong><\/section>\n
    <\/strong><\/section>\n
    1. \u300aAI X Science\u5341\u5927\u524d\u6cbf\u89c2\u5bdf\u300b\u53d1\u5e03\uff1a35\u4e2a\u7814\u7a76\u524d\u6cbf\uff0c\u5f00\u542f\u79d1\u5b66\u53d1\u73b0\u7684\u9ec4\u91d1\u65f6\u4ee3<\/span><\/a><\/span><\/strong>
    <\/span><\/strong><\/span><\/section>\n
    2. \u300aAI\u00d7SCIENCE\u5341\u5927\u524d\u6cbf\u89c2\u5bdf\u300b1\uff1a\u57fa\u4e8eLLM\u7684\u79d1\u5b66\u7814\u7a76<\/span><\/a><\/span><\/strong>
    <\/span><\/strong><\/span><\/section>\n
    3. \u300aAI\u00d7SCIENCE\u5341\u5927\u524d\u6cbf\u89c2\u5bdf\u300b2\uff1a\u5782\u76f4\u9886\u57df\u79d1\u5b66\u5927\u6a21\u578b<\/span><\/a><\/span><\/strong><\/section>\n
    4. \u300aAI\u00d7SCIENCE\u5341\u5927\u524d\u6cbf\u89c2\u5bdf\u300b3\uff1a\u878d\u5165\u5148\u9a8c\u77e5\u8bc6\u7684AI\u6a21\u578b<\/span><\/a><\/span><\/strong><\/section>\n
    5. \u300aAI\u00d7SCIENCE\u5341\u5927\u524d\u6cbf\u89c2\u5bdf\u300b4\uff1aAI\u79d1\u5b66\u5bb6<\/span><\/a><\/span><\/strong><\/section>\n
    6. \u300aAI\u00d7SCIENCE\u5341\u5927\u524d\u6cbf\u89c2\u5bdf\u300b5\uff1a\u590d\u6742\u4e16\u754c\u7684\u591a\u667a\u80fd\u4f53\u5efa\u6a21<\/span><\/a><\/span><\/strong><\/span><\/section>\n
    7. \u300aAI\u00d7SCIENCE\u5341\u5927\u524d\u6cbf\u89c2\u5bdf\u300b6\uff1aAI \u4eff\u771f\u4e0e\u7cfb\u7edf\u6a21\u62df<\/span><\/a><\/span><\/strong><\/span><\/section>\n
    8. \u300aAI\u00d7SCIENCE\u5341\u5927\u524d\u6cbf\u89c2\u5bdf\u300b7\uff1a\u7269\u7406\u4e16\u754c\u7684\u7b2c\u4e00\u6027\u539f\u7406<\/span><\/a><\/span>
    <\/span><\/strong><\/span><\/strong><\/span><\/section>\n
    9. \u300aAI\u00d7SCIENCE\u5341\u5927\u524d\u6cbf\u89c2\u5bdf\u300b8\uff1a\u79d1\u5b66\u542f\u53d1\u7684AI\u65b0\u67b6\u6784<\/span><\/a><\/span><\/strong><\/span><\/strong><\/span><\/section>\n
    <\/strong><\/section>\n
    10. <\/span><\/strong><\/span>\u6e38\u620f\u5316\u79d1\u7814\u2014\u2014\u8ba9\u6211\u4eec\u7a81\u7834\u5185\u5377\u3001\u5171\u7814\u6d8c\u73b0<\/a><\/span><\/strong><\/strong><\/span><\/section>\n
    <\/strong>11. <\/span><\/strong>\u63a2\u7d22\u8005\u8ba1\u5212 | \u96c6\u667a\u4ff1\u4e50\u90e82025\u5185\u5bb9\u56e2\u961f\u62db\u52df\uff08\u5168\u804c&\u517c\u804c\uff09<\/a><\/span><\/strong><\/span><\/section>\n

    12. \u52a0\u5165\u96c6\u667a\uff0c\u73a9\u8f6c\u590d\u6742\uff0c\u5171\u521b\u6591\u56fe\uff01\u96c6\u667a\u4ff1\u4e50\u90e8\u7ebf\u4e0b\u5fd7\u613f\u8005\u62db\u52df<\/span><\/a><\/span><\/strong>
    <\/span><\/strong><\/span><\/p>\n


    <\/span><\/p>\n


    <\/span><\/span><\/p>\n

    \u70b9\u51fb\u201c\u9605\u8bfb\u539f\u6587\u201d\uff0c\u62a5\u540d\u8bfb\u4e66\u4f1a<\/span><\/span><\/section>\n<\/p>\n<\/div>\n","protected":false},"excerpt":{"rendered":"

    \u5bfc\u8bed \u4e0a\u6d77\u79d1\u5b66\u667a\u80fd\u7814\u7a76\u9662\u3001\u96c6\u667a\u79d1\u5b66\u7814\u7a76\u4e2d\u5fc3\u548c\u963f\u91cc\u4e91\u8054\u5408\u53d1\u5e03\u4e86\u300aAI \u00d7 Science\u5341\u5927\u524d\u6cbf\u89c2\u5bdf\u300b\uff0c\u68b3\u7406\u51fa35\u4e2a\u7814\u7a76\u524d\u6cbf\uff0c\u6765\u63a8\u52a8\u79d1\u5b66\u53d1\u5c55\u7684\u9ec4\u91d1\u65f6\u4ee3\u5230\u6765\u3002\u672c\u7bc7\u4e3a\u524d\u6cbf\u89c2\u5bdf9\uff0c\u626b\u63cf\u4e0b\u65b9\u4e8c\u7ef4\u7801\uff0c\u53ef\u83b7\u5f97\u5b8c\u6574\u7248\u4e0b\u8f7d\u3001\u5feb\u901f\u94fe\u63a5\u8bba\u6587\u539f\u6587\u3002 \u626b\u63cf\u4e8c\u7ef4\u7801 \u4e0b\u8f7d\u5b8c\u6574\u62a5\u544a \u5408\u6210\u6570\u636e\u548c\u6570\u636e\u57fa\u7840\u8bbe\u65bd \u80cc\u666f\u4ecb\u7ecd \u5728AI\u5927\u6a21\u578b\u65f6…<\/p>\n","protected":false},"author":1,"featured_media":59126,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[1],"tags":[],"special":[],"_links":{"self":[{"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/posts\/59139"}],"collection":[{"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/swarma.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=59139"}],"version-history":[{"count":0,"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/posts\/59139\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/swarma.org\/index.php?rest_route=\/wp\/v2\/media\/59126"}],"wp:attachment":[{"href":"https:\/\/swarma.org\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=59139"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/swarma.org\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=59139"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/swarma.org\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=59139"},{"taxonomy":"special","embeddable":true,"href":"https:\/\/swarma.org\/index.php?rest_route=%2Fwp%2Fv2%2Fspecial&post=59139"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}