{"id":38299,"date":"2025-02-13T18:00:52","date_gmt":"2025-02-13T10:00:52","guid":{"rendered":"https:\/\/17aitech.com\/?p=38299"},"modified":"2025-02-13T18:00:52","modified_gmt":"2025-02-13T10:00:52","slug":"%e8%a7%86%e8%a7%89-ai-%e7%9a%84%e3%80%8cfoundation-model%e3%80%8d%ef%bc%8c%e5%b7%b2%e7%bb%8f%e5%8f%91%e5%b1%95%e5%88%b0%e5%93%aa%e4%b8%80%e6%ad%a5%ef%bc%9f%e4%b8%a8cvpr-2024-%e7%8e%b0%e5%9c%ba","status":"publish","type":"post","link":"https:\/\/17aitech.com\/?p=38299","title":{"rendered":"\u89c6\u89c9 AI \u7684\u300cFoundation Model\u300d\uff0c\u5df2\u7ecf\u53d1\u5c55\u5230\u54ea\u4e00\u6b65\uff1f\u4e28CVPR 2024 \u73b0\u573a\u76f4\u51fb"},"content":{"rendered":"

<\/p>\n

2024 \u5e74\u7f8e\u56fd\u65f6\u95f4 6 \u6708 17 \u65e5\u81f3 21 \u65e5\uff0cIEEE \u56fd\u9645\u8ba1\u7b97\u673a\u89c6\u89c9\u4e0e\u6a21\u5f0f\u8bc6\u522b\u4f1a\u8bae\uff08CVPR\uff09\u5728\u7f8e\u56fd\u897f\u96c5\u56fe\u53ec\u5f00\u3002<\/p>\n

\u5982\u5927\u5bb6\u9884\u6599\uff0c\u89c6\u89c9 Foundation Model \u6210\u4e3a\u4eca\u5e74 CVPR \u9664\u81ea\u52a8\u9a7e\u9a76\u30013D \u89c6\u89c9\u7b49\u4f20\u7edf\u7814\u7a76\u8bfe\u9898\u4ee5\u5916\u7684\u6838\u5fc3\u4f1a\u8bae\u4e3b\u9898\u3002<\/p>\n

\u6b64\u5916\uff0c\u7531\u4e8e\u4f1a\u8bae\u53ec\u5f00\u524d\u540e\uff0c\u56fd\u5185\u5feb\u624b\u300c\u53ef\u7075\u300d\u5f00\u653e\u56fe\u751f\u89c6\u9891\u529f\u80fd\u706b\u7206\u51fa\u5708\u3001Runway \u65f6\u9694\u4e00\u5e74\u63a8\u51fa\u65b0\u6a21\u578b Gen-3 Alpha\uff0c\u6587\u751f\u89c6\u9891\u4e5f\u6210\u4e3a CVPR 2024 \u7684\u4e00\u5927\u70ed\u8bcd\u3002<\/p>\n

\u4eca\u5e74\uff0c CVPR \u7684\u4e24\u7bc7\u6700\u4f73\u8bba\u6587\u90fd\u9881\u7ed9\u4e86\u00a0AIGC\u3002\u4ece\u8bba\u6587\u63a5\u6536\u6570\u91cf\u7684\u89d2\u5ea6\u770b\uff0c\u56fe\u50cf\u548c\u89c6\u9891\u5408\u6210\u4e0e\u751f\u6210\uff08Image and video synthesis and generation\uff09\u4ee5329 \u7bc7\u8bba\u6587\u6210\u529f\u5360\u636e\u699c\u9996\u3002\u800c\u6587\u751f\u89c6\u9891\u4e5f\u5c5e\u4e8e\u89c6\u89c9 Foundation Model \u7684\u7814\u7a76\u8ba8\u8bba\u8303\u7574\u3002<\/p>\n

\u4e8b\u5b9e\u4e0a\uff0cFoundation Model \u5728\u4eba\u5de5\u667a\u80fd\u9886\u57df\u7684\u6700\u65e9\u51fa\u5708\u5c31\u662f\u5728\u8ba1\u7b97\u673a\u89c6\u89c9\u9886\u57df\u3002<\/p>\n

2021 \u5e74 8 \u6708\uff0c\u65af\u5766\u798f\u5927\u5b66\u767e\u4f4d\u5b66\u8005\u8054\u540d\u53d1\u8868 Foundation Model \u7efc\u8ff0\uff0c\u4f5c\u8005\u961f\u5217\u91cc\u5c31\u6709\u591a\u4f4d\u8ba1\u7b97\u673a\u89c6\u89c9\u9886\u57df\u7684\u7fd8\u695a\uff0c\u5982\u674e\u98de\u98de\u3001Percy Liang \u7b49\u3002\u4f46OpenAI \u51ed\u501f\u4e00\u5df1\u4e4b\u529b\uff0c\u5728\u81ea\u7136\u8bed\u8a00\u5904\u7406\u9886\u57df\u7684 Foundation Model \u4e0a\u7387\u5148\u5f2f\u9053\u8d85\u8f66\uff0c\u901a\u8fc7\u5806\u53c2\u6570\u91cf\u4e0e\u62fc\u7b97\u529b\uff0c\u5c06\u8bed\u8a00\u5927\u6a21\u578b\u505a\u5230\u6781\u81f4\uff0c\u8bed\u8a00\u57fa\u5ea7\u6a21\u578b\u7684\u98ce\u5934\u4e5f\u5728 2023 \u5e74\u4e00\u5ea6\u76d6\u8fc7\u4e86\u89c6\u89c9\u57fa\u5ea7\u6a21\u578b\u3002<\/p>\n

\u7136\u800c\uff0c\u7531\u4e8e Sora \u4e0e\u53ef\u7075\u7b49\u5de5\u4f5c\u7684\u70ab\u4e3d\u6548\u679c\uff0cCV \u9886\u57df\u5185\u5173\u4e8e\u89c6\u89c9\u300cFoundation Model\u300d\u7684\u8bdd\u9898\u53c8\u91cd\u56de\u724c\u684c\u3002<\/p>\n

\u5728 CVPR 2024 \u7684\u5927\u4f1a\u73b0\u573a\uff0cAI \u79d1\u6280\u8bc4\u8bba\u8d70\u8bbf\u4e86\u591a\u4f4d\u4ece\u4e8b\u8fc7\u89c6\u89c9\u57fa\u5ea7\u6a21\u578b\u7684\u7814\u7a76\u8005\uff0c\u8bd5\u56fe\u6c42\u89e3\u5728\u73b0\u9636\u6bb5\u9886\u57df\u5185\u7684\u4e13\u5bb6\u4eba\u58eb\u5982\u4f55\u770b\u5f85\u300cFoundation Model\u300d\u3002<\/p>\n

\u6211\u4eec\u53d1\u73b0\uff1a<\/p>\n

1.\u89c6\u89c9 Foundation Model \u7684\u7814\u7a76\u601d\u8def\u4e5f\u501f\u9274 OpenAI \u7684\u8def\u7ebf\uff0c\u5c06\u4e0b\u4e00\u6b65\u91cd\u8981\u7a81\u7834\u653e\u5728\u5982\u9884\u6d4b\u4e0b\u4e00\u4e2a visual token\u3001\u6269\u5927\u7b97\u529b\u89c4\u6a21\u7b49\u601d\u8def\u4e0a\uff1b<\/p>\n

2.\u4e0d\u6b62\u4e00\u4f4d\u7814\u7a76\u8005\u8ba4\u4e3a\uff0c\u65e0\u8bba\u662f\u8bed\u8a00\u8fd8\u662f\u89c6\u89c9\uff0cFoundation Model \u7684\u6982\u5ff5\u5d1b\u8d77\u540e\uff0cAI \u5df2\u7ecf\u4ece\u4e00\u4e2a\u5f00\u653e\u7684\u7814\u7a76\u95ee\u9898\u53d8\u6210\u4e86\u4e00\u6761\u5b9e\u5b9e\u5728\u5728\u7684\u300c\u5de5\u4e1a\u751f\u4ea7\u7ebf\u300d\uff0c\u7814\u7a76\u5458\u7684\u76ee\u6807\u53ea\u6709\u4e24\u4e2a\u2014\u2014\u300c\u641e\u6570\u636e\u300d\u4e0e\u300c\u641e\u7b97\u529b\u300d\uff1b<\/p>\n

3.\u300c\u591a\u6a21\u6001\u300d\u6210\u4e3a\u89c6\u89c9\u57fa\u7840\u6a21\u578b\u7814\u7a76\u7684\u4e00\u95e8\u663e\u5b66\uff0c\u4f46\u89c6\u89c9\u4e0e\u8bed\u8a00\u4e24\u6d3e\u7684\u5408\u4f5c\u901a\u9053\u4ecd\u672a\u6709\u6548\u5efa\u7acb\u3002<\/p>\n

\u9664\u4e86 Foundation Model\uff0c\u6211\u4eec\u4e5f\u8bbf\u8c08\u4e86\u81ea\u52a8\u9a7e\u9a76\u30013D \u89c6\u89c9\u9886\u57df\u7684\u76f8\u5173\u4eba\u58eb\u3002\u6211\u4eec\u4e5f\u53d1\u73b0\uff0c\u8bf8\u5982 CVPR \u7b49\u4ece\u8bba\u6587\u63a5\u6536\u622a\u6b62\u5230\u4f1a\u8bae\u53ec\u5f00\u65f6\u95f4\u957f\u5ea6\u8de8\u8d8a\u534a\u5e74\u7684\u5b66\u672f\u4f1a\u8bae\uff0c\u6216\u8bb8\u5df2\u4e0d\u518d\u9002\u7528\u4e8e\u8ddf\u8e2a\u5982\u4eca\u53d8\u5316\u4e07\u8c61\u3001\u65e5\u65b0\u6708\u5f02\u7684\u7814\u7a76\u6210\u679c\u3002<\/p>\n

\u5f53 AI \u7814\u7a76\u4e2d\u5de5\u4e1a\u754c\u4e0e\u4ea7\u4e1a\u754c\u7684\u9694\u9602\u8d8a\u6765\u8d8a\u5c0f\u3001\u878d\u5408\u8d8a\u6765\u8d8a\u591a\u65f6\uff0c\u54ea\u6015\u662f\u4e00\u4e2a\u4f20\u7edf\u7684\u5b66\u672f\u4f1a\u8bae\u4e5f\u8981\u6709\u8ddf\u4e0a\u65f6\u4ee3\u6f6e\u6d41\u7684\u610f\u8bc6\u3002<\/p>\n

\u300cFoundation Model\u300d\u7684\u74f6\u9888\u4e0e\u7a81\u7834<\/span><\/strong><\/p>\n

\u4e8b\u5b9e\u4e0a\uff0c\u57fa\u4e8e Transformer \u5f00\u53d1\u901a\u7528\u7684\u89c6\u89c9\u57fa\u7840\u6a21\u578b\u5e76\u4e0d\u662f 2023 \u5e74 ChatGPT \u706b\u8d77\u6765\u540e\u624d\u6709\u7684\u7814\u7a76\u601d\u8def\u3002<\/p>\n

\u56fd\u5916\u4ece\u5fae\u8f6f Swin Tranformer \u5230\u8c37\u6b4c ViT\uff0c\u518d\u5230\u56fd\u5185\u4e0a\u6d77\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4\u7684\u300c\u4e66\u751f\u300d\uff08Intern\uff09\u7cfb\u5217\uff0c\u90fd\u5f88\u65e9\u5f00\u59cb\u4e86\u901a\u7528\u89c6\u89c9\u667a\u80fd\u7684\u63a2\u7d22\u3002\u53ea\u4e0d\u8fc7\u4e0e\u00a0BERT\u00a0\u88ab GPT-3 \u78be\u538b\u7684\u547d\u8fd0\u4e00\u6837\uff0c\u5b83\u4eec\u90fd\u88ab\u540e\u6765\u51fa\u73b0\u7684 Sora \u5149\u8292\u63a9\u76d6\uff1b\u540c\u65f6\uff0c\u7531\u4e8e Sora \u7684\u6280\u672f\u8def\u5f84\u72ec\u8f9f\u8e4a\u5f84\uff0c\u4e5f\u5f00\u59cb\u5b66\u4e60 Sora\u3001\u501f\u9274 Sora\u3002<\/p>\n

\u5728\u4eca\u5e74\u7684 CVPR \u4e0a\uff0c\u4e0a\u6d77\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4\u7684\u901a\u7528\u89c6\u89c9\u56e2\u961f\uff08OpenGVLab\uff09\u5c55\u793a\u4e86\u4ed6\u4eec\u6700\u65b0\u7684\u89c6\u89c9\u591a\u6a21\u6001\u57fa\u7840\u6a21\u578b\u7814\u7a76\u6210\u679c InternVL-1.5\u3002\u8be5\u5de5\u4f5c\u51ed\u501f\u5f3a\u5927\u7684\u89c6\u89c9\u7f16\u7801\u5668 InternViT-6B\u3001\u9ad8\u52a8\u6001\u5206\u8fa8\u7387\uff0c\u4ee5\u53ca\u9ad8\u8d28\u91cf\u53cc\u8bed\u6570\u636e\u96c6\uff0c\u5728\u4e1a\u5185\u5e7f\u53d7\u8ffd\u6367\u3002CMU\u3001\u9762\u58c1\u667a\u80fd\u7b49\u56fd\u5185\u5916\u7684\u591a\u4e2a\u5f00\u6e90\u5bf9\u6bd4\u6570\u636e\u4e5f\u8868\u793a\uff0cInternVL \u7684\u6548\u679c\u540d\u5217\u524d\u8305\uff1a<\/p>\n

<\/a><\/p>\n

\u4e0a\u6d77\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4OpenGVLab\u300c\u4e66\u751f\u300d\u591a\u6a21\u6001\u5927\u6a21\u578b\u56e2\u961f\u8ba4\u4e3a\uff0c\u89c6\u89c9\u57fa\u7840\u6a21\u578b\u533a\u522b\u4e8e\u4ee5\u5f80\u5de5\u4f5c\u7684\u4e00\u4e2a\u76f4\u63a5\u4f53\u73b0\u662f\u591a\u6a21\u6001\u5bf9\u8bdd\u7cfb\u7edf\u7684\u6784\u5efa\u3002<\/p>\n

InternVL-26B \u7684\u7814\u7a76\u59cb\u4e8e 2023 \u5e74 3 \u6708\u3002\u6b64\u524d\uff0c\u89c6\u89c9\u57fa\u7840\u6a21\u578b\u7684\u76f8\u5173\u7814\u7a76\u4ee3\u8868\u5de5\u4f5c\u662f OpenAI \u5728 2021 \u5e74\u53d1\u8868\u7684CLIP\u3002\u300cCLIP \u4f5c\u4e3a\u53e4\u65e9\u7684\u89c6\u89c9\u57fa\u7840\u6a21\u578b\uff0c\u901a\u8fc7\u4e0e BERT \u5bf9\u9f50\uff0c\u4f7f ViT \u83b7\u5f97\u4e00\u5b9a\u7a0b\u5ea6\u7684\u8bed\u8a00\u8868\u5f81\u5bf9\u9f50\u80fd\u529b\uff0c\u4f46\u53c2\u6570\u91cf\u53ea\u6709 300 M \u5de6\u53f3\uff0c\u89c4\u6a21\u592a\u5c0f\uff0c\u4e14\u5bf9\u9f50\u7684\u5bf9\u8c61\u4e0d\u662f LLM\u3002\uff08\u591a\u6a21\u6001\u5bf9\u8bdd\u7cfb\u7edf\u7684\u6784\u9020\uff09\u5fc5\u987b\u4f7f\u7528\u66f4\u591a\u8bad\u7ec3\u6570\u636e\u624d\u80fd\u8fdb\u884c\u8868\u5f81\u5bf9\u9f50\u3002\u300d<\/p>\n

\u8fd9\u542f\u53d1\u4e86\u4e0a\u6d77 AI Lab \u56e2\u961f\u5f00\u59cb\u7814\u7a76 InternVL\u3002\u4ed6\u4eec\u7684\u76ee\u6807\u662f\u5728\u4fdd\u6301\u57fa\u7840\u6a21\u578b\u5f3a\u5927\u6027\u548c\u591a\u529f\u80fd\u6027\u7684\u524d\u63d0\u4e0b\uff0c\u5c06\u5176\u4f5c\u4e3a\u5bf9\u8bdd\u7cfb\u7edf\u7684 backbone\uff0c\u65e2\u652f\u6301\u56fe\u50cf\u68c0\u6d4b\u3001\u5206\u5272\uff0c\u4e5f\u80fd\u591f\u50cf CLIP \u652f\u6301\u591a\u6a21\u6001\u4efb\u52a1\uff0c\u4f8b\u5982\u56fe\u6587\u68c0\u7d22\u3002\u6700\u5f00\u59cb\u662f\u7814\u7a76\u4e86\u4e00\u4e2a 6+7 \u7684 13B \u6a21\u578b\uff08\u5373 InternVL-Chat-V1.2\uff09\uff0c\u4f46\u7531\u4e8e\u5728\u5bf9\u8bdd\u7cfb\u7edf\u7684\u5b9e\u9645\u5e94\u7528\u4e2d\u8868\u73b0\u4e00\u822c\uff0c\u53c8\u6295\u5165\u5927\u91cf\u7cbe\u529b\u4f18\u5316\u5bf9\u8bdd\u529f\u80fd\uff0c\u53c8\u5f97\u51fa\u4e86\u4e00\u4e2a 26B \u6a21\u578b\uff0c\u5373\u98ce\u9761\u4e00\u65f6\u7684 InternVL-Chat-V1.5\u3002<\/p>\n

\u4ece InternVL-1.5 \u6280\u672f\u62a5\u544a\u5f97\u77e5\uff0c\u89c6\u89c9\u57fa\u7840\u6a21\u578b\u7814\u7a76\u7684\u4e09\u4e2a\u5173\u952e\u70b9\u662f\uff1a<\/p>\n

\u4e00\uff0c\u89c6\u89c9\u6a21\u578b\u5fc5\u987b\u63a5\u9a73\u80fd\u529b\u4e0e\u4e4b\u76f8\u5ab2\u7f8e\u7684\u8bed\u8a00\u6a21\u578b\u3002\u6bd4\u5982\uff0c\u4ed6\u4eec\u4e00\u5f00\u59cb\u7684 7B \u8bed\u8a00\u6a21\u578b\u65e0\u6cd5\u5145\u5206\u53d1\u6325 6B \u89c6\u89c9\u6a21\u578b\u7684\u4f18\u52bf\uff0c\u4f46\u5728\u4ed6\u4eec\u5c06\u8bed\u8a00\u6a21\u578b\u7684\u89c4\u6a21\u6269\u5927\u5230 20B \u540e\uff0c\u95ee\u9898\u5f97\u5230\u4e86\u5927\u5e45\u6539\u5584\u3002InternVL-Chat-V1.5 \u91c7\u7528\u7684\u662f\u4e66\u751f\u00b7\u6d66\u8bed\u7684 20B \u6a21\u578b\uff0c\u4f7f\u6a21\u578b\u5177\u5907\u4e86\u5f3a\u5927\u7684\u4e2d\u6587\u8bc6\u522b\u80fd\u529b\uff1b\u4e8c\u662f\u8981\u9002\u914d\u9ad8\u5206\u8fa8\u7387\uff1b\u4e09\u662f\u8981\u91c7\u7528\u9ad8\u8d28\u91cf\u6570\u636e\u96c6\u3002<\/p>\n

\u5728\u4eca\u5e74\u7684 CVPR \u4e0a\uff0cGPT-4o \u56e2\u961f\u4f5c\u8005\u9996\u6b21\u516c\u5f00\u5206\u4eab\u4e86\u80cc\u540e\u7684\u6280\u672f\u8def\u7ebf\uff1aGPT-4o \u7684\u6587\u5b57\u8f6c\u56fe\u50cf\u91c7\u7528\u4e86 DALL\u00b7E \u8def\u7ebf\uff0c\u6587\u5b57\u8f6c\u6587\u5b57\u662f GPT\uff0c\u6587\u5b57\u8f6c\u8bed\u97f3\u662f TTS\u3002InternVL \u7814\u7a76\u5458\u8bc4\u4ef7\uff0cGPT-4o \u6ce8\u91cd\u4e0d\u540c\u6a21\u578b\u95f4\u7684\u8de8\u6a21\u6001\u8f6c\u5316\uff0c\u4f46 InternVL \u7684\u8def\u7ebf\u662f\u4e13\u6ce8\u4e8e\u540c\u4e00\u4e2a\u6a21\u578b\u4e0a\u4e0d\u540c\u6a21\u6001\u7684\u8f93\u5165\u4e0e\u6587\u672c\u7406\u89e3\u7684\u8f93\u51fa\u3002OpenAI \u8def\u7ebf\u5e76\u4e0d\u662f\u6240\u6709\u89c6\u89c9Foundation Model \u7814\u7a76\u7684\u6743\u5a01\u8def\u7ebf\u3002<\/p>\n

\u76ee\u524d\u9886\u57df\u5185\u6709\u4e00\u79cd\u58f0\u97f3\u8ba4\u4e3a\uff0c\u89c6\u89c9\u57fa\u7840\u6a21\u578b\u5e94\u5177\u5907\u66f4\u5f3a\u7684\u79bb\u6563\u5316\u7279\u6027\uff0c\u5373\u5404\u4e2a\u6a21\u6001\uff08\u5305\u62ec\u89c6\u89c9\u3001\u8bed\u97f3\u548c3D\u8f93\u5165\uff09\u90fd\u8f6c\u6362\u4e3a\u79bb\u6563\u8868\u793a\u3001\u800c\u975e\u9ad8\u7ef4\u5411\u91cf\uff0c\u5e76\u5c06\u5176\u5b58\u50a8\u5728\u540c\u4e00\u6846\u67b6\u4e0b\uff0c\u89e3\u8026\u5bf9\u5916\u611f\u77e5\u4fa7\u6a21\u578b\u548c LLM \u5927\u8111\u6a21\u578b\uff0c\u5982\u6b64\u4e00\u6765\uff0c\u591a\u6a21\u6001\u66f4\u52a0\u7edf\u4e00\uff0c\u8bad\u7ec3\u66f4\u52a0\u72ec\u7acb\uff0c\u4e0d\u7528\u518d\u5173\u6ce8\u89c6\u89c9\u6a21\u578b\u662f\u5426\u4f20\u68af\u5ea6\u3002<\/p>\n

\u5bf9\u6b64\uff0c\u7814\u7a76\u5458\u8ba4\u4e3a\uff0c\u300c\u8fd9\u662f\u5bf9\u539f\u751f\u591a\u6a21\u6001\u652f\u6301\u7684\u4e00\u79cd\u5c1d\u8bd5\uff0c\u4fbf\u4e8e\u8fdb\u884c\u7aef\u5230\u7aef\u7684\u8bad\u7ec3\u548c\u8de8\u6a21\u6001\u80fd\u529b\u7684\u652f\u6301\u3002\u79bb\u6563\u538b\u7f29\u53ef\u80fd\u4f1a\u635f\u5931\u4e00\u4e9b\u7ec6\u5fae\u4f46\u5173\u952e\u7684\u4fe1\u606f\uff0c\u6b64\u6280\u672f\u8def\u7ebf\u8fd8\u6709\u5f88\u591a\u5173\u952e\u95ee\u9898\u6709\u5f85\u63a2\u7d22\u3002\u300d<\/p>\n

\u9488\u5bf9\u89c6\u89c9\u57fa\u7840\u6a21\u578b\u7684\u74f6\u9888\u4e0e\u7a81\u7834\u65b9\u5411\uff0c\u601d\u8c0b\u79d1\u6280\u7814\u7a76\u5458\u3001\u9999\u6e2f\u4e2d\u6587\u5927\u5b66 DV Lab \u5b9e\u9a8c\u5ba4\u6210\u5458\u5f20\u5cb3\u6668\u4e5f\u63d0\u51fa\u4e86\u76f8\u4f3c\u7684\u770b\u6cd5\u3002<\/p>\n

\u4ed6\u8ba4\u4e3a\uff0c\u76ee\u524d\u89c6\u89c9\u57fa\u7840\u6a21\u578b\u7684\u96be\u70b9\u4e3b\u8981\u5728\u4e8e\u5927\u89c4\u6a21\u9ad8\u8d28\u91cf\u6570\u636e\u5982\u4f55\u6536\u96c6\u548c\u52a9\u529b\u5927\u89c4\u6a21\u7684\u8bad\u7ec3\u3002\u4e0d\u4ec5\u5982\u6b64\uff0c\u89c6\u89c9\u57fa\u7840\u6a21\u578b\u5982\u4f55\u8df3\u51fa\u6a21\u578b\u8f93\u51fa\u8bed\u8a00\u7684\u9650\u5236\uff0c\u652f\u6301\u539f\u751f\u591a\u6a21\u6001\uff08\u5982GPT-4o\uff09\u4e5f\u662f\u63a5\u4e0b\u6765\u503c\u5f97\u601d\u8003\u4e0e\u7814\u7a76\u7684\u95ee\u9898\u3002<\/p>\n

\u636e\u96f7\u5cf0\u7f51\u4e86\u89e3\uff0c\u76ee\u524d DV Lab \u81ea\u7814\u7684\u89c6\u89c9\u57fa\u5ea7 Mini-Gemini \u5728\u5f00\u6e90\u793e\u533a\u5f15\u8d77\u4e86\u5e7f\u6cdb\u7684\u5173\u6ce8\u548c\u53cd\u54cd\uff0c\u4e00\u5ea6\u4fdd\u6301 SOTA \u7684\u4f4d\u7f6e\uff0c\u83b7\u5f97\u4e86 3k+ \u7684 stars\u3002\u5728\u4eca\u5e74\u7684 CVPR \u4e0a\uff0c\u8d3e\u4f73\u4e9a DV Lab \u56e2\u961f\u7684 LISA \u6a21\u578b\u3001Video-P2P \u7b49\u5de5\u4f5c\u4e5f\u83b7\u5f97\u4e86\u9ad8\u5ea6\u8bc4\u4ef7\u3002<\/p>\n

\u800c\u5357\u6d0b\u7406\u5de5\u5927\u5b66\u526f\u6559\u6388\u5f20\u542b\u671b\u5219\u8ba4\u4e3a\uff0c\u5728\u89c6\u89c9\u57fa\u7840\u6a21\u578b\u7684\u7814\u7a76\u4e2d\uff0c\u5927\u5bb6\u7ecf\u5e38\u5ffd\u89c6\u201c\u7406\u89e3\u4efb\u52a1\u201d\u548c\u201c\u751f\u6210\u201d\u4efb\u52a1\u672c\u8d28\u662f\u4e92\u65a5\u7684\u95ee\u9898\uff1a\u524d\u8005\u662f\u8981\u8ba9\u5927\u6a21\u578b\u4e22\u6389\u89c6\u89c9\u4fe1\u606f\uff0c\u800c\u540e\u8005\u662f\u8ba9\u5927\u6a21\u578b\u5c3d\u53ef\u80fd\u4fdd\u7559\u89c6\u89c9\u4fe1\u606f\u3002\u7136\u800c\uff0c\u5728\u8bed\u8a00\u5927\u6a21\u578b\u5f53\u4e2d\uff0c\u8fd9\u79cd\u4e92\u65a5\u73b0\u8c61\u786e\u4ece\u6765\u6ca1\u5b58\u5728\u8fc7\u3002<\/p>\n

\u5f20\u6559\u6388\u8ba4\u4e3a\uff0c\u75c7\u7ed3\u5c31\u5728\u4e8e\u76ee\u524dvisual token \u53ea\u662f\u7b80\u5355\u5730\u628a\u89c6\u89c9\u4fe1\u53f7\u201c\u5206\u5757\u201d\uff0c\u8fd9\u79cd\u5757\u72b6\u7684\u7a7a\u95f4\u5e8f\u5217\u548c\u8bed\u8a00\u7684\u201c\u9012\u5f52\u7ed3\u6784\u201d\u662f\u6709\u672c\u8d28\u533a\u522b\u3002\u300c\u5982\u679c\u4e0d\u628a\u56fe\u7247\u6216\u662f\u89c6\u9891\u53d8\u6210\u9012\u5f52\u5e8f\u5217 token \u7684\u8bdd\uff0c\u662f\u65e0\u6cd5\u63a5\u5165\u5927\u8bed\u8a00\u6a21\u578b\u7684\uff0c\u800c\u5927\u8bed\u8a00\u6a21\u578b\u662f\u4e00\u4e2a\u5f88\u91cd\u8981\u7684\u63a8\u7406\u673a\u5668\u3002\u4f46\u76ee\u524d\u8fd9\u4e00\u5757\uff0c\u4ece\u884c\u4e1a\u6765\u770b\uff0c\u8fd8\u6ca1\u6709\u7279\u522b\u597d\u7684\u7814\u7a76\u6210\u679c\u51fa\u73b0\uff0c\u672a\u6765\u503c\u5f97\u52a0\u5927\u6295\u5165\u7814\u7a76\u529b\u5ea6\u3002\u300d<\/p>\n

<\/a>\u56fe\u6ce8\uff1a\u300c\u7406\u89e3\u300d\u4e0e\u300c\u751f\u6210\u300d\u7684\u533a\u522b\uff0c\u6765\u81ea\u5357\u6d0b\u7406\u5de5\u5927\u5b66\u5f20\u542b\u671b\u6559\u6388\u7684\u5206\u4eab<\/span><\/p>\n

\u6b64\u5916\uff0c\u4e0d\u6b62\u4e00\u4f4d\u7814\u7a76\u8005\u8ba4\u4e3a\uff0c\u65e0\u8bba\u662f\u8bed\u8a00\u8fd8\u662f\u89c6\u89c9\uff0cFoundation Model \u7684\u6982\u5ff5\u5d1b\u8d77\u540e\uff0cAI \u5df2\u7ecf\u4ece\u4e00\u4e2a\u5f00\u653e\u7684\u7814\u7a76\u95ee\u9898\u53d8\u6210\u4e86\u4e00\u6761\u5b9e\u5b9e\u5728\u5728\u7684\u300c\u5de5\u4e1a\u751f\u4ea7\u7ebf\u300d\uff0c\u7814\u7a76\u5458\u7684\u76ee\u6807\u53ea\u6709\u4e24\u4e2a\u2014\u2014\u300c\u641e\u6570\u636e\u300d\u4e0e\u300c\u641e\u7b97\u529b\u300d\u3002<\/p>\n

\u63a5\u8fd1 OpenAI \u7684\u77e5\u60c5\u4eba\u58eb\u4e5f\u79f0\uff0c\u4e00\u5f00\u59cb\u51b2\u7740OpenAI \u7684 AGI \u5149\u73af\u52a0\u5165\u7684\u9876\u7ea7\u9ad8\u6821\u535a\u58eb\u6bd5\u4e1a\u751f\u5728\u52a0\u5165\u7814\u7a76\u540e\uff0c\u4e5f\u53d1\u751f\u81ea\u5df1\u5728\u5b9e\u9645\u7814\u7a76\u4e2d\u4e5f\u66f4\u591a\u626e\u6f14\u7740\u87ba\u4e1d\u9489\u822c\u7684\u89d2\u8272\uff0c\u6bd4\u5982\u82b1\u5927\u91cf\u7684\u65f6\u95f4\u5904\u7406\u6570\u636e\u3002\u4e00\u53e5\u9010\u6e10\u5728 OpenAI \u5185\u90e8\u6210\u4e3a\u7ecf\u5178\u7684 Slogan \u662f\uff1a<\/p>\n

There is no magic\u3002<\/p>\n

\u81ea\u52a8\u9a7e\u9a76\u3001\u7aef\u4fa7 AI<\/span><\/strong><\/p>\n

\u81ea\u52a8\u9a7e\u9a76\u5728\u4eca\u5e74\u7684 CVPR \u4e0a\u5360\u636e\u4e86\u975e\u5e38\u91cd\u8981\u7684\u4f4d\u7f6e\uff0c\u5c06\u8bed\u8a00\u5927\u6a21\u578b\u843d\u5730\u5230\u81ea\u52a8\u9a7e\u9a76\u662f\u7279\u8272\u3002<\/p>\n

\u5176\u4e2d\uff0c\u6838\u5fc3\u5c31\u5728\u4e8e\u5982\u4f55\u628a\u5927\u6a21\u578b\u653e\u5230\u81ea\u52a8\u9a7e\u9a76\u7684\u573a\u666f\u4e2d\uff0c\u56e0\u4e3a\u9a7e\u9a76\u9700\u8981\u7406\u89e3\u73af\u5883\u3001\u9884\u6d4b\u4e0b\u4e00\u4e2a\u65f6\u523b\u8be5\u5982\u4f55\u524d\u884c\uff0c\u9047\u5230\u8fb9\u7f18\u573a\u666f(corner case\uff09\u7684\u65f6\u5019\u80fd\u5426\u786e\u4fdd\u5b89\u5168\u6027\u7b49\uff0c\u8fd9\u4e9b\u90fd\u662f\u81ea\u52a8\u9a7e\u9a76\u9886\u57df\u63a5\u4e0b\u6765\u8981\u91cd\u70b9\u7814\u7a76\u7684\u65b9\u5411\u3002<\/p>\n

\u4eca\u5e74\u81ea\u52a8\u9a7e\u9a76\u7684\u4e00\u4e2a\u63a2\u7d22\u8d8b\u52bf\u5c31\u662f\uff0c\u5927\u8bed\u8a00\u6a21\u578b\u4e3a\u81ea\u52a8\u9a7e\u9a76\u7aef\u5230\u7aef\u6280\u672f\u7684\u7b97\u6cd5\u548cinfra\u63d0\u4f9b\u4e86\u65b0\u7684\u601d\u8def\u548c\u89e3\u51b3\u65b9\u6848\u3002\u4ee5\u4eff\u771f\u5e73\u53f0\u4e3a\u4f8b\uff0c\u4e4b\u524d\u7684\u4eff\u771f\u5e73\u53f0\uff0c\u591a\u534a\u662f\u4ee5\u8ba1\u7b97\u673a\u56fe\u5f62\u5b66\u7684\u80fd\u529b\u53bb\u505a\u56fa\u5b9a\u5f15\u64ce\uff0c\u4ece\u800c\u751f\u6210\u4eff\u771f\u5e73\u53f0\uff0c\u4eca\u5e74\u5c31\u6709\u591a\u5bb6\u516c\u53f8\u901a\u8fc7\u751f\u6210\u5f0fAI\u7684\u65b9\u5f0f\u53bb\u505a\u4eff\u771f\u5e73\u53f0\u3002<\/p>\n

CVPR 2024 \u81ea\u52a8\u9a7e\u9a76\u56fd\u9645\u6311\u6218\u8d5b\u662f\u4e1a\u754c\u548c\u5b66\u754c\u90fd\u5173\u6ce8\u7684\u91cd\u8981\u8d5b\u4e8b\u3002\u8be5\u6bd4\u8d5b\u7531\u4e0a\u6d77\u4eba\u5de5\u667a\u80fd\u5b9e\u9a8c\u5ba4\u8054\u5408\u6e05\u534e\u5927\u5b66\u3001\u56fe\u5bbe\u6839\u5927\u5b66\u3001\u7f8e\u56e2\u7b49\u56fd\u5185\u5916\u9ad8\u6821\u548c\u79d1\u6280\u4f01\u4e1a\u5171\u540c\u4e3e\u529e\uff0c\u56f4\u7ed5\u5f53\u524d\u81ea\u52a8\u9a7e\u9a76\u9886\u57df\u7684\u524d\u6cbf\u6280\u672f\u3001\u5b9e\u8df5\u843d\u5730\u573a\u666f\u96be\u9898\u7b49\u5171\u8bbe\u7f6e\u4e86 7 \u5927\u8d5b\u9053\uff0c\u5438\u5f15\u4e86\u5168\u7403\u8fd1 500 \u652f\u961f\u4f0d\u53c2\u8d5b\u3002<\/p>\n

\u6311\u6218\u8d5b\u8981\u6c42\u53c2\u8d5b\u8005\u5f00\u53d1\u4e00\u4e2a\u7aef\u5230\u7aef\u7684 AV \u6a21\u578b\uff0c\u4f7f\u7528 nuPlan \u6570\u636e\u96c6\u8fdb\u884c\u8bad\u7ec3\uff0c\u6839\u636e\u4f20\u611f\u5668\u6570\u636e\u751f\u6210\u884c\u9a76\u8f68\u8ff9\u3002\u636e AI \u79d1\u6280\u8bc4\u8bba\u4e86\u89e3\uff0c\u7aef\u5230\u7aef\u81ea\u52a8\u9a7e\u9a76\u662f\u4eca\u5e74 7 \u5927\u8d5b\u9053\u4e2d\u7ade\u4e89\u6700\u4e3a\u6fc0\u70c8\u7684\u8d5b\u9053\u4e4b\u4e00\uff0c\u51a0\u519b\u6765\u81ea\u4e8e\u82f1\u4f1f\u8fbe\u8054\u5408\u590d\u65e6\u5927\u5b66\u7684\u81ea\u52a8\u9a7e\u9a76\u7b97\u6cd5\u53c2\u8d5b\u56e2\u961f\uff0c\u4e9a\u519b\u5219\u662f\u6765\u81ea\u4e2d\u56fd\u7684\u96f6\u4e00\u6c7d\u8f66\u81ea\u52a8\u9a7e\u9a76\u7814\u53d1\u56e2\u961f\u3002<\/p>\n

<\/a><\/p>\n

\u82f1\u4f1f\u8fbe\u7684\u7814\u7a76\u4eba\u5458\u544a\u8bc9\u96f7\u5cf0\u7f51(\u516c\u4f17\u53f7\uff1a\u96f7\u5cf0\u7f51)<\/span>\uff0cL2++ \u7ea7\u522b\u7684\u7aef\u5230\u7aef\u81ea\u52a8\u9a7e\u9a76\uff0c\u5176\u80fd\u529b\u4e3b\u8981\u4f53\u73b0\u4e8e\u4e24\u5927\u677f\u5757\uff0c\u5206\u522b\u662f Planning \u548c Percetion\u3002<\/p>\n

\u5728\u81ea\u52a8\u9a7e\u9a76\u9886\u57df\u4e2d\u975e\u5e38\u91cd\u8981\u7684\u591a\u6a21\u6001\u6570\u636e\u96c6 nuScenes\uff0c\u5176\u4e2d\u6709 93% \u7684\u6570\u636e\u53ea\u662f\u5305\u542b\u76f4\u884c\u5728\u5185\u7684\u7b80\u5355\u9a7e\u9a76\u573a\u666f\uff0c\u5929\u7136\u65e0\u6cd5\u5b9e\u73b0\u5de5\u4e1a\u754c\u4ea7\u54c1\u7ea7\u522b\u7684\u81ea\u52a8\u9a7e\u9a76\u3002\u8fd9\u4e9b\u573a\u666f\u591a\u4e3a\u81ea\u52a8\u8ddf\u8f66\u3001\u81ea\u52a8\u6cca\u8f66\uff0c\u4ee5\u53ca\u9759\u6001\u73af\u5883\u4fe1\u606f\uff0c\u5982\u4ea4\u901a\u6807\u5fd7\u3001\u9053\u8def\u6807\u793a\u7ebf\u3001\u4ea4\u901a\u706f\u4f4d\u7f6e\u7b49\u3002<\/p>\n

Perception \u662f\u81ea\u52a8\u9a7e\u9a76\u7cfb\u7edf\u4e2d\u7684\u611f\u77e5\u90e8\u5206\uff0c\u8d1f\u8d23\u901a\u8fc7\u5404\u79cd\u4f20\u611f\u5668\u6765\u611f\u77e5\u5468\u56f4\u73af\u5883\u7684\u80fd\u529b\u3002\u5b83\u76f8\u5f53\u4e8e\u81ea\u52a8\u9a7e\u9a76\u8f66\u8f86\u7684\u201c\u773c\u775b\u201d\uff0c\u4e3a\u7cfb\u7edf\u63d0\u4f9b\u5173\u4e8e\u9053\u8def\u3001\u8f66\u8f86\u3001\u884c\u4eba\u3001\u969c\u788d\u7269\u7b49\u5143\u7d20\u7684\u4fe1\u606f\u3002\u800c Planning \u6a21\u5757\u76f8\u5f53\u4e8e\u81ea\u52a8\u9a7e\u9a76\u7cfb\u7edf\u4e2d\u7684\u201c\u5927\u8111\u201d\uff0c\u8d1f\u8d23\u51b3\u7b56\u548c\u89c4\u5212\u8f66\u8f86\u7684\u884c\u9a76\u8def\u5f84\u3002\u5b83\u63a5\u6536\u6765\u81ea\u4e0a\u6e38\u6a21\u5757\uff08\u5982\u5730\u56fe\u3001\u5bfc\u822a\u3001\u611f\u77e5\u3001\u9884\u6d4b\uff09\u7684\u4fe1\u606f\uff0c\u5e76\u5728\u5f53\u524d\u5468\u671f\u5185\u8fdb\u884c\u601d\u8003\u5e76\u505a\u51fa\u5224\u65ad\u3002<\/p>\n

<\/a><\/p>\n

\u82f1\u4f1f\u8fbe\u56e2\u961f\u544a\u8bc9\u96f7\u5cf0\u7f51\uff0c\u4ed6\u4eec\u6240\u4f5c\u51fa\u7684\u521b\u65b0\u5728\u4e8e\uff0c\u5728\u8fb9\u7f18\u573a\u666f\u7684\u6570\u636e\u91cf\u4e0d\u8db3\u591f\u7684\u60c5\u51b5\u4e0b\uff0c\u4f7f\u7528\u57fa\u4e8e\u89c4\u5219\u7684\u4e13\u5bb6\uff08rule-based expert\uff09\u4f5c\u4e3a\u6559\u5e08\uff0c\u5c06\u89c4\u5219\u77e5\u8bc6\u84b8\u998f\u7ed9\u795e\u7ecf\u7f51\u7edc\u89c4\u5212\u5668\u3002\u201c\u6211\u4eec\u8ba4\u4e3a\uff0c\u5373\u4fbf\u5728\u6570\u636e\u91cf\u8db3\u591f\u591a\u7684\u60c5\u51b5\u4e0b\uff0c\u8fd9\u4e00\u65b9\u6cd5\u4e5f\u5c06\u4f7f\u5f97\u795e\u7ecf\u7f51\u7edc\u89c4\u5212\u5668\u53d8\u5f97\u66f4\u5177\u6709\u89e3\u91ca\u6027\u3002\u201d<\/p>\n

\u9664\u4e86\u8fd9\u4e9b\u70ed\u95e8\u8bdd\u9898\uff0c\u5728 CVPR \u73b0\u573a\uff0c\u8fd8\u6709\u5f88\u591a\u5382\u5546\u5e26\u6765\u4e86\u4eae\u773c\u7684\u6280\u672f\u4e0e\u4ea7\u54c1\uff0c\u82f9\u679c\u5c31\u662f\u5176\u4e2d\u4e00\u5bb6\u3002<\/p>\n

\u4ece\u53bb\u5e74\u5f00\u59cb\uff0c\u82f9\u679c\u5bf9\u5927\u6a21\u578b\u7684\u6295\u5165\u529b\u5ea6\u8089\u773c\u53ef\u89c1\u5730\u52a0\u5927\uff0c\u5c24\u5176\u662f\u751f\u6210\u5f0f\u4eba\u5de5\u667a\u80fd\uff08GenAI\uff09\u3002\u867d\u7136\u672c\u8eab\u5e76\u4e0d\u662f\u4e00\u5bb6 AI \u80fd\u529b\u7279\u522b\u5f3a\u5927\u7684\u516c\u53f8\uff0c\u4f46\u4e0d\u61c8\u7684\u52aa\u529b\u8ffd\u8d76\u540e\uff0c\u82f9\u679c\u5df2\u7136\u6210\u529f\u4ece\u4e00\u4e2a\u4e09\u6d41\u6c34\u5e73\u7684 AI \u73a9\u5bb6\u6324\u8fdb\u4e86\u4e8c\u6d41\u6c34\u5e73\u6218\u961f\u3002<\/p>\n

\u4eca\u5e74 3 \u6708\uff0c\u82f9\u679c\u6b63\u5f0f\u53d1\u5e03\u591a\u6a21\u6001 LLM \u7cfb\u5217\u6a21\u578b\uff0c\u5e76\u5728\u8bba\u6587\u300aMM1: Methods, Analysis & Insights from Multimodal LLM Pre-training\u300b\u4e2d\uff0c\u901a\u8fc7\u6784\u5efa\u5927\u6a21\u578b MM1\uff0c\u9610\u8ff0\u4e86\u591a\u6a21\u6001\u5927\u6a21\u578b\u9884\u8bad\u7ec3\u7684\u65b9\u6cd5\u3001\u5206\u6790\u548c\u89c1\u89e3\uff0c\u5f15\u8d77\u5927\u4f17\u56f4\u89c2\u3002<\/p>\n

\u6b64\u6b21\u5728 CVPR \u5927\u4f1a\u4e0a\uff0c\u8bba\u6587\u7684\u4f5c\u8005\u4e4b\u4e00 Zhe Gan \u73b0\u8eab\u5927\u4f1a\u8bba\u575b\uff0c\u7cfb\u7edf\u4ecb\u7ecd\u4e86\u82f9\u679c\u5728\u66f4\u597d\u5730\u8fdb\u884c\u591a\u6a21\u6001\u5927\u6a21\u578b\u9884\u8bad\u7ec3\u6240\u505a\u7684\u6700\u65b0\u7814\u7a76\u8fdb\u5c55\u3002\u4ed6\u8868\u793a\uff0c\u5927\u89c4\u6a21\u4e14\u4e0e\u4efb\u52a1\u76f8\u5173\u7684\u6570\u636e\u5bf9\u4e8e\u8bad\u7ec3\u9ad8\u6027\u80fd\u6a21\u578b\u975e\u5e38\u91cd\u8981\uff0c\u56e0\u6b64\uff0c\u7740\u91cd\u5206\u6790\u4e86\u5982\u4f55\u901a\u8fc7\u57fa\u4e8e\u6a21\u578b\u7684\u8fc7\u6ee4\u548c\u66f4\u591a\u6837\u5316\u7684\u6570\u636e\u6e90\uff0c\u6765\u83b7\u5f97\u9ad8\u8d28\u91cf\u7684\u9884\u8bad\u7ec3\u6570\u636e\u3002<\/span><\/p>\n

<\/a><\/p>\n

\u636e Zhe Gan \u4ecb\u7ecd\uff0c\u5728\u5b9e\u9a8c\u4e2d\uff0c\u4ed6\u4eec\u4f7f\u7528 45% \u6709\u5b57\u5e55\u63cf\u8ff0\u56fe\u50cf\u300145% \u4ea4\u9519\u56fe\u50cf\u6587\u672c\u548c 10% \u7684\u7eaf\u6587\u672c\u6570\u636e\u6df7\u5408\uff0c\u4f5c\u4e3a\u9884\u8bad\u7ec3\u7684\u6570\u636e\u6df7\u5408\uff0c\u5e76\u4e3a\u4e86\u8bc4\u4f30\uff0c\u5728\u5404\u79cd\u5b57\u5e55\u548c VQA \u6570\u636e\u96c6\u4f7f\u7528 zero-shot \uff080-shot\uff09\u548c few-shot (4-shot \u548c 8-shot)\u3002<\/p>\n

<\/a><\/p>\n

\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0c\u4ea4\u9519\u6570\u636e\u5bf9\u4e8e few-shot \u548c\u7eaf\u6587\u672c\u6027\u80fd\u81f3\u5173\u91cd\u8981\uff0c\u800c\u5b57\u5e55\u6570\u636e\u63d0\u9ad8\u4e86 zero-shot \u6027\u80fd\uff1b\u7eaf\u6587\u672c\u6570\u636e\u6709\u52a9\u4e8e\u63d0\u9ad8 few-shot \u548c\u7eaf\u6587\u672c\u6027\u80fd\uff1b\u7cbe\u5fc3\u6df7\u5408\u56fe\u50cf\u548c\u6587\u672c\u6570\u636e\u53ef\u4ee5\u5b9e\u73b0\u6700\u4f73\u591a\u6a21\u6001\u6027\u80fd\uff0c\u540c\u65f6\u4fdd\u6301\u5f3a\u5927\u7684\u6587\u672c\u7406\u89e3\u80fd\u529b\uff1b\u5408\u6210\u6570\u636e\u6709\u52a9\u4e8e few-shot \u5b66\u4e60\u3002<\/p>\n

<\/a><\/a><\/a><\/p>\n

\u53e6\u5916\uff0cZhe Gan \u8868\u793a\uff0c\u5bf9\u4e8e\u5f53\u524d\u70ed\u95e8\u7684 MoE \u67b6\u6784\u6765\u8bf4\uff0c\u53ef\u4ee5\u5728\u4fdd\u6301\u6fc0\u6d3b\u53c2\u6570\u4e0d\u53d8\u7684\u60c5\u51b5\u4e0b\u6269\u5927\u6a21\u578b\u603b\u53c2\u6570\u91cf\uff0c\u4ed6\u4eec\u6b63\u5728\u7814\u7a76\u5982\u4f55\u4e3a\u591a\u6a21\u6001\u5927\u6a21\u578b\u8bbe\u8ba1\u66f4\u597d\u7684 MoE \u67b6\u6784\u3002<\/p>\n

<\/a><\/p>\n

\u5199\u5728\u6700\u540e<\/span><\/strong><\/p>\n

\u4eca\u5e74\u7684 CVPR \u662f\u4e00\u573a\u522b\u5f00\u751f\u9762\u7684\u76db\u4f1a\u3002<\/p>\n

\u5f88\u591a\u53c2\u4f1a\u4eba\u5458\u90fd\u5411\u96f7\u5cf0\u7f51\u8868\u793a\uff0c\u76f8\u8f83\u4e8e\u5f80\u5e74\uff0c\u4eca\u5e74 CVPR \u7684 AIGC \u5143\u7d20\u5f02\u5e38\u6d53\u539a\uff0c\u65b0\u6280\u672f\u3001\u65b0\u4ea7\u54c1\u63a5\u8fde\u6d8c\u73b0\uff0c\u8ba9\u4eba\u5370\u8c61\u6df1\u523b\u3002\u4f46\u4e5f\u6709\u4e00\u4e9b\u5b66\u8005\u8ba4\u4e3a\uff0c\u8fd8\u5e94\u8be5\u6709\u66f4\u591a\u66f4\u65b0\u7684\u6280\u672f\u51fa\u73b0\u3002<\/p>\n

\u9999\u6e2f\u4e2d\u6587\u5927\u5b66\u6df1\u5733\uff08CUHKSZ\uff09\u52a9\u7406\u6559\u6388\u97e9\u6653\u5149\u53c2\u52a0\u5b8c\u6b64\u6b21 CVPR \u4e4b\u540e\uff0c\u8ba4\u4e3a CVPR \u7684\u8bba\u6587\u6295\u7a3f\u65f6\u53ef\u4ee5\u8003\u8651\u8bbe\u7f6e\u4e24\u6761\u8f68\u9053\uff0c\u4e00\u4e2a\u662f\u5de5\u7a0b\u8f68\u9053\uff0c\u4ee5\u6548\u679c\u4f5c\u4e3a\u52a8\u673a\u9a71\u52a8\u70b9\uff0c\u4e00\u4e2a\u662f\u7814\u7a76\u8f68\u9053\uff0c\u4e13\u95e8\u4ee5\u597d\u5947\u5fc3\u4e3a\u9a71\u52a8\u3002\u4e24\u6761\u8f68\u9053\u90fd\u5e94\u8be5\u9700\u8981\u6709\u6700\u4f73\u8bba\u6587\u5956\u9879\uff0cSora \u662f\u4ed6\u5fc3\u91cc\u5de5\u7a0b\u8f68\u9053\u7684\u6700\u4f73\u7814\u7a76\uff0c\u800c\u4eca\u5e74\u7684\u300cGenerative Image Dynamics\u300d\u5219\u6ee1\u8db3\u4e86\u4ed6\u5bf9\u6700\u4f73\u7814\u7a76\u8bba\u6587\u7684\u60f3\u8c61\u3002<\/p>\n

\u300c\u4e00\u76f4\u601d\u8003 CV \u7684\u672a\u6765\u662f\u4ec0\u4e48\u666f\u8c61\uff0c\u6597\u80c6\u9884\u6d4b\uff08\u6216\u8005\u662f\u4e00\u79cd\u5e0c\u671b\uff09\u672a\u6765\u5c06 from \u2018virtual\u2019 to \u2018physical\u2019\uff0c\u53ef\u80fd\u4ee5\u5404\u79cd\u4e0d\u540c\u7684\u5f62\u5f0f\u3002\u300d\u97e9\u6653\u5149\u8bf4\u9053\u3002<\/p>\n

<\/p>\n

\u96f7\u5cf0\u7f51\u539f\u521b\u6587\u7ae0\uff0c\u672a\u7ecf\u6388\u6743\u7981\u6b62\u8f6c\u8f7d\u3002\u8be6\u60c5\u89c1\u8f6c\u8f7d\u987b\u77e5<\/a>\u3002<\/p>\n

\n
\n <\/a>
\n <\/a>\n <\/p>\n

\u6587\u7ae0\u6765\u6e90\u4e8e\u4e92\u8054\u7f51:\u89c6\u89c9 AI \u7684\u300cFoundation Model\u300d\uff0c\u5df2\u7ecf\u53d1\u5c55\u5230\u54ea\u4e00\u6b65\uff1f\u4e28CVPR 2024 \u73b0\u573a\u76f4\u51fb<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

2024 \u5e74\u7f8e\u56fd\u65f6\u95f4 6 \u6708 17 \u65e5\u81f3 […]<\/p>\n","protected":false},"author":3,"featured_media":0,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[27],"tags":[73,68,67,66],"class_list":["post-38299","post","type-post","status-publish","format-standard","hentry","category-news","tag-73","tag-68","tag-67","tag-66"],"yoast_head":"\n\u89c6\u89c9 AI \u7684\u300cFoundation Model\u300d\uff0c\u5df2\u7ecf\u53d1\u5c55\u5230\u54ea\u4e00\u6b65\uff1f\u4e28CVPR 2024 \u73b0\u573a\u76f4\u51fb - \u4e00\u8d77AI\u6280\u672f<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/17aitech.com\/?p=38299\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=38299#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=38299\"},\"author\":{\"name\":\"AI\u5c0f\u52a9\u624b\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\"},\"headline\":\"\u89c6\u89c9 AI \u7684\u300cFoundation Model\u300d\uff0c\u5df2\u7ecf\u53d1\u5c55\u5230\u54ea\u4e00\u6b65\uff1f\u4e28CVPR 2024 \u73b0\u573a\u76f4\u51fb\",\"datePublished\":\"2025-02-13T10:00:52+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=38299\"},\"wordCount\":220,\"commentCount\":0,\"image\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=38299#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/frc-49d124dfb97112abd419fd99739fed58.jpeg\",\"keywords\":[\"\u591a\u6a21\u6001\",\"\u5927\u6a21\u578b\",\"\u81ea\u7136\u8bed\u8a00\u5904\u7406\",\"\u8ba1\u7b97\u673a\u89c6\u89c9\"],\"articleSection\":[\"\u884c\u4e1a\u8d44\u8baf\"],\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/17aitech.com\\\/?p=38299#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=38299\",\"url\":\"https:\\\/\\\/17aitech.com\\\/?p=38299\",\"name\":\"\u89c6\u89c9 AI \u7684\u300cFoundation Model\u300d\uff0c\u5df2\u7ecf\u53d1\u5c55\u5230\u54ea\u4e00\u6b65\uff1f\u4e28CVPR 2024 \u73b0\u573a\u76f4\u51fb - \u4e00\u8d77AI\u6280\u672f\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=38299#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=38299#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/frc-49d124dfb97112abd419fd99739fed58.jpeg\",\"datePublished\":\"2025-02-13T10:00:52+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\"},\"breadcrumb\":{\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=38299#breadcrumb\"},\"inLanguage\":\"zh-Hans\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/17aitech.com\\\/?p=38299\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=38299#primaryimage\",\"url\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/frc-49d124dfb97112abd419fd99739fed58.jpeg\",\"contentUrl\":\"https:\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/frc-49d124dfb97112abd419fd99739fed58.jpeg\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/?p=38299#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"\u9996\u9875\",\"item\":\"https:\\\/\\\/17aitech.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"\u89c6\u89c9 AI \u7684\u300cFoundation Model\u300d\uff0c\u5df2\u7ecf\u53d1\u5c55\u5230\u54ea\u4e00\u6b65\uff1f\u4e28CVPR 2024 \u73b0\u573a\u76f4\u51fb\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/#website\",\"url\":\"https:\\\/\\\/17aitech.com\\\/\",\"name\":\"\u4e00\u8d77AI\u6280\u672f\",\"description\":\"\u8ba9AI\u77e5\u8bc6\u89e6\u624b\u53ef\u53ca\",\"alternateName\":\"\u4e00\u8d77AI\u6280\u672f\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/17aitech.com\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"zh-Hans\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/17aitech.com\\\/#\\\/schema\\\/person\\\/60225458499e817ae0af73e67e440b9d\",\"name\":\"AI\u5c0f\u52a9\u624b\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"zh-Hans\",\"@id\":\"\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/04\\\/robot_3.png\",\"url\":\"\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/04\\\/robot_3.png\",\"contentUrl\":\"\\\/\\\/17aitech.com\\\/wp-content\\\/uploads\\\/2024\\\/04\\\/robot_3.png\",\"caption\":\"AI\u5c0f\u52a9\u624b\"},\"description\":\"\u8fd9\u4e2a\u4eba\u5f88\u61d2\uff0c\u4ec0\u4e48\u90fd\u6ca1\u6709\u7559\u4e0b\uff5e\",\"url\":\"https:\\\/\\\/17aitech.com\\\/?page_id=33738&user=3\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"\u89c6\u89c9 AI \u7684\u300cFoundation Model\u300d\uff0c\u5df2\u7ecf\u53d1\u5c55\u5230\u54ea\u4e00\u6b65\uff1f\u4e28CVPR 2024 \u73b0\u573a\u76f4\u51fb - \u4e00\u8d77AI\u6280\u672f","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/17aitech.com\/?p=38299","schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/17aitech.com\/?p=38299#article","isPartOf":{"@id":"https:\/\/17aitech.com\/?p=38299"},"author":{"name":"AI\u5c0f\u52a9\u624b","@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"headline":"\u89c6\u89c9 AI \u7684\u300cFoundation Model\u300d\uff0c\u5df2\u7ecf\u53d1\u5c55\u5230\u54ea\u4e00\u6b65\uff1f\u4e28CVPR 2024 \u73b0\u573a\u76f4\u51fb","datePublished":"2025-02-13T10:00:52+00:00","mainEntityOfPage":{"@id":"https:\/\/17aitech.com\/?p=38299"},"wordCount":220,"commentCount":0,"image":{"@id":"https:\/\/17aitech.com\/?p=38299#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/06\/frc-49d124dfb97112abd419fd99739fed58.jpeg","keywords":["\u591a\u6a21\u6001","\u5927\u6a21\u578b","\u81ea\u7136\u8bed\u8a00\u5904\u7406","\u8ba1\u7b97\u673a\u89c6\u89c9"],"articleSection":["\u884c\u4e1a\u8d44\u8baf"],"inLanguage":"zh-Hans","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/17aitech.com\/?p=38299#respond"]}]},{"@type":"WebPage","@id":"https:\/\/17aitech.com\/?p=38299","url":"https:\/\/17aitech.com\/?p=38299","name":"\u89c6\u89c9 AI \u7684\u300cFoundation Model\u300d\uff0c\u5df2\u7ecf\u53d1\u5c55\u5230\u54ea\u4e00\u6b65\uff1f\u4e28CVPR 2024 \u73b0\u573a\u76f4\u51fb - \u4e00\u8d77AI\u6280\u672f","isPartOf":{"@id":"https:\/\/17aitech.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/17aitech.com\/?p=38299#primaryimage"},"image":{"@id":"https:\/\/17aitech.com\/?p=38299#primaryimage"},"thumbnailUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/06\/frc-49d124dfb97112abd419fd99739fed58.jpeg","datePublished":"2025-02-13T10:00:52+00:00","author":{"@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d"},"breadcrumb":{"@id":"https:\/\/17aitech.com\/?p=38299#breadcrumb"},"inLanguage":"zh-Hans","potentialAction":[{"@type":"ReadAction","target":["https:\/\/17aitech.com\/?p=38299"]}]},{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"https:\/\/17aitech.com\/?p=38299#primaryimage","url":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/06\/frc-49d124dfb97112abd419fd99739fed58.jpeg","contentUrl":"https:\/\/17aitech.com\/wp-content\/uploads\/2024\/06\/frc-49d124dfb97112abd419fd99739fed58.jpeg"},{"@type":"BreadcrumbList","@id":"https:\/\/17aitech.com\/?p=38299#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"\u9996\u9875","item":"https:\/\/17aitech.com\/"},{"@type":"ListItem","position":2,"name":"\u89c6\u89c9 AI \u7684\u300cFoundation Model\u300d\uff0c\u5df2\u7ecf\u53d1\u5c55\u5230\u54ea\u4e00\u6b65\uff1f\u4e28CVPR 2024 \u73b0\u573a\u76f4\u51fb"}]},{"@type":"WebSite","@id":"https:\/\/17aitech.com\/#website","url":"https:\/\/17aitech.com\/","name":"\u4e00\u8d77AI\u6280\u672f","description":"\u8ba9AI\u77e5\u8bc6\u89e6\u624b\u53ef\u53ca","alternateName":"\u4e00\u8d77AI\u6280\u672f","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/17aitech.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"zh-Hans"},{"@type":"Person","@id":"https:\/\/17aitech.com\/#\/schema\/person\/60225458499e817ae0af73e67e440b9d","name":"AI\u5c0f\u52a9\u624b","image":{"@type":"ImageObject","inLanguage":"zh-Hans","@id":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","url":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","contentUrl":"\/\/17aitech.com\/wp-content\/uploads\/2024\/04\/robot_3.png","caption":"AI\u5c0f\u52a9\u624b"},"description":"\u8fd9\u4e2a\u4eba\u5f88\u61d2\uff0c\u4ec0\u4e48\u90fd\u6ca1\u6709\u7559\u4e0b\uff5e","url":"https:\/\/17aitech.com\/?page_id=33738&user=3"}]}},"_links":{"self":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/38299","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=38299"}],"version-history":[{"count":0,"href":"https:\/\/17aitech.com\/index.php?rest_route=\/wp\/v2\/posts\/38299\/revisions"}],"wp:attachment":[{"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=38299"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=38299"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/17aitech.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=38299"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}