.<\/span><\/a><\/p>\n\u4eba\u5de5\u667a\u80fdAIGC\u884c\u4e1a\u63a2\u7d22\u5206\u4eab\uff0c\u5305\u62ec\u76f8\u5173\u6280\u672f\u5206\u4eab\u548c\u8d44\u8baf\u5206\u4eab\uff0c\u4ee5\u53ca\u76f8\u5173\u5546\u52a1\u6d3d\u8c08\u5408\u4f5c\u3002<\/p>\n
<\/p>\n\n\n<\/section>\n<\/section>\n\u591a\u6a21\u6001\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff08<\/span>MM-LLMs\uff09\u8fd1\u671f\u53d6\u5f97\u663e\u8457\u8fdb\u6b65\uff0c\u901a\u8fc7\u6709\u6548\u8bad\u7ec3\u7b56\u7565\uff0c\u4f7f\u6a21\u578b\u80fd\u5904\u7406\u591a\u6a21\u6001\u8f93\u5165\u8f93\u51fa\uff0c\u4fdd\u7559\u5e76\u589e\u5f3a\u8bed\u8a00\u6a21\u578b\u7684\u63a8\u7406\u51b3\u7b56\u80fd\u529b\u3002<\/span>MM-LLMs\u4e0d\u4ec5\u9650\u4e8e\u6587\u672c\uff0c\u8fd8\u80fd\u5904\u7406\u56fe\u50cf\u3001\u89c6\u9891\u548c\u97f3\u9891\u7b49\u5176\u4ed6\u6a21\u6001\u6570\u636e\uff0c\u63a8\u52a8\u4e0b\u6e38\u4efb\u52a1\u6027\u80fd\u63d0\u5347\u3002<\/span><\/span><\/section>\n
<\/span><\/span><\/section>\nMM-LLMs\u6539\u8fdb\u4e86\u8bad\u7ec3\u6d41\u7a0b\uff0c\u589e\u5f3a\u4ea4\u4e92\u80fd\u529b\u548c\u91c7\u7528\u66f4\u9ad8\u6548\u67b6\u6784\u3002\u5176\u8fed\u4ee3\u5305\u62ec\u4ece\u9884\u8bad\u7ec3\u5230\u5fae\u8c03\uff0c\u518d\u5230\u57fa\u4e8e\u4eba\u7c7b\u53cd\u9988\u7684\u5f3a\u5316\u5b66\u4e60\uff0c\u4f53\u73b0\u4e86\u6a21\u578b\u7684\u6301\u7eed\u4f18\u5316\u548c\u8fdb\u6b65\u3002\u603b\u4f53\u800c\u8a00\uff0c<\/span>MM-LLMs\u6b63\u671d\u7740\u66f4\u5e7f\u6cdb\u6a21\u6001\u652f\u6301\u3001\u66f4\u9ad8\u8d28\u91cf\u6570\u636e\u5229\u7528\u548c\u66f4\u6709\u6548\u6a21\u578b\u8bbe\u8ba1\u7684\u65b9\u5411\u53d1\u5c55\u3002<\/span><\/span><\/section>\n![\"\"](\"https:\/\/linguaresources.com\/wp-content\/uploads\/2024\/08\/640-23.jpg\")
<\/section>\n1 MM-LLMs\u7684\u57fa\u672c\u67b6\u6784<\/strong><\/span><\/h2>\n\u591a\u6a21\u6001\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff08MM-LLMs<\/span>\uff09\u7684\u67b6\u6784\u7531\u4e94\u4e2a\u4e3b\u8981\u7ec4\u4ef6\u6784\u6210\uff1a\u6a21\u6001\u7f16\u7801\u5668\u3001\u8f93\u5165\u6295\u5f71\u5668\u3001LLM<\/span>\u4e3b\u5e72\u3001\u8f93\u51fa\u6295\u5f71\u5668\u548c\u6a21\u6001\u751f\u6210\u5668<\/strong>\u3002<\/span><\/span><\/section>\n![\"\"](\"https:\/\/linguaresources.com\/wp-content\/uploads\/2024\/08\/640-20-5.png\")
<\/section>\n\n- \n\u6a21\u6001\u7f16\u7801\u5668\uff08<\/span>Modality Encoder\uff09<\/span><\/span><\/strong>\uff1a<\/span><\/span><\/section>\n\u6a21\u6001\u7f16\u7801\u5668\u7684\u4e3b\u8981\u529f\u80fd\u662f\u5c06\u4e0d\u540c\u6a21\u6001\u7684\u8f93\u5165\u8f6c\u6362\u6210\u5bf9\u5e94\u7684\u7279\u5f81\u8868\u793a<\/span>FX\u3002\u4f8b\u5982\uff0c\u5b83\u53ef\u4ee5\u5c06\u56fe\u50cf\u3001\u89c6\u9891\u3001\u97f3\u9891\u7b49\u975e\u6587\u672c\u4fe1\u606f\u8f6c\u5316\u4e3a\u6a21\u578b\u53ef\u7406\u89e3\u7684\u7279\u5f81\u5411\u91cf\u3002\u8fd9\u4e2a\u8f6c\u6362\u8fc7\u7a0b\u901a\u8fc7\u9884\u8bad\u7ec3\u7684\u7f16\u7801\u5668\u5b9e\u73b0\uff0c\u5982\u9488\u5bf9\u56fe\u50cf\u7684<\/span>ViT\u6216\u9488\u5bf9\u97f3\u9891\u7684\u7279\u5b9a\u97f3\u9891\u7f16\u7801\u5668\u3002<\/span><\/span><\/section>\n<\/li>\n
- \n\u8f93\u5165\u6295\u5f71\u5668\uff08<\/span>Input Projector\uff09<\/span><\/span><\/strong>\uff1a<\/span><\/span><\/section>\n\u8f93\u5165\u6295\u5f71\u5668\u7684\u4f5c\u7528\u662f\u5c06\u6a21\u6001\u7f16\u7801\u5668\u4ea7\u751f\u7684\u7279\u5f81\u6620\u5c04\u5230\u8bed\u8a00\u6a21\u578b\u4e3b\u5e72\u53ef\u4ee5\u7406\u89e3\u7684\u7a7a\u95f4\u3002\u5b83\u786e\u4fdd\u4e86\u4e0d\u540c\u6a21\u6001\u4fe1\u606f\u53ef\u4ee5\u88ab\u6574\u5408\u5e76\u4f20\u9012\u7ed9\u6a21\u578b\u7684\u6838\u5fc3\u90e8\u5206\u8fdb\u884c\u5904\u7406\u3002<\/span><\/section>\n<\/li>\n
- \nLLM\u4e3b\u5e72\uff08<\/span>LLM Backbone\uff09<\/span><\/span><\/strong>\uff1a<\/span><\/span><\/section>\n\u8fd9\u662f\u6a21\u578b\u7684\u4e2d\u5fc3\u90e8\u5206\uff0c\u57fa\u4e8e\u9884\u8bad\u7ec3\u7684\u6587\u672c\u578b\u8bed\u8a00\u6a21\u578b\u3002\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0c\u8fd9\u90e8\u5206\u4e00\u822c\u4fdd\u6301\u51bb\u7ed3\u72b6\u6001\uff0c\u4ee5\u4fdd\u7559\u6a21\u578b\u539f\u6709\u7684\u8bed\u8a00\u7406\u89e3\u548c\u751f\u6210\u80fd\u529b\u3002<\/span>LLM\u4e3b\u5e72\u8d1f\u8d23\u5904\u7406\u4ece\u8f93\u5165\u6295\u5f71\u5668\u63a5\u6536\u5230\u7684\u7279\u5f81\uff0c\u5e76\u751f\u6210\u4e2d\u95f4\u8868\u793a\u3002<\/span><\/span><\/section>\n<\/li>\n
- \n\u8f93\u51fa\u6295\u5f71\u5668\uff08<\/span>Output Projector\uff09<\/span><\/span><\/strong>\uff1a<\/span><\/span><\/section>\n\u8f93\u51fa\u6295\u5f71\u5668\u4e0e\u8f93\u5165\u6295\u5f71\u5668\u76f8\u53cd\uff0c\u5b83\u5c06\u4ece<\/span>LLM\u4e3b\u5e72\u5f97\u5230\u7684\u4e2d\u95f4\u8868\u793a\u8f6c\u6362\u4e3a\u76ee\u6807\u6a21\u6001\u7684\u8f93\u51fa\uff0c\u4f8b\u5982\u5c06\u4e2d\u95f4\u8868\u793a\u8f6c\u5316\u4e3a\u6587\u672c\u3001\u56fe\u50cf\u6216\u5176\u4ed6\u6a21\u6001\u7684\u8f93\u51fa\u3002<\/span><\/span><\/section>\n<\/li>\n
- \n\u6a21\u6001\u751f\u6210\u5668\uff08<\/span>Modality Generator\uff09<\/span><\/span><\/strong>\uff1a<\/span><\/span><\/section>\n\u6a21\u6001\u751f\u6210\u5668\u7528\u4e8e\u751f\u6210\u65b0\u6a21\u6001\u7684\u6570\u636e\uff0c\u6bd4\u5982\u6839\u636e\u6587\u672c\u63cf\u8ff0\u751f\u6210\u56fe\u50cf\u3002\u5b83\u5728\u9700\u8981\u751f\u6210\u8f93\u51fa\u6a21\u6001\u65f6\u53d1\u6325\u4f5c\u7528\uff0c\u4f8b\u5982\u89c6\u89c9\u63cf\u8ff0\u6216\u8bed\u97f3\u5408\u6210\u3002<\/span><\/section>\n<\/li>\n<\/ol>\n
2 MM-LLMs\u7684\u8bad\u7ec3\u6d41\u7a0b<\/strong><\/span><\/h2>\n\u591a\u6a21\u6001\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff08MM-LLMs<\/span>\uff09\u7684\u8bad\u7ec3\u6d41\u7a0b\u4e3b\u8981\u5305\u62ec\u4e24\u4e2a\u9636\u6bb5\uff1a\u591a\u6a21\u6001\u9884\u8bad\u7ec3\uff08MM PT<\/span>\uff09\u548c\u591a\u6a21\u6001\u6307\u4ee4\u5fae\u8c03\uff08MM IT<\/span>\uff09\u3002<\/span><\/span><\/section>\n
<\/span><\/span><\/section>\n\n- \n\u591a\u6a21\u6001\u9884\u8bad\u7ec3\uff08<\/span>MM PT\uff09<\/span><\/span><\/strong>\uff1a<\/span><\/strong><\/span><\/section>\nMM PT\u9636\u6bb5\u5229\u7528<\/span>X-Text\u6570\u636e\u96c6\uff0c\u65e8\u5728\u8bad\u7ec3\u8f93\u5165\u6295\u5f71\u5668\u548c\u8f93\u51fa\u6295\u5f71\u5668\uff0c\u4f7f\u5b83\u4eec\u80fd\u591f\u5b9e\u73b0\u4e0d\u540c\u6a21\u6001\u95f4\u7684\u5bf9\u9f50\u3002\u5bf9\u4e8e\u4ec5\u6d89\u53ca\u591a\u6a21\u6001\u7406\u89e3\u7684\u6a21\u578b\uff0c\u4f18\u5316\u4e3b\u8981\u96c6\u4e2d\u5728\u4f7f\u8f93\u5165\u6295\u5f71\u5668\u80fd\u5c06\u4e0d\u540c\u6a21\u6001\u8f93\u5165\u6620\u5c04\u5230\u7edf\u4e00\u8868\u793a\u7a7a\u95f4\u3002\u5bf9\u4e8e\u6d89\u53ca\u591a\u6a21\u6001\u751f\u6210\u7684\u6a21\u578b\uff0c\u4f18\u5316\u5305\u62ec\u4f7f\u8f93\u5165\u6295\u5f71\u5668\u3001\u8f93\u51fa\u6295\u5f71\u5668\u4ee5\u53ca\u6a21\u6001\u751f\u6210\u5668\u534f\u540c\u5de5\u4f5c\uff0c\u5b9e\u73b0\u6a21\u6001\u95f4\u7684\u8f6c\u6362\u548c\u751f\u6210\u3002<\/span>X-Text\u6570\u636e\u96c6\u5305\u542b<\/span>Image-Text\u3001<\/span>Video-Text\u548c<\/span>Audio-Text\u5bf9\uff0c\u5176\u4e2d<\/span>Image-Text\u6709<\/span>Image-Text\u914d\u5bf9\u548c\u4ea4\u9519<\/span>Image-Text\u8bed\u6599\u5e93\u4e24\u79cd\u7c7b\u578b\u3002<\/span><\/span><\/section>\n<\/li>\n
- \n\u591a\u6a21\u6001\u6307\u4ee4\u5fae\u8c03\uff08<\/span>MM IT\uff09<\/span><\/span><\/strong>\uff1a<\/span><\/strong><\/span><\/section>\nMM IT\u662f\u6307\u4f7f\u7528\u6307\u4ee4\u683c\u5f0f\u5316\u6570\u636e\u96c6\u5bf9\u9884\u8bad\u7ec3\u7684<\/span>MM-LLMs\u8fdb\u884c\u5fae\u8c03\uff0c\u8ba9\u6a21\u578b\u80fd\u591f\u6cdb\u5316\u5230\u672a\u89c1\u8fc7\u7684\u4efb\u52a1\u3002\u8fd9\u4e00\u8fc7\u7a0b\u4f7f<\/span>MM-LLMs\u80fd\u591f\u66f4\u597d\u5730\u7406\u89e3\u4eba\u7c7b\u610f\u56fe\uff0c\u589e\u5f3a\u5bf9\u8bdd\u4ea4\u4e92\u80fd\u529b\uff0c\u4ece\u800c\u63d0\u5347\u6a21\u578b\u5728\u5b9e\u9645\u5e94\u7528\u4e2d\u7684\u6027\u80fd\u3002<\/span><\/span><\/section>\n<\/li>\n<\/ol>\n
3 MM-LLMs\u7684\u672a\u6765\u53d1\u5c55<\/strong><\/span><\/h2>\n\n- \n\u6700\u65b0\u53d1\u5c55\u8d8b\u52bf\uff1a<\/span><\/strong><\/span><\/section>\n<\/li>\n<\/ol>\n
\n- \n\u63d0\u5347\u89c6\u89c9\u5206\u8fa8\u7387\uff1a\u6a21\u578b\u91c7\u7528\u66f4\u9ad8\u5206\u8fa8\u7387\u56fe\u50cf\uff0c\u4ee5\u6355\u83b7\u66f4\u7cbe\u7ec6\u7684\u7ec6\u8282\uff0c\u5982LLaVA-1.5\u3001VILA\u3001Qwen-VL\u548cMiniGPT-v2\u3002<\/span><\/section>\n<\/li>\n
- \n\u9ad8\u8d28\u91cf\u76d1\u7763\u5fae\u8c03\u6570\u636e\uff1a\u6574\u5408\u9ad8\u8d28\u91cf<\/span>SFT<\/span>\u6570\u636e\uff0c\u663e\u8457\u63d0\u5347\u6a21\u578b\u5728\u7279\u5b9a\u4efb\u52a1\u4e0a\u7684\u6027\u80fd\u3002<\/span><\/span><\/section>\n<\/li>\n
- \n\u6539\u8fdb\u8bad\u7ec3\u7b56\u7565\uff1a\u91c7\u7528\u53c2\u6570\u9ad8\u6548\u5fae\u8c03\u3001\u4ea4\u9519\u56fe\u50cf-\u6587\u672c\u6570\u636e\u8bad\u7ec3\u7b49\u65b9\u6cd5\u3002<\/span><\/section>\n<\/section>\n<\/li>\n<\/ul>\n2. \u672a\u6765\u7814\u7a76\u65b9\u5411\uff1a<\/span><\/strong><\/span><\/section>\n
\n- \n\u6784\u5efa\u66f4\u901a\u7528\u667a\u80fd\u6a21\u578b\uff1a\u6269\u5c55\u6a21\u6001\u652f\u6301\uff0c\u591a\u6837\u5316LLMs\uff0c\u63d0\u5347\u6570\u636e\u96c6\u8d28\u91cf\uff0c\u52a0\u5f3a\u751f\u6210\u80fd\u529b\u3002<\/span><\/section>\n<\/li>\n
- \n\u66f4\u5177\u6311\u6218\u6027\u7684\u57fa\u51c6\u6d4b\u8bd5\uff1a\u5b8c\u5584\u8bc4\u4f30\u6807\u51c6\uff0c\u5168\u9762\u8861\u91cfMM-LLMs\u7684\u9ad8\u7ea7\u529f\u80fd\u3002<\/span><\/section>\n<\/li>\n
- \n\u79fb\u52a8\/\u8f7b\u91cf\u7ea7\u90e8\u7f72\uff1a\u7814\u7a76\u8f7b\u91cf\u5316\u7b56\u7565\uff0c\u4ee5\u9002\u5e94\u8d44\u6e90\u53d7\u9650\u7684\u5e73\u53f0\u3002<\/span><\/section>\n<\/li>\n
- \n\u5177\u8eab\u667a\u80fd\uff1a\u4f7f\u6a21\u578b\u80fd\u591f\u50cf\u4eba\u7c7b\u4e00\u6837\u611f\u77e5\u548c\u4e92\u52a8\uff0c\u6709\u6548\u7406\u89e3\u73af\u5883\u5e76\u8fdb\u884c\u64cd\u4f5c\u3002<\/span><\/section>\n<\/li>\n<\/ul>\n
4\u00a0\u7ed3\u8bed<\/strong><\/span><\/h2>\n\u8bba\u6587\u5168\u9762\u6982\u8ff0\u4e86\u591a\u6a21\u6001\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff08MM-LLMs<\/span>\uff09\u7684\u6700\u65b0\u8fdb\u5c55\uff0c\u4ece\u6a21\u578b\u67b6\u6784\u3001\u8bad\u7ec3\u7b56\u7565\u5230\u6027\u80fd\u8bc4\u4f30\u548c\u672a\u6765\u65b9\u5411\uff0c\u4e3a\u63a8\u52a8\u8be5\u9886\u57df\u7814\u7a76\u63d0\u4f9b\u4e86\u8be6\u5b9e\u7684\u7efc\u8ff0\u548c\u8d44\u6e90\u3002\u4e3a\u4e86\u7d27\u8ddf\u591a\u6a21\u6001\u5927\u8bed\u8a00\u6a21\u578b\uff08MM-LLMs<\/span>\uff09\u7684\u6700\u65b0\u8fdb\u5c55\uff0c\u4f5c\u8005\u521b\u5efa\u4e86\u4e00\u4e2a\u5b9e\u65f6\u8ddf\u8e2a\u5e73\u53f0\uff1ahttps:\/\/mm-llms.github.io<\/span><\/strong>\u3002\u8be5\u7f51\u7ad9\u65e8\u5728\u4fc3\u8fdb\u77e5\u8bc6\u5171\u4eab\uff0c\u8bb0\u5f55\u7814\u7a76\u52a8\u6001\uff0c\u5e76\u9f13\u52b1\u793e\u533a\u53c2\u4e0e\uff0c\u5bf9\u63a8\u52a8\u9886\u57df\u521b\u65b0\u3001\u907f\u514d\u91cd\u590d\u52b3\u52a8\u548c\u52a0\u5f3a\u5b66\u672f\u4ea4\u6d41\u5177\u6709\u91cd\u8981\u4f5c\u7528\u3002<\/span><\/span><\/section>\n<\/section>\n\u8bba\u6587\u9898\u76ee\uff1aMM-LLMs: Recent Advances in MultiModal Large Language Models<\/span><\/span><\/section>\n\u8bba\u6587\u94fe\u63a5\uff1ahttps:\/\/arxiv.org\/abs\/2401.13601<\/span><\/span><\/span><\/section>\n<\/section>\nPS:\u00a0<\/span>\u6b22\u8fce\u5927\u5bb6\u626b\u7801\u5173\u6ce8\u516c\u4f17\u53f7^_^<\/span>\uff0c\u6211\u4eec\u4e00\u8d77\u5728AI\u7684\u4e16\u754c\u4e2d\u63a2\u7d22\u524d\u884c\uff0c\u671f\u5f85\u5171\u540c\u8fdb\u6b65\uff01<\/span><\/section>\n\n\n<\/section>\n
\n\n\u7279\u522b\u8bf4\u660e\uff1a\u672c\u6587\u4ec5\u7528\u4e8e\u5b66\u672f\u4ea4\u6d41\uff0c\u5982\u6709\u4fb5\u6743\u8bf7\u540e\u53f0\u8054\u7cfb\u5c0f\u7f16\u5220\u9664\u3002<\/span><\/span><\/section>\n<\/section>\n<\/blockquote>\n\n\u8f6c\u8f7d\u6765\u6e90\uff1a\u5927\u6a21\u578b\u591a\u6a21\u6001\u8bba\u6587\u89e3\u8bfb<\/span><\/section>\n<\/section>\n<\/section>\n<\/section>\n <\/p>\n<\/section>\n","protected":false},"excerpt":{"rendered":"
ChatGPT | \u591a\u6a21\u6001\u5927\u8bed\u8a00\u6a21\u578b\u6700\u65b0\u8fdb\u5c55 \u7ffb\u8bd1\u6280\u672f\u6559\u80b2\u4e0e\u7814\u7a76\u00a02024\u5e7407\u670820\u65e5 00:01\u00a0\u9655\u897f \u4ee5 […]<\/p>\n","protected":false},"author":1,"featured_media":32341,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[391],"tags":[],"class_list":["post-32340","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-391"],"_links":{"self":[{"href":"https:\/\/linguaresources.com\/index.php?rest_route=\/wp\/v2\/posts\/32340","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/linguaresources.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/linguaresources.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/linguaresources.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/linguaresources.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=32340"}],"version-history":[{"count":1,"href":"https:\/\/linguaresources.com\/index.php?rest_route=\/wp\/v2\/posts\/32340\/revisions"}],"predecessor-version":[{"id":32343,"href":"https:\/\/linguaresources.com\/index.php?rest_route=\/wp\/v2\/posts\/32340\/revisions\/32343"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/linguaresources.com\/index.php?rest_route=\/wp\/v2\/media\/32341"}],"wp:attachment":[{"href":"https:\/\/linguaresources.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=32340"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/linguaresources.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=32340"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/linguaresources.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=32340"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}