{"id":5410,"date":"2024-12-25T14:26:14","date_gmt":"2024-12-25T06:26:14","guid":{"rendered":"https:\/\/nullthought.net\/?p=5410"},"modified":"2024-12-25T14:27:19","modified_gmt":"2024-12-25T06:27:19","slug":"metamorph%ef%bc%9a%e5%ae%9e%e7%8e%b0%e8%a7%86%e8%a7%89%e7%90%86%e8%a7%a3%e4%b8%8e%e7%94%9f%e6%88%90%e7%bb%9f%e4%b8%80%e7%9a%84%e5%a4%9a%e6%a8%a1%e6%80%81%e6%a8%a1%e5%9e%8b","status":"publish","type":"post","link":"https:\/\/nullthought.net\/?p=5410","title":{"rendered":"MetaMorph\uff1a\u5b9e\u73b0\u89c6\u89c9\u7406\u89e3\u4e0e\u751f\u6210\u7edf\u4e00\u7684\u591a\u6a21\u6001\u6a21\u578b"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">\u8bba\u6587<strong><a href=\"https:\/\/arxiv.org\/abs\/2412.14164\" target=\"_blank\" rel=\"noreferrer noopener\">MetaMorph: Multimodal Understanding and Generation via Instruction Tuning<\/a><\/strong>\u63d0\u51fa\u4e86\u4e00\u79cd\u65b0\u7684\u5fae\u8c03\u6846\u67b6\u2014\u2014<strong>Visual-Predictive Instruction Tuning (VPiT)<\/strong>\uff0c\u5e76\u6784\u5efa\u591a\u6a21\u6001\u6a21\u578b <strong>MetaMorph<\/strong>\uff0c<strong>\u5b9e\u73b0\u89c6\u89c9\u7406\u89e3\u548c\u751f\u6210\u7684\u7edf\u4e00<\/strong>\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u8bba\u6587\u4f5c\u8005\u4e3aShengbang Tong, David Fan, Jiachen Zhu, Yunyang Xiong, Xinlei Chen, Koustuv Sinha, Michael Rabbat, <strong>Yann LeCun<\/strong>, Saining Xie, Zhuang Liu\uff0c\u5747\u6765\u81eaMeta\u3002<a href=\"https:\/\/nullthought.net\/?s=%E6%9D%A8%E7%AB%8B%E6%98%86\" target=\"_blank\" rel=\"noreferrer noopener\">\u6768\u7acb\u6606\uff08Yann LeGun\uff09\u6559\u6388<\/a>\u4e5f\u662f\u8bba\u6587\u4f5c\u8005\u4e4b\u4e00\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"479\" src=\"https:\/\/nullthought.net\/wp-content\/uploads\/2024\/12\/image-17-1024x479.png\" alt=\"\" class=\"wp-image-5411\" srcset=\"https:\/\/nullthought.net\/wp-content\/uploads\/2024\/12\/image-17-1024x479.png 1024w, https:\/\/nullthought.net\/wp-content\/uploads\/2024\/12\/image-17-300x140.png 300w, https:\/\/nullthought.net\/wp-content\/uploads\/2024\/12\/image-17-768x359.png 768w, https:\/\/nullthought.net\/wp-content\/uploads\/2024\/12\/image-17.png 1030w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><strong><a href=\"https:\/\/arxiv.org\/abs\/2412.14164\" target=\"_blank\" rel=\"noreferrer noopener\">MetaMorph: Multimodal Understanding and Generation via Instruction Tuning<\/a><\/strong><\/figcaption><\/figure>\n\n\n\n<h5 class=\"wp-block-heading\">1. \u80cc\u666f\u4e0e\u7814\u7a76\u52a8\u673a<\/h5>\n\n\n\n<h6 class=\"wp-block-heading\">1.1 \u591a\u6a21\u6001\u6a21\u578b\u7684\u73b0\u72b6<\/h6>\n\n\n\n<p class=\"wp-block-paragraph\">\u591a\u6a21\u6001\u5927\u578b\u8bed\u8a00\u6a21\u578b\uff08MLLMs\uff09\u7684\u53d1\u5c55\uff0c\u4f7f\u5f97\u673a\u5668\u80fd\u591f\u540c\u65f6\u7406\u89e3\u548c\u5904\u7406\u56fe\u50cf\u4e0e\u6587\u672c\u3002\u7136\u800c\uff0c\u5f53\u524d\u6a21\u578b\u5927\u591a\u5c40\u9650\u4e8e\u5355\u4e00\u4efb\u52a1\uff08\u5982\u56fe\u50cf\u7406\u89e3\u6216\u751f\u6210\uff09\uff0c\u4e14\u5b9e\u73b0\u8fd9\u4e9b\u4efb\u52a1\u901a\u5e38\u9700\u8981\u5927\u89c4\u6a21\u7684\u6570\u636e\u8bad\u7ec3\u548c\u590d\u6742\u7684\u6a21\u578b\u67b6\u6784\u8c03\u6574\u3002\u4ee5\u5f80\u7684\u7814\u7a76\u8868\u660e\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u89c6\u89c9\u7406\u89e3\u4efb\u52a1\uff0c\u5982\u56fe\u50cf\u95ee\u7b54\uff08ImageQA\uff09\uff0c\u901a\u8fc7\u89c6\u89c9\u7f16\u7801\u5668\u5c06\u56fe\u50cf\u5d4c\u5165\u5230\u8bed\u8a00\u6a21\u578b\u4e2d\uff0c\u53ef\u4ee5\u9ad8\u6548\u5730\u5b9e\u73b0\u3002<\/li>\n\n\n\n<li>\u89c6\u89c9\u751f\u6210\u4efb\u52a1\u5219\u901a\u5e38\u88ab\u8ba4\u4e3a\u662f\u72ec\u7acb\u4e8e\u7406\u89e3\u7684\uff0c\u9700\u8981\u989d\u5916\u7684\u751f\u6210\u6a21\u5757\uff0c\u5982\u6269\u6563\u6a21\u578b\uff08Diffusion Models\uff09\u3002<\/li>\n<\/ul>\n\n\n\n<h6 class=\"wp-block-heading\">1.2 \u5b58\u5728\u7684\u95ee\u9898<\/h6>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u751f\u6210\u4e0e\u7406\u89e3\u5272\u88c2<\/strong>\uff1a\u73b0\u6709\u65b9\u6cd5\u4e2d\uff0c\u89c6\u89c9\u751f\u6210\u548c\u7406\u89e3\u901a\u5e38\u88ab\u5206\u5f00\u5904\u7406\uff0c\u8fd9\u9650\u5236\u4e86\u6a21\u578b\u7684\u591a\u4efb\u52a1\u80fd\u529b\u3002<\/li>\n\n\n\n<li><strong>\u9ad8\u6210\u672c\u9700\u6c42<\/strong>\uff1a\u89c6\u89c9\u751f\u6210\u4efb\u52a1\u901a\u5e38\u9700\u8981\u5927\u91cf\u7684\u6570\u636e\u548c\u8ba1\u7b97\u8d44\u6e90\u3002<\/li>\n\n\n\n<li><strong>\u67b6\u6784\u590d\u6742\u6027<\/strong>\uff1a\u5b9e\u73b0\u591a\u6a21\u6001\u80fd\u529b\u7684\u6a21\u578b\u8bbe\u8ba1\u5f80\u5f80\u9700\u8981\u5728\u73b0\u6709\u67b6\u6784\u4e0a\u5f15\u5165\u5927\u91cf\u8c03\u6574\uff0c\u589e\u52a0\u4e86\u5b9e\u73b0\u96be\u5ea6\u3002<\/li>\n<\/ol>\n\n\n\n<h6 class=\"wp-block-heading\">1.3 \u7814\u7a76\u76ee\u6807<\/h6>\n\n\n\n<p class=\"wp-block-paragraph\">\u4f5c\u8005\u63d0\u51fa\u7684\u4e3b\u8981\u76ee\u6807\u662f\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u901a\u8fc7\u8f7b\u91cf\u7ea7\u5fae\u8c03\uff08VPiT\uff09\uff0c\u5728\u9884\u8bad\u7ec3\u8bed\u8a00\u6a21\u578b\u7684\u57fa\u7840\u4e0a\u5b9e\u73b0\u7edf\u4e00\u7684\u89c6\u89c9\u7406\u89e3\u4e0e\u751f\u6210\u80fd\u529b\u3002<\/li>\n\n\n\n<li>\u63a2\u8ba8\u89c6\u89c9\u751f\u6210\u548c\u7406\u89e3\u4efb\u52a1\u4e4b\u95f4\u7684\u534f\u540c\u5173\u7cfb\uff0c\u63ed\u793a\u4e24\u8005\u7684\u6f5c\u5728\u4e92\u8865\u6027\u3002<\/li>\n\n\n\n<li>\u901a\u8fc7\u8054\u5408\u591a\u6a21\u6001\u6570\u636e\u8bad\u7ec3\u4e00\u4e2a\u6a21\u578b\uff0c\u5b9e\u73b0\u89c6\u89c9\u751f\u6210\u4efb\u52a1\u6240\u9700\u7684\u6570\u636e\u6548\u7387\u63d0\u5347\u3002<\/li>\n<\/ul>\n\n\n\n<h5 class=\"wp-block-heading\">2. \u7814\u7a76\u65b9\u6cd5<\/h5>\n\n\n\n<h6 class=\"wp-block-heading\">2.1 VPiT\uff08Visual-Predictive Instruction Tuning\uff09<\/h6>\n\n\n\n<h6 class=\"wp-block-heading\">\u6838\u5fc3\u601d\u60f3<\/h6>\n\n\n\n<p class=\"wp-block-paragraph\">\u901a\u8fc7\u6269\u5c55\u89c6\u89c9\u6307\u4ee4\u5fae\u8c03\uff08Visual Instruction Tuning\uff09\uff0c\u5b9e\u73b0\u7edf\u4e00\u7684\u89c6\u89c9\u7406\u89e3\u4e0e\u751f\u6210\u3002\u5177\u4f53\u6b65\u9aa4\u5305\u62ec\uff1a<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u4ece\u5355\u6a21\u6001\u5230\u591a\u6a21\u6001\u7684\u9884\u6d4b\u4efb\u52a1<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u8f93\u5165\u6269\u5c55\u4e3a\u56fe\u50cf\u548c\u6587\u672c\u7684\u6df7\u5408\u5e8f\u5217\u3002<\/li>\n\n\n\n<li>\u8f93\u51fa\u65e2\u5305\u62ec\u79bb\u6563\u7684\u6587\u672c\u4ee4\u724c\uff0c\u4e5f\u5305\u62ec\u8fde\u7eed\u7684\u89c6\u89c9\u4ee4\u724c\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u591a\u6a21\u6001\u6570\u636e\u7684\u4ee4\u724c\u5316<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u6587\u672c\u6570\u636e\u901a\u8fc7\u8bed\u8a00\u6a21\u578b\u7684\u6807\u51c6\u4ee4\u724c\u5668\u8fdb\u884c\u79bb\u6563\u5316\u3002<\/li>\n\n\n\n<li>\u56fe\u50cf\u6570\u636e\u901a\u8fc7\u9884\u8bad\u7ec3\u89c6\u89c9\u7f16\u7801\u5668\uff08\u5982 SigLIP\uff09\u8f6c\u6362\u4e3a\u8fde\u7eed\u89c6\u89c9\u4ee4\u724c\uff0c\u5e76\u901a\u8fc7\u6295\u5f71\u5c42\u8c03\u6574\u7ef4\u5ea6\u4ee5\u9002\u914d\u8bed\u8a00\u6a21\u578b\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u6a21\u578b\u67b6\u6784<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u5728\u9884\u8bad\u7ec3\u8bed\u8a00\u6a21\u578b\u7684\u57fa\u7840\u4e0a\u589e\u52a0\u201c\u89c6\u89c9\u5934\u201d\uff0c\u7528\u4e8e\u9884\u6d4b\u89c6\u89c9\u4ee4\u724c\u3002<\/li>\n\n\n\n<li>\u5f15\u5165\u7279\u6b8a\u4ee4\u724c <code>&lt;image_start&gt;<\/code> \u548c <code>&lt;image_end&gt;<\/code> \u4ee5\u6807\u8bc6\u89c6\u89c9\u4ee4\u724c\u5e8f\u5217\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u635f\u5931\u51fd\u6570<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u6587\u672c\u751f\u6210\u4efb\u52a1\u91c7\u7528\u6807\u51c6\u7684\u4ea4\u53c9\u71b5\u635f\u5931\u3002<\/li>\n\n\n\n<li>\u89c6\u89c9\u751f\u6210\u4efb\u52a1\u91c7\u7528\u4f59\u5f26\u76f8\u4f3c\u6027\u635f\u5931\uff0c\u6700\u5927\u5316\u6a21\u578b\u9884\u6d4b\u7684\u89c6\u89c9\u4ee4\u724c\u4e0e\u89c6\u89c9\u7f16\u7801\u5668\u751f\u6210\u7684\u5d4c\u5165\u4e4b\u95f4\u7684\u76f8\u4f3c\u5ea6\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u4ee4\u724c\u53ef\u89c6\u5316<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u4f7f\u7528\u6269\u5c55\u7684\u6269\u6563\u6a21\u578b\u5c06\u9884\u6d4b\u7684\u89c6\u89c9\u4ee4\u724c\u6620\u5c04\u56de\u50cf\u7d20\u7a7a\u95f4\uff0c\u5b9e\u73b0\u89c6\u89c9\u751f\u6210\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ol>\n\n\n\n<h6 class=\"wp-block-heading\">2.2 \u6570\u636e\u7c7b\u578b\u4e0e\u5904\u7406<\/h6>\n\n\n\n<p class=\"wp-block-paragraph\">VPiT \u652f\u6301\u591a\u79cd\u7c7b\u578b\u7684\u591a\u6a21\u6001\u6570\u636e\uff0c\u6db5\u76d6\u89c6\u89c9\u7406\u89e3\u548c\u751f\u6210\u4efb\u52a1\uff0c\u5177\u4f53\u5305\u62ec\uff1a<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u89c6\u89c9\u7406\u89e3\u6570\u636e<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li><strong>\u56fe\u50cf\u95ee\u7b54\uff08ImageQA\uff09<\/strong>\uff1a\u4f8b\u5982 Cambrian-7M \u6570\u636e\u96c6\uff0c\u63d0\u4f9b\u56fe\u50cf\u548c\u5bf9\u5e94\u7684\u95ee\u9898-\u56de\u7b54\u5bf9\u3002<\/li>\n\n\n\n<li><strong>\u89c6\u9891\u95ee\u7b54\uff08VideoQA\uff09<\/strong>\uff1a\u5982 VideoStar \u6570\u636e\u96c6\uff0c\u8f93\u5165\u4e3a\u89c6\u9891\u5e27\u548c\u6587\u672c\u95ee\u9898\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u89c6\u89c9\u751f\u6210\u6570\u636e<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li>\u901a\u8fc7\u63cf\u8ff0\u751f\u6210\u56fe\u50cf\uff0c\u5982 MetaCLIP \u6570\u636e\u96c6\uff0c\u8f93\u5165\u4e3a\u6587\u672c\u63cf\u8ff0\uff0c\u8f93\u51fa\u4e3a\u89c6\u89c9\u4ee4\u724c\u3002<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>\u6df7\u5408\u89c6\u89c9\u6570\u636e<\/strong>\uff1a\n<ul class=\"wp-block-list\">\n<li><strong>\u89c6\u9891\u9884\u6d4b<\/strong>\uff1a\u6839\u636e\u90e8\u5206\u89c6\u9891\u5e27\u9884\u6d4b\u672a\u6765\u5e27\u3002<\/li>\n\n\n\n<li><strong>\u56fe\u50cf\u63a8\u7406<\/strong>\uff1a\u901a\u8fc7\u8f93\u5165\u56fe\u50cf\u751f\u6210\u53d8\u6362\u540e\u7684\u56fe\u50cf\uff08\u5982 InstructPix2Pix \u6570\u636e\u96c6\uff09\u3002<\/li>\n<\/ul>\n<\/li>\n<\/ol>\n\n\n\n<h6 class=\"wp-block-heading\">2.3 \u53ef\u89c6\u5316\u89c6\u89c9\u4ee4\u724c<\/h6>\n\n\n\n<p class=\"wp-block-paragraph\">\u901a\u8fc7\u4e00\u4e2a\u5fae\u8c03\u7684\u6269\u6563\u6a21\u578b\uff08\u5982 Stable Diffusion\uff09\uff0c\u5c06\u8fde\u7eed\u89c6\u89c9\u4ee4\u724c\u6620\u5c04\u4e3a\u50cf\u7d20\u6570\u636e\u3002\u6b64\u8fc7\u7a0b\u9700\u8981\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u4f7f\u7528\u9884\u8bad\u7ec3\u7684\u6269\u6563\u6a21\u578b\u3002<\/li>\n\n\n\n<li>\u5f15\u5165\u4e00\u4e2a\u4e24\u5c42\u7684\u591a\u5c42\u611f\u77e5\u673a\uff08MLP\uff09\u6295\u5f71\u5668\uff0c\u7528\u4e8e\u5c06\u89c6\u89c9\u4ee4\u724c\u7ef4\u5ea6\u8c03\u6574\u4e3a\u6269\u6563\u6a21\u578b\u7684\u6761\u4ef6\u8f93\u5165\u7ef4\u5ea6\u3002<\/li>\n<\/ul>\n\n\n\n<h5 class=\"wp-block-heading\">3. \u5b9e\u9a8c\u8bbe\u8ba1\u4e0e\u7ed3\u679c\u5206\u6790<\/h5>\n\n\n\n<h6 class=\"wp-block-heading\">3.1 \u5b9e\u9a8c\u95ee\u9898<\/h6>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u89c6\u89c9\u751f\u6210\u80fd\u5426\u901a\u8fc7\u8f7b\u91cf\u7ea7\u5fae\u8c03\u5b9e\u73b0\uff1f<\/strong><\/li>\n\n\n\n<li><strong>\u89c6\u89c9\u7406\u89e3\u4e0e\u751f\u6210\u4efb\u52a1\u662f\u5426\u76f8\u4e92\u8865\u5145\uff1f<\/strong><\/li>\n\n\n\n<li><strong>\u589e\u52a0\u4e0d\u540c\u6570\u636e\u7c7b\u578b\u5bf9\u6a21\u578b\u6027\u80fd\u7684\u5177\u4f53\u8d21\u732e\u5982\u4f55\uff1f<\/strong><\/li>\n<\/ol>\n\n\n\n<h6 class=\"wp-block-heading\">3.2 \u5b9e\u9a8c\u53d1\u73b0<\/h6>\n\n\n\n<h6 class=\"wp-block-heading\">\u53d1\u73b01\uff1a\u89c6\u89c9\u751f\u6210\u80fd\u529b\u901a\u8fc7\u8054\u5408\u8bad\u7ec3\u9ad8\u6548\u89e3\u9501<\/h6>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u5355\u72ec\u8bad\u7ec3\u89c6\u89c9\u751f\u6210\u4efb\u52a1\u65f6\uff0c\u6a21\u578b\u9700\u8981\u8d85\u8fc7 300 \u4e07\u6761\u6570\u636e\u624d\u80fd\u5b9e\u73b0\u8f83\u9ad8\u8d28\u91cf\u751f\u6210\u3002<\/li>\n\n\n\n<li>\u8054\u5408\u8bad\u7ec3\u89c6\u89c9\u7406\u89e3\u548c\u751f\u6210\u4efb\u52a1\u540e\uff0c\u4ec5\u9700 20 \u4e07\u6761\u751f\u6210\u6570\u636e\u5373\u53ef\u8fbe\u5230\u76f8\u4f3c\u7684\u6027\u80fd\u3002\u8fd9\u8868\u660e\u89c6\u89c9\u751f\u6210\u5e76\u975e\u72ec\u7acb\u80fd\u529b\uff0c\u800c\u662f\u4e0e\u89c6\u89c9\u7406\u89e3\u9ad8\u5ea6\u76f8\u5173\u3002<\/li>\n<\/ul>\n\n\n\n<h6 class=\"wp-block-heading\">\u53d1\u73b02\uff1a\u7406\u89e3\u4e0e\u751f\u6210\u4efb\u52a1\u7684\u534f\u540c\u6548\u5e94<\/h6>\n\n\n\n<ol class=\"wp-block-list\">\n<li>\u589e\u52a0\u89c6\u89c9\u7406\u89e3\u6570\u636e\uff08\u5982 VQA \u6570\u636e\uff09\u53ef\u4ee5\u663e\u8457\u63d0\u5347\u751f\u6210\u8d28\u91cf\u3002<\/li>\n\n\n\n<li>\u589e\u52a0\u89c6\u89c9\u751f\u6210\u6570\u636e\u4e5f\u80fd\u63d0\u5347\u7406\u89e3\u4efb\u52a1\uff08\u5982 VQA\uff09\u7684\u6027\u80fd\uff0c\u4f46\u6548\u679c\u4e0d\u5982\u589e\u52a0\u7406\u89e3\u6570\u636e\u663e\u8457\u3002<\/li>\n<\/ol>\n\n\n\n<h6 class=\"wp-block-heading\">\u53d1\u73b03\uff1a\u89c6\u89c9\u7406\u89e3\u6570\u636e\u8d21\u732e\u66f4\u5927<\/h6>\n\n\n\n<p class=\"wp-block-paragraph\">\u901a\u8fc7\u5bf9\u4e0d\u540c\u7c7b\u578b\u6570\u636e\u7684\u5bf9\u6bd4\u5b9e\u9a8c\u53d1\u73b0\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u89c6\u89c9\u7406\u89e3\u6570\u636e\uff08\u5982 ImageQA \u548c VideoQA\uff09\u5bf9\u751f\u6210\u4efb\u52a1\u7684\u63d0\u5347\u6700\u663e\u8457\u3002<\/li>\n\n\n\n<li>\u89c6\u89c9\u751f\u6210\u6570\u636e\u7684\u8d21\u732e\u8f83\u5c0f\uff0c\u5c24\u5176\u662f\u5728\u7406\u89e3\u6570\u636e\u8f83\u591a\u7684\u60c5\u51b5\u4e0b\u3002<\/li>\n<\/ul>\n\n\n\n<h6 class=\"wp-block-heading\">\u53d1\u73b04\uff1a\u7279\u5b9a\u4efb\u52a1\u7684\u8d21\u732e\u5dee\u5f02<\/h6>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u4e00\u822c\u6027\u4efb\u52a1\uff08\u5982\u9ad8\u5206\u8fa8\u7387\u56fe\u50cf\u95ee\u7b54\uff09\u4e0e\u751f\u6210\u4efb\u52a1\u7684\u76f8\u5173\u6027\u8f83\u5f3a\u3002<\/li>\n\n\n\n<li>\u77e5\u8bc6\u5bc6\u96c6\u578b\u4efb\u52a1\uff08\u5982\u79d1\u5b66\u95ee\u7b54\uff09\u7684\u8d21\u732e\u8f83\u4f4e\uff0c\u8bf4\u660e\u751f\u6210\u4efb\u52a1\u66f4\u4f9d\u8d56\u4e8e\u89c6\u89c9\u5185\u5bb9\u7684\u76f4\u63a5\u7406\u89e3\u3002<\/li>\n<\/ul>\n\n\n\n<h6 class=\"wp-block-heading\">3.3 \u5b9a\u91cf\u7ed3\u679c<\/h6>\n\n\n\n<p class=\"wp-block-paragraph\">\u5728\u591a\u4e2a\u57fa\u51c6\u4efb\u52a1\u4e0a\uff0cMetaMorph \u7684\u8868\u73b0\u4f18\u4e8e\u73b0\u6709\u65b9\u6cd5\uff0c\u5305\u62ec\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u5728\u89c6\u89c9\u7406\u89e3\u4efb\u52a1\uff08\u5982 ImageQA\uff09\u4e0a\uff0c\u4e0e GPT-4V \u7b49\u5148\u8fdb\u6a21\u578b\u8868\u73b0\u76f8\u5f53\u3002<\/li>\n\n\n\n<li>\u5728\u89c6\u89c9\u751f\u6210\u4efb\u52a1\u4e0a\uff0c\u8d85\u8fc7\u90e8\u5206\u4e13\u7528\u751f\u6210\u6a21\u578b\uff08\u5982 Stable Diffusion\uff09\u3002<\/li>\n<\/ul>\n\n\n\n<h5 class=\"wp-block-heading\">4. \u6a21\u578b\u80fd\u529b\u5c55\u793a<\/h5>\n\n\n\n<h6 class=\"wp-block-heading\">4.1 \u5229\u7528 LLM \u7684\u77e5\u8bc6\u751f\u6210\u89c6\u89c9\u5185\u5bb9<\/h6>\n\n\n\n<p class=\"wp-block-paragraph\">MetaMorph \u80fd\u591f\u5229\u7528\u9884\u8bad\u7ec3\u8bed\u8a00\u6a21\u578b\u7684\u77e5\u8bc6\u751f\u6210\u590d\u6742\u7684\u89c6\u89c9\u6982\u5ff5\uff0c\u4f8b\u5982\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u751f\u6210\u201cChhogori\u201d\uff08\u4e54\u6208\u91cc\u5cf0\uff0c\u4e16\u754c\u7b2c\u4e8c\u9ad8\u5cf0\uff09\u7684\u56fe\u50cf\u3002<\/li>\n\n\n\n<li>\u751f\u6210\u201cOncilla\u201d\uff08\u5357\u7f8e\u5c0f\u91ce\u732b\uff09\u7b49\u7279\u5b9a\u9886\u57df\u77e5\u8bc6\u76f8\u5173\u7684\u56fe\u50cf\u3002<\/li>\n<\/ul>\n\n\n\n<h6 class=\"wp-block-heading\">4.2 \u9690\u5f0f\u591a\u6b65\u63a8\u7406<\/h6>\n\n\n\n<p class=\"wp-block-paragraph\">\u901a\u8fc7\u9690\u5f0f\u5b8c\u6210\u591a\u6b65\u903b\u8f91\u63a8\u7406\uff0cMetaMorph \u80fd\u591f\u76f4\u63a5\u751f\u6210\u6b63\u786e\u7684\u89c6\u89c9\u5185\u5bb9\u3002\u4f8b\u5982\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u8f93\u5165\u201c\u751f\u6210\u79d1\u5b66\u5bb6\u7231\u56e0\u65af\u5766\u559c\u6b22\u6f14\u594f\u7684\u4e50\u5668\u201d\uff0c\u6a21\u578b\u63a8\u7406\u51fa\u7b54\u6848\u4e3a\u5c0f\u63d0\u7434\u5e76\u751f\u6210\u5bf9\u5e94\u7684\u56fe\u50cf\u3002<\/li>\n<\/ul>\n\n\n\n<h5 class=\"wp-block-heading\">5. \u610f\u4e49\u4e0e\u672a\u6765\u5c55\u671b<\/h5>\n\n\n\n<h6 class=\"wp-block-heading\">5.1 \u610f\u4e49<\/h6>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong>\u7edf\u4e00\u80fd\u529b<\/strong>\uff1a\u901a\u8fc7\u7b80\u5355\u7684\u5fae\u8c03\uff0cMetaMorph \u5b9e\u73b0\u4e86\u89c6\u89c9\u7406\u89e3\u4e0e\u751f\u6210\u7684\u7edf\u4e00\u3002<\/li>\n\n\n\n<li><strong>\u6570\u636e\u6548\u7387<\/strong>\uff1a\u8054\u5408\u8bad\u7ec3\u663e\u8457\u51cf\u5c11\u4e86\u89c6\u89c9\u751f\u6210\u4efb\u52a1\u5bf9\u5927\u89c4\u6a21\u6570\u636e\u7684\u4f9d\u8d56\u3002<\/li>\n\n\n\n<li><strong>\u6f5c\u5728\u80fd\u529b\u6fc0\u6d3b<\/strong>\uff1a\u5b9e\u9a8c\u8868\u660e\uff0c\u9884\u8bad\u7ec3\u7684\u8bed\u8a00\u6a21\u578b\u4e2d\u5df2\u5b58\u5728\u201c\u9690\u6027\u201d\u7684\u89c6\u89c9\u751f\u6210\u80fd\u529b\uff0c\u901a\u8fc7 VPiT \u53ef\u4ee5\u9ad8\u6548\u6fc0\u6d3b\u3002<\/li>\n<\/ol>\n\n\n\n<h6 class=\"wp-block-heading\">5.2 \u672a\u6765\u65b9\u5411<\/h6>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u63a2\u7d22\u66f4\u591a\u591a\u6a21\u6001\u6570\u636e\u7c7b\u578b\uff0c\u8fdb\u4e00\u6b65\u63d0\u5347\u6a21\u578b\u7684\u7edf\u4e00\u80fd\u529b\u3002<\/li>\n\n\n\n<li>\u4f18\u5316\u5fae\u8c03\u6280\u672f\uff0c\u4f7f\u5176\u9002\u914d\u66f4\u591a\u4efb\u52a1\u573a\u666f\uff0c\u63a8\u52a8\u901a\u7528\u4eba\u5de5\u667a\u80fd\u7684\u53d1\u5c55\u3002<\/li>\n<\/ul>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"wp-block-paragraph\">Metamorph\u9879\u76ee\u4e3b\u9875: <a href=\"https:\/\/tsb0601.github.io\/metamorph\/\" target=\"_blank\" rel=\"noreferrer noopener\">tsb0601.github.io\/metamorph<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u8bba\u6587MetaMorph: Multimodal Understanding and Generation vi [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","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":"","ast-disable-related-posts":"","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":"set","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"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":""},"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":[8],"tags":[39,77,90,86,80],"class_list":["post-5410","post","type-post","status-publish","format-standard","hentry","category-tech","tag-ai","tag-meta","tag-multimodal","tag-llm","tag-cv"],"rttpg_featured_image_url":null,"rttpg_author":{"display_name":"NullThought","author_link":"https:\/\/nullthought.net\/?author=1"},"rttpg_comment":0,"rttpg_category":"<a href=\"https:\/\/nullthought.net\/?cat=8\" rel=\"category\">Tech<\/a>","rttpg_excerpt":"\u8bba\u6587MetaMorph: Multimodal Understanding and Generation vi&hellip;","_links":{"self":[{"href":"https:\/\/nullthought.net\/index.php?rest_route=\/wp\/v2\/posts\/5410","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/nullthought.net\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/nullthought.net\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/nullthought.net\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/nullthought.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=5410"}],"version-history":[{"count":2,"href":"https:\/\/nullthought.net\/index.php?rest_route=\/wp\/v2\/posts\/5410\/revisions"}],"predecessor-version":[{"id":5413,"href":"https:\/\/nullthought.net\/index.php?rest_route=\/wp\/v2\/posts\/5410\/revisions\/5413"}],"wp:attachment":[{"href":"https:\/\/nullthought.net\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=5410"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/nullthought.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=5410"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/nullthought.net\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=5410"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}