{"id":9724,"date":"2026-06-27T14:09:25","date_gmt":"2026-06-27T06:09:25","guid":{"rendered":"https:\/\/infernews.com\/blog\/eccv2026-viq-text-aligned-visual-quantized-representations-at-any-resolution\/"},"modified":"2026-06-27T14:09:25","modified_gmt":"2026-06-27T06:09:25","slug":"eccv2026-viq-text-aligned-visual-quantized-representations-at-any-resolution","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/eccv2026-viq-text-aligned-visual-quantized-representations-at-any-resolution\/","title":{"rendered":"ViQ \u60f3\u628a\u5f71\u50cf\u8b8a\u6210\u66f4\u61c2\u8a9e\u610f\u7684\u96e2\u6563\u78bc"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/06\/hunyuan_logo.jpg\" alt=\"hunyuan logo\"><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">ViQ \u662f\u4e00\u500b\u8996\u89ba\u91cf\u5316\u8868\u793a\u7814\u7a76\u6846\u67b6\uff0c\u4e5f\u662f\u628a\u5f71\u50cf\u8f49\u6210\u96e2\u6563 codes \u7684\u6a21\u578b\u65b9\u6cd5\u3002\u5b83\u8981\u89e3\u6c7a\u7684\u554f\u984c\uff0c\u662f\u8b93\u5716\u7247\u50cf\u6587\u5b57 token \u4e00\u6a23\u53ef\u4ea4\u7d66\u591a\u6a21\u614b\u5927\u6a21\u578b\u8655\u7406\uff0c\u540c\u6642\u76e1\u91cf\u4e0d\u8981\u5728\u91cf\u5316\u904e\u7a0b\u4e1f\u5931\u592a\u591a\u8a9e\u610f\u8207\u756b\u9762\u7d30\u7bc0\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u73fe\u6709\u505a\u6cd5\u5e38\u898b\u5169\u689d\u8def\uff1a\u4e00\u985e\u504f\u91cd\u91cd\u5efa\uff0c\u9084\u539f\u756b\u9762\u80fd\u529b\u8f03\u597d\uff0c\u4f46\u8a9e\u610f\u8cc7\u8a0a\u4e0d\u8db3\uff1b\u53e6\u4e00\u985e\u4f9d\u8cf4 contrastive vision-language learning \u7684\u9023\u7e8c\u7279\u5fb5\uff0c\u8a9e\u610f\u8f03\u5f37\uff0c\u537b\u4e0d\u5bb9\u6613\u76f4\u63a5\u8b8a\u6210\u9ad8\u54c1\u8cea\u96e2\u6563\u8868\u793a\u3002ViQ \u7684\u5207\u5165\u9ede\u662f\u5148\u505a Text-Aligned Pre-training\uff0c\u518d\u505a\u91cf\u5316\u5b78\u7fd2\uff0c\u628a\u300c\u5148\u5c0d\u9f4a\u8a9e\u8a00\u8a9e\u610f\u3001\u5f8c\u9010\u6b65\u96e2\u6563\u5316\u300d\u62c6\u6210\u6e05\u695a\u5169\u6bb5\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5b83\u7684\u6838\u5fc3\u8a2d\u8a08\u6709\u5e7e\u500b\u8fa8\u8b58\u5ea6\u5f88\u9ad8\u7684\u90e8\u4ef6\uff1a\u4ee5 pretrained language model \u76e3\u7763\u8996\u89ba\u7de8\u78bc\u5668\u3001\u7528 resized positional embedding \u8207 native patchify \u652f\u63f4 any-resolution input\u3001\u518d\u7528 Proximal Representation Learning \u914d\u5408 L\u221e-norm \u7d04\u675f\uff0c\u628a\u7279\u5fb5\u9010\u6b65\u63a8\u8fd1\u91cf\u5316\u9328\u9ede\uff0c\u6700\u5f8c\u4ea4\u7d66 position-aware\u3001head-wise FSQ\uff08Finite Scalar Quantization\uff09\u8655\u7406\u3002\u8ad6\u6587\u4ea6\u63d0\u5230\u57fa\u5ea7\u53ef\u63a5 SigLIP2 vision tower\u3001Qwen2.5 backbone\uff0c\u4e26\u900f\u904e LoRA \u7b49\u8f15\u91cf\u7d44\u4ef6\u8a13\u7df4\u91cf\u5316\u90e8\u5206\uff0c\u800c\u4e0d\u662f\u5168\u9762\u5fae\u8abf\u6574\u500b\u7cfb\u7d71\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u652f\u63f4\u4efb\u610f\u89e3\u6790\u5ea6\u8f38\u5165\uff0c\u4e0d\u7528\u88ab\u56fa\u5b9a\u5c3a\u5bf8\u7d81\u6b7b<\/li>\n<li>\u76ee\u6a19\u4e0d\u662f\u53ea\u91cd\u5efa\u5716\u7247\uff0c\u800c\u662f\u517c\u9867\u8a9e\u610f\u7406\u89e3\u8207\u7d30\u7bc0<\/li>\n<li>\u591a\u6a21\u614b\u8a13\u7df4\u53ef\u76f4\u63a5\u5403\u96e2\u6563\u8996\u89ba codes\uff0c\u8ad6\u6587\u7a31\u6548\u7387\u53ef\u63d0\u5347\u7d04 20% \u81f3 70%<\/li>\n<li>\u5df2\u516c\u958b\u8a13\u7df4\u8207\u63a8\u8ad6\u7a0b\u5f0f\uff0c\u4e26\u63d0\u4f9b HuggingFace \u6b0a\u91cd<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u5f9e\u90e8\u7f72\u8207\u6e2c\u8a66\u89d2\u5ea6\u770b\uff0c\u9019\u500b GitHub \u5132\u5b58\u5eab\u8f03\u9069\u5408\u7576\u7814\u7a76\u5be6\u4f5c\u8207\u6a21\u578b\u9a57\u8b49\u9805\u76ee\u4f86\u7406\u89e3\uff1a\u53ef\u5148\u7528\u5df2\u516c\u958b\u6b0a\u91cd\u8dd1 inference\uff0c\u89c0\u5bdf\u5f71\u50cf\u5982\u4f55\u88ab\u7de8\u6210\u96e2\u6563 codes\uff0c\u518d\u9032\u4e00\u6b65\u91cd\u73fe\u55ae\u968e\u6bb5\u8a13\u7df4\u793a\u4f8b\uff0c\u4e4b\u5f8c\u624d\u5617\u8a66\u8ad6\u6587\u4e2d\u7684\u5169\u968e\u6bb5 recipe\u3002\u8f03\u53d7\u60e0\u7684\u6703\u662f\u505a MLLM\u3001\u8996\u89ba tokenization\u3001\u5f71\u50cf\u91cd\u5efa\u6216\u8a13\u7df4\u52a0\u901f\u7684\u5718\u968a\uff1b\u9650\u5236\u5247\u662f\u6982\u5ff5\u8207\u8a13\u7df4\u6d41\u7a0b\u90fd\u4e0d\u7b97\u8f15\uff0c\u91cd\u9ede\u8f03\u504f\u7814\u7a76\u50f9\u503c\uff0c\u672a\u5fc5\u662f\u5373\u88dd\u5373\u7528\u7684\u901a\u7528\u5de5\u5177\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>GitHub\uff1a<\/strong> <a href=\"https:\/\/github.com\/yuxumin\/ViQ\" rel=\"noopener noreferrer\" target=\"_blank\">https:\/\/github.com\/yuxumin\/ViQ<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Paper\uff1a<\/strong> <a href=\"https:\/\/arxiv.org\/pdf\/2606.27313\" rel=\"noopener noreferrer\" target=\"_blank\">https:\/\/arxiv.org\/pdf\/2606.27313<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>ViQ \u662f\u628a\u5f71\u50cf\u91cf\u5316\u6210\u96e2\u6563\u8868\u793a\u7684\u7814\u7a76\u6846\u67b6\uff0c\u76ee\u6a19\u662f\u5728\u8a9e\u610f\u7406\u89e3\u8207\u7d30\u7bc0\u4fdd\u7559\u4e4b\u9593\u53d6\u5f97\u5e73\u8861\u3002\u5b83\u4e5f\u652f\u63f4\u4efb\u610f\u89e3\u6790\u5ea6\u8f38\u5165\uff0c\u5c0d\u591a\u6a21\u614b\u8a13\u7df4\u6548\u7387\u6709\u660e\u986f\u5438\u5f15\u529b\u3002<\/p>\n","protected":false},"author":8,"featured_media":9723,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,185,172,131,119,76,127,149,196,197],"tags":[],"class_list":["post-9724","post","type-post","status-publish","format-standard","hentry","category-133","category-qwen","category-172","category-embedding","category-119","category-76","category-127","category-149","category-196","category-framework"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9724","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=9724"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9724\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/9723"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=9724"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=9724"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=9724"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}