{"id":10309,"date":"2026-07-17T18:32:33","date_gmt":"2026-07-17T10:32:33","guid":{"rendered":"https:\/\/infernews.com\/blog\/a-generalist-video-mllm-built-for-fine-grained-motion-long-form-reasoning-tempor\/"},"modified":"2026-07-17T18:32:33","modified_gmt":"2026-07-17T10:32:33","slug":"a-generalist-video-mllm-built-for-fine-grained-motion-long-form-reasoning-tempor","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/a-generalist-video-mllm-built-for-fine-grained-motion-long-form-reasoning-tempor\/","title":{"rendered":"VideoChat3 \u4e00\u500b\u7747\u5f97\u8010\u53c8\u7747\u5f97\u6e96\u7684\u5f71\u7247\u6a21\u578b"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/07\/pasted-ba3dc554afe7.jpg\" alt=\"VideoChat3 logo\"><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u5f71\u7247\u7406\u89e3\u6700\u9ebb\u7169\u7684\u5730\u65b9\uff0c\u5f80\u5f80\u5514\u4fc2\u300c\u8b58\u5514\u8b58\u7747\u300d\uff0c\u800c\u4fc2\u8981\u4e00\u908a\u4fdd\u7559\u52d5\u4f5c\u7d30\u7bc0\uff0c\u4e00\u908a\u6371\u5f97\u4f4f\u9577\u6642\u9593\u7247\u6bb5\u3002VideoChat3 \u5c31\u4fc2\u671d\u4f4f\u5462\u500b\u77db\u76fe\u843d\u624b\uff1a\u5b83\u5c6c\u65bc\u591a\u6a21\u614b\u6a21\u578b\uff08Multimodal Large Language Model, MLLM\uff09\uff0c\u76ee\u6a19\u4fc2\u7528\u540c\u4e00\u500b 4B \u6a21\u578b\u8655\u7406\u7d30\u5fae\u52d5\u4f5c\u3001\u9577\u7247\u63a8\u7406\u3001temporal grounding \u540c live streaming \u56de\u61c9\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u540c\u985e\u9805\u76ee\u597d\u591a\u6642\u53ea\u6703\u5c08\u6ce8\u5176\u4e2d\u4e00\u6bb5\u5de5\u4f5c\u6d41\uff0c\u4f8b\u5982\u77ed\u7247\u52d5\u4f5c\u8fa8\u8b58\uff0c\u6216\u8005\u9577\u7247\u554f\u7b54\u3002VideoChat3 \u7684\u53d6\u5411\u4fc2\u505a generalist video understanding\uff0c\u4ee3\u50f9\u5c31\u5514\u4fc2\u8ffd\u6c42\u55ae\u4e00\u5834\u666f\u6700\u6975\u81f4\u7684\u898f\u683c\uff0c\u800c\u4fc2\u7528 I3D-ViT \u540c Adaptive Frame Resolution \u5e73\u8861 token \u6210\u672c\u3001\u6642\u5e8f\u8b49\u64da\u540c\u5ef6\u9072\uff0c\u4ee4\u6a21\u578b\u5514\u9700\u8981\u5168\u7a0b\u7528\u9ad8\u6210\u672c\u65b9\u5f0f\u8b80\u5b8c\u6574\u6bb5\u5f71\u7247\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u91cd\u9ede\u5514\u53ea\u4fc2\u7747\u55ae\u683c\u756b\u9762\uff0c\u800c\u4fc2\u4fdd\u7559\u8de8\u6642\u9593\u7684\u8b49\u64da<\/li>\n<li>I3D-ViT \u63d0\u4f9b 16\u00d7 spatiotemporal compression\uff0c\u4e3b\u6253\u6548\u7387<\/li>\n<li>Adaptive Frame Resolution \u6703\u6309\u9700\u8981\u63d0\u9ad8\u756b\u9762\u89e3\u6790\u5ea6\uff0c\u8f03\u9069\u5408 streaming \u5834\u666f<\/li>\n<li>\u5df2\u516c\u958b model weights \u540c\u5b8c\u6574\u8a13\u7df4\u6578\u64da\uff0c\u4f46 training code \u4ecd\u672a\u91cb\u51fa<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u90e8\u7f72\u540c\u6e2c\u8a66\u7684\u7406\u89e3\u65b9\u5f0f\u5e7e\u76f4\u63a5\uff1a\u73fe\u968e\u6bb5\u8f03\u63a5\u8fd1\u7814\u7a76\u91cb\u51fa\u8207\u6a21\u578b\u9ad4\u9a57\uff0c\u9069\u5408\u5148\u7d93 Hugging Face \u53d6\u7528 models &amp; data\uff0c\u518d\u6309\u793a\u7bc4\u5834\u666f\u9a57\u8b49\u9577\u7247\u554f\u7b54\u3001\u6642\u9593\u5b9a\u4f4d\u540c\u4e32\u6d41\u56de\u61c9\u8868\u73fe\u3002README \u5df2\u5217\u660e\u5b8c\u6574\u8a13\u7df4\u8cc7\u6599\u5305\u62ec Academic2M\u3001LV116K\u3001OL617K\uff0c\u5c0d\u7814\u7a76\u5718\u968a\u3001\u505a video agent\u3001\u6216\u8005\u8981\u5efa\u69cb\u5f71\u7247\u6aa2\u7d22\u8207\u76e3\u5bdf\u6d41\u7a0b\u7684\u5718\u968a\u6700\u6709\u53c3\u8003\u50f9\u503c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u516c\u958b\u8cc7\u8a0a\u4ea6\u4ea4\u4ee3\u5497\u5e7e\u500b\u95dc\u9375\u6578\u5b57\uff1a4B parameters\u30013M curated instruction samples\u30012,048 frames \u4e0b\u7d04 20.4s latency\u3002\u5462\u5572\u6578\u64da\u672a\u5fc5\u4ee3\u8868\u6240\u6709\u74b0\u5883\u90fd\u6703\u6709\u540c\u6a23\u6548\u679c\uff0c\u4f46\u81f3\u5c11\u8b1b\u6e05\u695a\u5b83\u60f3\u8b49\u660e\u7684\u65b9\u5411\uff1a\u5514\u9760\u8d85\u5927\u6a21\u578b\uff0c\u90fd\u53ef\u4ee5\u628a\u5f71\u7247\u4e2d\u7684\u6642\u9593\u7dda\u7d22\u3001\u4e8b\u4ef6\u95dc\u806f\u540c\u5373\u6642\u53cd\u61c9\u653e\u5165\u540c\u4e00\u5957\u67b6\u69cb\u3002\u76f8\u95dc\u6a21\u578b\u8207\u6a21\u7d44\u5247\u4ee5 VideoChat3\u3001I3D-ViT\u3001Adaptive Frame Resolution \u70ba\u6838\u5fc3\uff0c\u6574\u9ad4\u66f4\u4f3c\u4e00\u500b\u9762\u5411\u7814\u7a76\u8207\u9032\u968e\u61c9\u7528\u7684\u958b\u6e90\u5f71\u7247\u7406\u89e3\u9805\u76ee\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/mcg-nju.github.io\/VideoChat3\/\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/MCG-NJU\/VideoChat3\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a> \u00b7 <a href=\"https:\/\/huggingface.co\/collections\/MCG-NJU\/videochat3\" rel=\"noopener noreferrer\"><strong>\u6a21\u578b<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u8981\u540c\u4e00\u500b\u6a21\u578b\u540c\u6642\u8655\u7406\u7d30\u5fae\u52d5\u4f5c\u3001\u9577\u7247\u7406\u89e3\u540c\u76f4\u64ad\u56de\u61c9\uff0c\u4e00\u5411\u597d\u96e3\u3002VideoChat3 \u628a\u91cd\u9ede\u653e\u5728\u6642\u5e8f\u7406\u89e3\u540c\u6548\u7387\uff0c\u5b9a\u4f4d\u76f8\u7576\u6e05\u6670\u3002<\/p>\n","protected":false},"author":8,"featured_media":10308,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,116,31,132,119,76,149,197,198],"tags":[],"class_list":["post-10309","post","type-post","status-publish","format-standard","hentry","category-133","category-agentic","category-video","category-3d","category-119","category-76","category-149","category-framework","category-198"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10309","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=10309"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10309\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/10308"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=10309"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=10309"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=10309"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}