{"id":9137,"date":"2026-06-16T01:00:10","date_gmt":"2026-06-15T17:00:10","guid":{"rendered":"https:\/\/infernews.com\/blog\/a-dataset-for-audio-visual-reasoning-through-structured-scripts-and-evidence-cha\/"},"modified":"2026-06-16T03:32:05","modified_gmt":"2026-06-15T19:32:05","slug":"a-dataset-for-audio-visual-reasoning-through-structured-scripts-and-evidence-cha","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/a-dataset-for-audio-visual-reasoning-through-structured-scripts-and-evidence-cha\/","title":{"rendered":"OmniVideo-100K\uff1a\u589e\u5f37\u5f71\u97f3\u63a8\u7406\u8a13\u7df4\u6578\u64da\u96c6"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/06\/pasted-3768c9f3f3a9.jpg\" alt=\"Framework Overview\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u73fe\u6642\u4e0d\u5c11\u5f71\u97f3\u554f\u7b54\u8cc7\u6599\u5efa\u7acb\u6d41\u7a0b\uff0c\u666e\u904d\u6cbf\u7528\u300cvideo-caption-QA\u300d\u7bc4\u5f0f\uff1a\u5148\u628a\u5f71\u7247\u5207\u6210\u77ed\u7247\u6bb5\uff0c\u518d\u5206\u958b\u5beb\u8996\u89ba\u8207\u97f3\u8a0a\u63cf\u8ff0\uff0c\u6700\u5f8c\u751f\u6210 QA\u3002\u8ad6\u6587\u4f5c\u8005\u8a8d\u70ba\u9019\u7a2e\u505a\u6cd5\u5bb9\u6613\u51fa\u73fe modality bias\u3001temporal misalignment\uff0c\u540c\u4e00\u89d2\u8272\u5728\u4e0d\u540c\u7247\u6bb5\u4ea6\u53ef\u80fd\u63cf\u8ff0\u4e0d\u4e00\u81f4\uff0c\u4ee4\u554f\u984c\u591a\u6578\u53ea\u570d\u7e5e\u5c40\u90e8\u4e8b\u4ef6\uff0c\u96e3\u4ee5\u8003\u9a57\u9577\u6642\u9593\u8de8\u5ea6\u7684 audio-visual reasoning\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">OmniVideo-100K \u662f\u4e00\u500b Dataset \u6578\u64da\u96c6\uff0c\u76ee\u6a19\u662f\u70ba Multimodal Large Language Models (MLLMs) \u63d0\u4f9b\u8f03\u5b8c\u6574\u7684\u5f71\u97f3\u63a8\u7406\u8a13\u7df4\u6750\u6599\u3002\u5b83\u63d0\u51fa\u5169\u500b\u6838\u5fc3\u6a5f\u5236\uff1aEntity-Anchored Video Scripting \u5148\u628a\u539f\u59cb\u5f71\u7247\u6574\u7406\u6210\u7d50\u69cb\u5316 script\uff0c\u5305\u62ec\u6458\u8981\u3001\u4e3b\u8981\u5be6\u9ad4\u6e05\u55ae\uff0c\u4ee5\u53ca\u5e36\u6642\u9593\u6233\u7684\u5206\u6bb5\u97f3\u756b\u63cf\u8ff0\uff1bClue-Guided QA Generation \u5247\u5148\u62bd\u53d6\u8de8\u7247\u6bb5\u3001\u8de8\u6a21\u614b\u7dda\u7d22\uff0c\u518d\u751f\u6210\u8f03\u8907\u96dc\u7684\u554f\u7b54\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u9019\u7a2e\u8a2d\u8a08\u7684\u95dc\u9375\uff0c\u4e0d\u5728\u65bc\u984c\u76ee\u6578\u91cf\u5927\uff0c\u800c\u5728\u65bc\u5148\u6574\u7406\u8b49\u64da\u93c8\u518d\u51fa\u984c\u3002\u5c0d\u6bd4\u820a\u65b9\u6cd5\u628a\u9577\u6587\u672c\u7406\u89e3\u548c QA \u5408\u4f75\u6210\u4e00\u6b65\uff0cOmniVideo-100K \u628a\u7dda\u7d22\u6316\u6398\u7368\u7acb\u51fa\u4f86\uff0c\u8f03\u6709\u6a5f\u6703\u7522\u751f\u6d89\u53ca\u56e0\u679c\u3001\u672a\u4f86\u9810\u6e2c\u8207\u5047\u8a2d\u63a8\u7406\u7684\u984c\u76ee\uff0c\u800c\u4e0d\u53ea\u662f\u554f\u756b\u9762\u4e2d\u300c\u898b\u5230\u4e5c\u300d\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5982\u679c\u4f60\u60f3\u6e2c\u8a66\u9019\u500b\u9805\u76ee\uff0c\u53ef\u4ee5\u5148\u7559\u610f Hugging Face \u4e0a\u7684 OmniVideo-100K \u8207\u4eba\u624b\u9a57\u8b49\u7684 OmniVideo-Test\uff0c\u770b\u770b\u8cc7\u6599\u7d50\u69cb\u662f\u5426\u9069\u5408\u81ea\u5df1\u7684\u8a13\u7df4\u6d41\u7a0b\uff1b\u505a\u7814\u7a76\u7684\u4eba\u5247\u53ef\u76f4\u63a5\u6bd4\u8f03\u6a21\u578b\u5728\u5916\u90e8 benchmark \u7684\u8b8a\u5316\u3002\u8ad6\u6587\u63d0\u5230\uff0cVITA-1.5\u3001Qwen2.5-Omni-7B\u3001Qwen3-Omni-30B \u7d93\u6b64\u8cc7\u6599\u96c6\u5fae\u8abf\u5f8c\uff0c\u5728 OmniVideo-Test \u6700\u591a\u63d0\u5347 20.59%\uff0c\u5728 Daily-Omni\u3001JointAVBench \u4e5f\u6709\u6700\u591a 12.64% \u589e\u5e45\uff0c\u540c\u6642\u76e1\u91cf\u4fdd\u7559\u5728 Video-MME \u9019\u985e\u4e00\u822c\u5f71\u7247 benchmark \u7684\u80fd\u529b\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u91dd\u5c0d\u820a\u5f0f\u300cvideo-caption-QA\u300d\u6d41\u7a0b\u7684\u4e09\u500b\u75db\u9ede\uff1amodality bias\u3001temporal misalignment\u3001\u6558\u4e8b\u4e0d\u9023\u8cab<\/li>\n\n\n\n<li>\u7528 structured scripts \u52a0 entity list\uff0c\u88dc\u56de\u8de8\u7247\u6bb5\u6307\u4ee3\u4e00\u81f4\u6027\u8207\u8072\u97f3\u4f86\u6e90\u5c0d\u61c9<\/li>\n\n\n\n<li>\u4efb\u52d9\u8986\u84cb 10 \u985e\uff0c\u5305\u62ec FGP\u3001STD\u3001CU\u3001CP\u3001SA\u3001ESO\u3001SM\u3001CR\u3001FP\u3001HR<\/li>\n\n\n\n<li>\u9069\u5408\u505a\u5f71\u97f3\u7406\u89e3\u3001\u8de8\u6a21\u614b\u63a8\u7406\u3001\u6307\u4ee4\u5fae\u8abf\u8cc7\u6599\u7814\u7a76\u7684\u4eba\u53c3\u8003<\/li>\n\n\n\n<li>\u76f8\u95dc\u6a21\u578b\u5305\u62ec VITA-1.5\u3001Qwen2.5-Omni-7B\u3001Qwen3-Omni-30B<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u6574\u9ad4\u4f86\u770b\uff0c\u9019\u500b\u9805\u76ee\u7684\u50f9\u503c\u5728\u65bc\u5b83\u4e0d\u53ea\u65b0\u589e\u4e00\u6279\u8cc7\u6599\uff0c\u800c\u662f\u91cd\u5beb\u5f71\u97f3 QA \u8cc7\u6599\u7684\u7d44\u7e54\u65b9\u6cd5\u3002\u82e5\u4f60\u95dc\u5fc3\u7684\u4e0d\u662f\u55ae\u5f35\u756b\u9762\u554f\u7b54\uff0c\u800c\u662f\u5f71\u7247\u5167\u89d2\u8272\u3001\u8072\u97f3\u3001\u4e8b\u4ef6\u5148\u5f8c\u8207\u63a8\u8ad6\u4e4b\u9593\u7684\u9023\u7d50\uff0cOmniVideo-100K \u6703\u6bd4\u4e00\u822c\u81ea\u52d5\u5408\u6210\u8cc7\u6599\u96c6\u66f4\u6709\u7814\u7a76\u53c3\u8003\u50f9\u503c\u3002<\/p>\n\n\n\n<div class=\"wp-block-group is-layout-constrained wp-block-group-is-layout-constrained\">\n<p class=\"wp-block-paragraph\"><strong>GitHub\uff1a<\/strong> <a href=\"https:\/\/github.com\/MiG-NJU\/OmniVideo-100K\" rel=\"noopener noreferrer\">https:\/\/github.com\/MiG-NJU\/OmniVideo-100K<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u9805\u76ee\uff1a<\/strong><a href=\"https:\/\/yzlmhzz.github.io\/OmniVideo-100K\/\" rel=\"noopener\">https:\/\/yzlmhzz.github.io\/OmniVideo-100K\/<\/a><\/p>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>\u9019\u662f\u4e00\u500b\u9762\u5411\u97f3\u8a0a\u8207\u5f71\u7247\u806f\u5408\u7406\u89e3\u7684\u6578\u64da\u96c6\u9805\u76ee\uff0c\u91cd\u9ede\u4e0d\u5728\u7247\u6bb5\u6a19\u8a3b\uff0c\u800c\u5728\u8f03\u5b8c\u6574\u7684\u8de8\u6a21\u614b\u63a8\u7406\u3002\u5b83\u66f4\u50cf\u662f\u5e6b\u6a21\u578b\u88dc\u8ab2\u7684\u8a13\u7df4\u6750\u6599\u5eab\u3002<\/p>\n","protected":false},"author":8,"featured_media":9136,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,185,31,29,38,162,119,150,76,127,128,199],"tags":[],"class_list":["post-9137","post","type-post","status-publish","format-standard","hentry","category-133","category-qwen","category-video","category-audio","category-38","category-ai-productions","category-119","category-150","category-76","category-127","category-128","category-dataset-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9137","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=9137"}],"version-history":[{"count":3,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9137\/revisions"}],"predecessor-version":[{"id":9143,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9137\/revisions\/9143"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/9136"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=9137"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=9137"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=9137"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}