{"id":10032,"date":"2026-07-08T19:44:47","date_gmt":"2026-07-08T11:44:47","guid":{"rendered":"https:\/\/infernews.com\/blog\/official-code-for-musebench-benchmarking-intent-level-audiovisual-arts-understan\/"},"modified":"2026-07-08T19:44:47","modified_gmt":"2026-07-08T11:44:47","slug":"official-code-for-musebench-benchmarking-intent-level-audiovisual-arts-understan","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/official-code-for-musebench-benchmarking-intent-level-audiovisual-arts-understan\/","title":{"rendered":"MuseBench \u7528\u85dd\u8853\u7406\u89e3\u8003\u9a57 MLLMs"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/07\/pasted-3d4b1d4191ea.jpg\" alt=\"Repository image for musebench\/musebench-code\"><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u898b\u5230\u756b\u9762\u3001\u807d\u5230\u8072\u97f3\uff0c\u672a\u5fc5\u7b49\u65bc\u771f\u4fc2\u660e\u767d\u4f5c\u54c1\u60f3\u9ede\u8b1b\u3002MuseBench \u628a\u7126\u9ede\u653e\u5230 artistic intent\uff0c\u5c08\u9580\u6e2c multimodal large language models\uff08MLLMs\uff09\u80fd\u5426\u7531\u8996\u807d\u8b49\u64da\u63a8\u65b7\u5275\u4f5c\u9078\u64c7\u80cc\u5f8c\u7684\u610f\u601d\uff1b\u5b83\u5c6c\u65bc benchmark\uff0f\u6578\u64da\u96c6\u578b\u9805\u76ee\uff0c\u8655\u7406\u7684\u662f\u73fe\u6709\u8a55\u6e2c\u591a\u6578\u53ea\u505c\u7559\u5728 perceptual recognition\uff0c\u672a\u80fd\u53cd\u6620\u85dd\u8853\u7406\u89e3\u6df1\u5ea6\u7684\u554f\u984c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u73fe\u6709\u505a\u6cd5\u5e38\u7528\u4e00\u822c\u8996\u89ba\u554f\u7b54\u6216\u5f71\u7247\u7406\u89e3\u984c\uff0c\u6a21\u578b\u53ea\u8981\u8fa8\u8a8d\u7269\u4ef6\u3001\u60c5\u7bc0\u6216\u8868\u9762\u4e8b\u4ef6\u5c31\u6709\u6a5f\u6703\u5f97\u5206\uff1b\u4f5c\u8005\u8a8d\u70ba\u9019\u7a2e fixed paradigm \u5ffd\u7565 stylistic vocabulary\u3001cultural priors \u540c grounded audiovisual inference\uff0c\u6240\u4ee5\u6539\u7528 narrator-removed video clip\uff0c\u4e26\u914d\u5408\u53ef\u9078 audio transcript\uff0c\u8feb\u4f7f\u6a21\u578b\u76f4\u63a5\u7531\u93e1\u982d\u3001\u8072\u97f3\u3001\u7bc0\u594f\u8207\u6558\u4e8b\u7dda\u7d22\u4f5c\u5224\u65b7\u3002\u984c\u76ee\u8986\u84cb Cinematic Arts\u3001Static Visual Arts\u3001Stage Performing Arts \u540c Game Arts\uff0c\u5408\u5171 4,016 \u689d\u554f\u7b54\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u540c\u985e benchmark \u591a\u6578\u8457\u91cd\u300c\u7747\u5230\u4e5c\u300d\uff0cMuseBench \u5247\u66f4\u5728\u610f\u300c\u9ede\u89e3\u8981\u5481\u5448\u73fe\u300d\u3002\u5b83\u4ea6\u5514\u53ea\u7528\u55ae\u4e00\u9078\u64c7\u984c\uff0c\u4ef2\u6709 single-select \u540c multi-select \u5169\u7a2e\u683c\u5f0f\uff0c\u4e26\u52a0\u5165 Chance-Adjusted Accuracy\uff08CAA\uff09\u8655\u7406\u9078\u9805\u6578\u91cf\u4e0d\u540c\u5e36\u4f86\u7684\u504f\u5dee\uff0c\u4ee4\u6bd4\u8f03 28 \u500b MLLMs \u6642\u8f03\u516c\u5e73\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u6db5\u84cb 4 \u500b\u85dd\u8853\u9818\u57df\u300111 \u500b\u7d30\u5206\u985e\uff0c\u984c\u6750\u6bd4\u4e00\u822c\u5f71\u7247 QA \u66f4\u95ca<\/li>\n<li>\u8a55\u6e2c 28 \u500b MLLMs\uff0c\u5305\u542b proprietary\u3001open source \u540c video-specific \u8def\u7dda<\/li>\n<li>\u6700\u4f73\u6a21\u578b\u6e96\u78ba\u7387 48.29%\uff0c\u660e\u986f\u4f4e\u65bc human expert \u7684 87.18%<\/li>\n<li>\u5df2\u6574\u5408 VLMEvalKit\uff0c\u65b9\u4fbf\u628a\u65b0\u6a21\u578b\u63a5\u5165\u540c\u4e00\u5957\u6d41\u7a0b\u6e2c\u8a66<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u90e8\u7f72\u540c\u6e2c\u8a66\u7406\u89e3\u4e0a\uff0c\u9019\u500b code repository \u4e3b\u8981\u5514\u4fc2\u63d0\u4f9b\u8a13\u7df4\u6a21\u578b\uff0c\u800c\u662f\u628a MuseBench \u63a5\u5230 VLMEvalKit \u7684\u8a55\u6e2c\u6d41\u7a0b\uff0c\u8f03\u9069\u5408\u7814\u7a76\u5718\u968a\u3001\u6a21\u578b\u8a55\u4f30\u4eba\u54e1\u3001\u505a video understanding \u6216\u591a\u6a21\u614b\u63a8\u7406\u7684\u9805\u76ee\u76f4\u63a5\u6bd4\u8f03\u65b0\u820a\u6a21\u578b\u3002\u5df2\u516c\u958b\u7684\u7d50\u679c\u63d0\u5230 Claude-4.6-Opus\u3001Qwen-3.5-Plus\u3001Doubao-Seed\u3001GPT-5.4\u3001Gemini-3.1-Pro\u3001Grok-4.1 \u7b49\u90fd\u6e2c\u904e\uff0c\u5206\u6578\u6574\u9ad4\u4ecd\u8207\u5c08\u5bb6\u6709\u5927\u6bb5\u8ddd\u96e2\uff1b\u63db\u53e5\u8a71\u8aaa\uff0c\u9019\u500b\u9805\u76ee\u6700\u6709\u50f9\u503c\u7684\u5730\u65b9\uff0c\u5728\u65bc\u5b83\u6e05\u695a\u6307\u51fa\u73fe\u6642 MLLMs \u5728\u85dd\u8853\u5224\u8b80\u4ecd\u672a\u7b97\u63a5\u8fd1\u53ef\u9760\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/musebench.github.io\/\" rel=\"noopener noreferrer\" target=\"_blank\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/musebench\/musebench-code\" rel=\"noopener noreferrer\" target=\"_blank\"><strong>GitHub<\/strong><\/a> \u00b7 <a href=\"https:\/\/arxiv.org\/pdf\/2606.30026\" rel=\"noopener noreferrer\" target=\"_blank\"><strong>Paper<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>MuseBench\u5514\u4fc2\u6e2c\u6a21\u578b\u898b\u5230\u54a9\uff0c\u800c\u4fc2\u8ffd\u554f\u9ede\u89e3\u4f5c\u54c1\u8981\u5481\u8868\u9054\u3002\u5462\u5957 benchmark \u628a\u591a\u6a21\u614b\u7406\u89e3\u62c9\u5230\u85dd\u8853\u8a9e\u5883\uff0c\u96e3\u5ea6\u660e\u986f\u9ad8\u904e\u4e00\u822c\u8996\u807d\u554f\u7b54\u3002<\/p>\n","protected":false},"author":8,"featured_media":10031,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,185,163,169,176,140,153,31,29,119,191,199],"tags":[],"class_list":["post-10032","post","type-post","status-publish","format-standard","hentry","category-133","category-qwen","category-163","category-169","category-176","category-gemini","category-openai","category-video","category-audio","category-119","category-anthropic","category-dataset-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10032","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=10032"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10032\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/10031"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=10032"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=10032"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=10032"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}