{"id":8615,"date":"2026-05-30T08:52:27","date_gmt":"2026-05-30T00:52:27","guid":{"rendered":"https:\/\/infernews.com\/blog\/qwen-vla\/"},"modified":"2026-05-30T08:52:27","modified_gmt":"2026-05-30T00:52:27","slug":"qwen-vla","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/qwen-vla\/","title":{"rendered":"Qwen-VLA \u7528\u55ae\u4e00\u6a21\u578b\u6253\u901a\u6a5f\u68b0\u4eba\u4efb\u52d9"},"content":{"rendered":"<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/05\/pasted-4ea98b3ebabf.jpg\" alt=\"Hero image preview\"><\/figure>\n<p>Qwen-VLA \u662f\u4e00\u500b\u628a\u8996\u89ba\u3001\u8a9e\u8a00\u8207\u52d5\u4f5c\u6574\u5408\u8d77\u4f86\u7684\u9805\u76ee\uff0c\u76ee\u6a19\u662f\u7528\u540c\u4e00\u5957\u6a21\u578b\u8655\u7406\u6a5f\u68b0\u4eba\u64cd\u4f5c\u3001\u5c0e\u822a\uff0c\u4ee5\u53ca\u8ecc\u8de1\u9810\u6e2c\u3002\u904e\u5f80\u5f88\u591a embodied intelligence \u7cfb\u7d71\u90fd\u53ea\u91dd\u5c0d\u55ae\u4e00\u4efb\u52d9\u6216\u55ae\u4e00\u6a5f\u68b0\u4eba\u8a2d\u8a08\uff0c\u7d50\u679c\u662f\u80fd\u529b\u5206\u6563\uff0c\u63db\u5834\u666f\u3001\u63db\u4efb\u52d9\u6216\u63db\u6a5f\u9ad4\u5f8c\u4fbf\u96e3\u4ee5\u6cbf\u7528\u3002<\/p>\n<p>\u9019\u500b\u9805\u76ee\u4ee5 Qwen3.5 Vision-Language Model \u70ba\u57fa\u790e\uff0c\u518d\u52a0\u5165 DiT-based action decoder\uff0c\u4ee4\u6a21\u578b\u4e0d\u53ea\u770b\u5716\u548c\u7406\u89e3\u6587\u5b57\uff0c\u9084\u80fd\u7522\u751f\u9023\u7e8c\u52d5\u4f5c\u8207\u8ecc\u8de1\u3002\u7814\u7a76\u5718\u968a\u4ea6\u52a0\u5165 embodiment-aware prompt conditioning\uff0c\u7528\u6a5f\u68b0\u4eba\u5c08\u5c6c\u6587\u5b57\u63cf\u8ff0\u53bb\u6a19\u793a\u7576\u524d\u6a5f\u9ad4\u8207\u63a7\u5236\u65b9\u5f0f\uff0c\u8b93\u540c\u4e00\u6a21\u578b\u53ef\u652f\u63f4\u591a\u7a2e robot embodiments\u3002<\/p>\n<p>\u6838\u5fc3\u6982\u5ff5\u662f\u628a\u4e0d\u540c\u985e\u578b\u8cc7\u6599\u4e00\u540c\u8a13\u7df4\uff0c\u5305\u62ec robotics manipulation trajectories\u3001human egocentric demonstrations\u3001synthetic simulation data\u3001vision-and-language navigation data\uff0c\u4ee5\u53ca auxiliary vision-language data\u3002\u5c0d\u958b\u767c\u8005\u4f86\u8aaa\uff0c\u9019\u4ee3\u8868\u9805\u76ee\u4e0d\u662f\u53ea\u505a\u554f\u7b54\u6216\u53ea\u505a\u63a7\u5236\uff0c\u800c\u662f\u628a manipulation\u3001navigation \u8207 trajectory prediction \u653e\u9032\u540c\u4e00\u500b action-and-trajectory prediction \u6846\u67b6\u3002<\/p>\n<p>\u91cd\u9ede\u53ef\u5148\u770b\u4ee5\u4e0b\u5e7e\u9805\uff1a<br \/>\n&#8211; \u7528\u55ae\u4e00 Vision-Language-Action Model \u8655\u7406\u591a\u985e embodied \u4efb\u52d9<br \/>\n&#8211; \u540c\u6642\u652f\u63f4\u6587\u5b57\u56de\u61c9\u3001\u52d5\u4f5c\u751f\u6210\u8207\u8ecc\u8de1\u751f\u6210<br \/>\n&#8211; \u900f\u904e embodiment-aware prompt conditioning \u9069\u914d\u4e0d\u540c\u6a5f\u68b0\u4eba\u5e73\u53f0<br \/>\n&#8211; \u5f37\u8abf out-of-distribution generalization\uff0c\u6db5\u84cb\u5149\u7dda\u3001\u80cc\u666f\u3001\u7269\u4ef6\u914d\u7f6e\u8207\u6a5f\u9ad4\u8b8a\u5316<br \/>\n&#8211; \u9069\u5408\u7814\u7a76\u901a\u7528\u6a5f\u68b0\u4eba\u63a7\u5236\u3001\u8de8\u4efb\u52d9\u9077\u79fb\u8207\u591a\u5e73\u53f0\u90e8\u7f72\u7684\u4eba\u54e1<\/p>\n<p>\u8ad6\u6587\u5217\u51fa Qwen-VLA-Instruct \u5728\u591a\u500b\u57fa\u6e96\u4e0a\u6709\u4e0d\u932f\u8868\u73fe\uff0c\u5305\u62ec LIBERO 97.9%\u3001Simpler-WidowX 73.7%\u3001RoboTwin-Easy\/Hard 86.1\/87.2%\u3001R2R \u7684 69.0% OSR\u3001RxR \u7684 59.6% SR\uff1b\u5728\u771f\u5be6\u4e16\u754c ALOHA \u5be6\u9a57\u4ea6\u9304\u5f97 76.9% average OOD success\uff0c\u5728 DOMINO dynamic manipulation \u6709 26.6% zero-shot success rate\u3002\u82e5\u4f60\u95dc\u6ce8\u7684\u662f\u4e00\u500b\u53ef\u6a6b\u8de8\u4efb\u52d9\u3001\u74b0\u5883\u8207 robot embodiments \u7684\u901a\u7528\u578b\u9805\u76ee\uff0c\u5c55\u793a\u4e86\u7d71\u4e00\u5f0f Vision-Language-Action Modeling \u7684\u6e05\u6670\u65b9\u5411\u3002<\/p>\n<p><strong>Paper\uff1a<\/strong> <a href=\"https:\/\/arxiv.org\/pdf\/2605.30280\" rel=\"noopener noreferrer\">https:\/\/arxiv.org\/pdf\/2605.30280<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Qwen-VLA\u628a\u64cd\u4f5c\u3001\u5c0e\u822a\u8207\u8ecc\u8de1\u9810\u6e2c\u653e\u5165\u540c\u4e00\u500b\u6a21\u578b\u3002\u91cd\u9ede\u5728\u8de8\u6a5f\u68b0\u4eba\u5e73\u53f0\u8207\u8de8\u5834\u666f\u6cdb\u5316\u80fd\u529b\u3002<\/p>\n","protected":false},"author":8,"featured_media":8614,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","wpai_meta_description":"","footnotes":""},"categories":[171,116,76,186,189],"tags":[],"class_list":["post-8615","post","type-post","status-publish","format-standard","hentry","category-171","category-agentic","category-76","category-186","category-189"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8615","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=8615"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8615\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/8614"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=8615"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=8615"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=8615"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}