{"id":9998,"date":"2026-07-08T17:12:03","date_gmt":"2026-07-08T09:12:03","guid":{"rendered":"https:\/\/infernews.com\/blog\/internvla-a1-unifying-understanding-generation-and-action-for-robotic-manipulati\/"},"modified":"2026-07-08T17:12:03","modified_gmt":"2026-07-08T09:12:03","slug":"internvla-a1-unifying-understanding-generation-and-action-for-robotic-manipulati","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/internvla-a1-unifying-understanding-generation-and-action-for-robotic-manipulati\/","title":{"rendered":"InternVLA-A1.5\uff1a\u6a5f\u68b0\u4eba\u7b56\u7565\u4e00\u9ad4\u5316\u65b0\u8def\u7dda"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/07\/pasted-592ede82cdd8.jpg\" alt=\"InternVLA-A1.5 teaser\"><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u6a5f\u68b0\u4eba\u64cd\u4f5c\u6700\u9ebb\u7169\u7684\u5730\u65b9\uff0c\u5f80\u5f80\u4e0d\u662f\u55ae\u7d14\u8fa8\u8a8d\u756b\u9762\uff0c\u800c\u662f\u8981\u540c\u6642\u7406\u89e3\u6307\u4ee4\u3001\u4f30\u8a08\u63a5\u4e0b\u4f86\u6703\u767c\u751f\u4ec0\u9ebc\uff0c\u518d\u7a69\u5b9a\u5730\u505a\u51fa\u9023\u7e8c\u52d5\u4f5c\u3002InternVLA-A1.5\u5c6c\u65bc<strong>\u958b\u6e90\u6846\u67b6\u517c\u6a5f\u68b0\u4eba\u653f\u7b56\u6a21\u578b<\/strong>\uff0c\u7126\u9ede\u653e\u5728\u628a vision-language understanding\u3001latent visual foresight \u8207 action generation \u5408\u4f75\uff0c\u6e1b\u5c11\u591a\u6a21\u7d44\u4e32\u63a5\u5e36\u4f86\u7684\u5ef6\u9072\u8207\u5354\u8abf\u6210\u672c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5b83\u7684\u53d6\u5411\u5f88\u6e05\u695a\uff1a\u4e0d\u5c11\u505a\u6cd5\u6703\u628a\u611f\u77e5\u3001\u672a\u4f86\u9810\u6e2c\u3001\u63a7\u5236\u5206\u958b\u8a13\u7df4\u6216\u5206\u958b\u90e8\u7f72\uff0cInternVLA-A1.5\u5247\u628a foresight \u653e\u9032\u540c\u4e00\u689d\u653f\u7b56\u8def\u5f91\uff0c\u5728\u8a13\u7df4\u671f\u9593\u501f\u52a9\u51cd\u7d50\u7684 WAN2.2-5B video generation model \u63d0\u4f9b\u672a\u4f86\u52d5\u614b\u76e3\u7763\uff0c\u4f46\u63a8\u7406\u6642\u4e1f\u68c4 video branch\uff0c\u53ea\u4fdd\u7559\u52d5\u4f5c\u9810\u6e2c\u3002\u9019\u500b\u8a2d\u8a08\u7684\u597d\u8655\u662f\u4fdd\u4f4f\u300c\u5148\u60f3\u4e00\u6b65\u300d\u7684\u80fd\u529b\uff0c\u540c\u6642\u907f\u514d\u90e8\u7f72\u5230\u771f\u5be6\u6a5f\u68b0\u4eba\u6642\u63a8\u7406\u592a\u91cd\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u6a21\u578b\u9aa8\u5e79\u5efa\u57fa\u65bc Qwen3.5-2B VLM\uff0c\u900f\u904e shared full-attention layers \u63a5\u4e0a\u4e00\u500b\u8f15\u91cf unified action expert\uff0c\u4e26\u4fdd\u7559 modality-specific Gated DeltaNet processing\uff1b\u52d5\u4f5c\u8f38\u51fa\u5247\u7528 flow matching \u9810\u6e2c continuous action chunks\u3002README \u4ea6\u63d0\u5230\u5b83\u53ef\u7528\u65bc\u90e8\u7f72\u3001\u6578\u64da\u6536\u96c6\u548c\u8a55\u4f30\u771f\u5be6\u6a5f\u5668\u4eba\u4e0a\u8a13\u7df4\u6709\u7d20\u7684\u64cd\u4f5c\u7b56\u7565\uff0c\u4f46\u516c\u958b\u8cc7\u8a0a\u8f03\u96c6\u4e2d\u5728\u6a21\u578b\u8207 benchmark\uff0c\u5177\u9ad4\u5b89\u88dd\u6d41\u7a0b\u8207\u771f\u6a5f\u90e8\u7f72\u6b65\u9a5f\u672a\u898b\u5b8c\u6574\u5c55\u958b\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u820a\u8def\u7dda\u5e38\u628a\u7406\u89e3\u3001\u9810\u6e2c\u3001\u63a7\u5236\u62c6\u958b\uff0c\u9019\u500b\u9805\u76ee\u6539\u70ba\u55ae\u4e00 policy \u7d71\u4e00\u8655\u7406<\/li>\n<li>\u8a13\u7df4\u7528 WAN2.2-5B \u5b78\u672a\u4f86\u52d5\u614b\uff0c\u63a8\u7406\u6642\u79fb\u9664\u76f8\u95dc\u5206\u652f\uff0c\u63db\u53d6\u8f03\u5be6\u969b\u5ef6\u9072<\/li>\n<li>\u5df2\u5728 LeRobot V2.1 dataset \u5fae\u8abf\uff0c\u4ea6\u7d50\u5408\u5927\u898f\u6a21\u6a5f\u68b0\u4eba\u8207\u591a\u6a21\u614b\u8cc7\u6599<\/li>\n<li>\u57fa\u6e96\u6210\u7e3e\u7a81\u51fa\uff1aRoboTwin 2.0 \u70ba 93.2\uff0cLIBERO \u70ba 98.9\uff0cLIBERO-Plus \u70ba 84.8<\/li>\n<li>\u53ef\u53d6\u5f97\u7684\u76f8\u95dc\u6a21\u578b\u5305\u62ec InternVLA-A1.5-base\u3001InternVLA-A1.5-RoboTwin\u3001InternVLA-A1.5-Libero<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u5f9e\u5b9a\u4f4d\u4f86\u770b\uff0c\u5b83\u8f03\u9069\u5408\u60f3\u628a\u7814\u7a76\u539f\u578b\u63a8\u8fd1\u771f\u6a5f\u9a57\u8b49\u7684\u5718\u968a\uff0c\u5c24\u5176\u662f\u540c\u6642\u91cd\u8996\u8a9e\u8a00\u7406\u89e3\u3001\u8996\u89ba\u6cdb\u5316\u8207\u64cd\u4f5c\u6210\u529f\u7387\u7684\u4eba\u3002\u73fe\u968e\u6bb5\u6700\u503c\u5f97\u7559\u610f\u7684\uff0c\u4e0d\u53ea\u662f\u5206\u6578\u9ad8\uff0c\u800c\u662f\u5b83\u793a\u7bc4\u4e86\u4e00\u7a2e\u66f4\u63a5\u8fd1\u5b8c\u6574\u6a5f\u68b0\u4eba\u5de5\u4f5c\u6d41\u7684\u6574\u5408\u65b9\u5f0f\uff1b\u9650\u5236\u5247\u662f\u516c\u958b\u8aaa\u660e\u4ecd\u504f\u7814\u7a76\u5c0e\u5411\uff0c\u771f\u6b63\u843d\u5730\u524d\u4ecd\u9700\u81ea\u884c\u88dc\u8db3\u90e8\u7f72\u7d30\u7bc0\u8207\u786c\u4ef6\u6574\u5408\u8cc7\u8a0a\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/internrobotics.github.io\/internvla-a15.github.io\/\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/InternRobotics\/InternVLA-A-series\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a> \u00b7 <a href=\"https:\/\/huggingface.co\/collections\/InternRobotics\/internvla-a15\" rel=\"noopener noreferrer\"><strong>\u6a21\u578b<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u60f3\u50cf\u540c\u4e00\u500b\u6a21\u578b\u65e2\u7747\u5f97\u660e\u756b\u9762\u8207\u6307\u4ee4\uff0c\u53c8\u6703\u9810\u60f3\u4e0b\u4e00\u6b65\u5834\u666f\uff0c\u518d\u76f4\u63a5\u8f38\u51fa\u52d5\u4f5c\u3002InternVLA-A1.5\u5438\u5f15\u4e4b\u8655\uff0c\u6b63\u6b63\u5728\u65bc\u5b83\u628a\u9019\u4e09\u4ef6\u4e8b\u6536\u5728\u540c\u4e00\u5957\u6a5f\u68b0\u4eba\u653f\u7b56\u88e1\u3002<\/p>\n","protected":false},"author":8,"featured_media":9997,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,185,31,119,149,141,184,199,200,204],"tags":[],"class_list":["post-9998","post","type-post","status-publish","format-standard","hentry","category-133","category-qwen","category-video","category-119","category-149","category-141","category-robotic","category-dataset-","category-200","category-visionlanguageaction"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9998","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=9998"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9998\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/9997"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=9998"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=9998"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=9998"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}