{"id":13121,"date":"2026-10-09T23:29:59","date_gmt":"2026-10-09T15:29:59","guid":{"rendered":"https:\/\/infernews.com\/blog\/official-code-of-uniwam-unified-world-action-model\/"},"modified":"2026-10-09T23:29:59","modified_gmt":"2026-10-09T15:29:59","slug":"official-code-of-uniwam-unified-world-action-model","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/official-code-of-uniwam-unified-world-action-model\/","title":{"rendered":"UniWAM \u628a\u7406\u89e3\u3001\u9810\u6e2c\u8207\u6a5f\u68b0\u4eba\u52d5\u4f5c\u5408\u800c\u70ba\u4e00"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/10\/UniWAM.jpg\" alt=\"UniWAM teaser\"><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u9762\u5c0d\u9700\u8981\u7406\u89e3\u6307\u4ee4\u3001\u9810\u6e2c\u74b0\u5883\u8b8a\u5316\u518d\u63a7\u5236\u6a5f\u68b0\u4eba\u7684\u4efb\u52d9\uff0cUniWAM \u5c07\u4e09\u500b\u6b65\u9a5f\u653e\u9032\u540c\u4e00\u500b 8B-parameter \u6a21\u578b\uff0c\u5c6c\u65bc\u7d50\u5408\u4e16\u754c\u6a21\u578b\u8207 Vision-Language-Action\uff08VLA\uff09\u7684\u958b\u6e90\u7814\u7a76\u9805\u76ee\u3002\u5b83\u4ee5 Mixture-of-Transformers\uff08MoT\uff09\u9023\u63a5 physical reasoner\u3001world generator \u548c action predictor\uff0c\u8b93\u4e09\u8005\u900f\u904e joint multimodal attention \u4ea4\u63db\u8cc7\u8a0a\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u6a21\u578b\u4e0d\u53ea\u4f9d\u8cf4\u6a5f\u68b0\u4eba\u793a\u7bc4\uff0c\u9810\u8a13\u7df4\u4ea6\u6df7\u5408 human egocentric data\u3001robot data \u53ca visual question answering\uff08VQA\uff09\u8cc7\u6599\uff0c\u4e26\u628a\u4f4e\u5c64\u52d5\u4f5c\u8f49\u6210\u81ea\u7136\u8a9e\u8a00\uff0c\u6539\u5584\u8996\u89ba\u8a9e\u8a00\u80fd\u529b\u8207\u5be6\u9ad4\u64cd\u4f5c\u4e4b\u9593\u7684\u929c\u63a5\u3002Post-training \u518d\u52a0\u5165 future visual noise augmentation\uff0c\u914d\u5408 history-conditioned flow matching\uff0c\u4ee5\u8f03\u5c11 denoising steps \u751f\u6210\u52d5\u4f5c\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u6574\u5408\u7bc4\u570d\uff1a<\/strong> \u540c\u6642\u8655\u7406 physical reasoning\u3001\u8996\u89ba\u751f\u6210\u8207\u52d5\u4f5c\u9810\u6e2c<\/li>\n<li><strong>\u8a13\u7df4\u8a0a\u865f\uff1a<\/strong> \u7d50\u5408\u6a5f\u68b0\u4eba\u3001\u7b2c\u4e00\u8eab\u4eba\u985e\u8996\u89d2\u53ca VQA \u8cc7\u6599<\/li>\n<li><strong>\u6548\u7387\u8a2d\u8a08\uff1a<\/strong> \u7528\u52d5\u4f5c\u6b77\u53f2\u521d\u59cb\u5316\u751f\u6210\u6d41\u7a0b\uff0c\u6e1b\u5c11 denoising steps<\/li>\n<li><strong>\u786c\u4ef6\u6210\u672c\uff1a<\/strong> RoboTwin 2.0 post-training \u66fe\u4ee5 8 \u5f35 NVIDIA H100 GPU \u57f7\u884c 10 \u5c0f\u6642<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">GitHub \u5132\u5b58\u5eab\u63d0\u4f9b <code>models\/<\/code>\u3001<code>train\/<\/code> \u548c <code>utils\/<\/code> \u7b49\u6a21\u578b\u8207\u8a13\u7df4\u7a0b\u5f0f\uff0c\u4e26\u9644\u6709 RoboTwin 2.0\u3001LIBERO \u53ca LIBERO-plus \u7684\u8cc7\u6599\u8655\u7406\u548c\u8a55\u4f30\u5de5\u5177\u3002Python 3.10\u3001CUDA \u517c\u5bb9\u7684 PyTorch \u74b0\u5883\u8f03\u9069\u5408\u6e2c\u8a66\uff0c\u4f7f\u7528\u8005\u53ef\u5148\u4ee5\u6a21\u64ec\u5668\u8cc7\u6599\u548c\u914d\u7f6e\u6a94\u9a57\u8b49\u8a13\u7df4\u6216\u63a8\u7406\u6d41\u7a0b\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u9805\u76ee\u9996\u9801\u5ba3\u7a31 UniWAM \u5728 in-distribution\u3001robustness\u3001generalization \u53ca instruction following \u7b49\u591a\u9805\u8a55\u4f30\u53d6\u5f97 SOTA\uff0c\u4f46\u5132\u5b58\u5eab\u63d0\u4f9b\u7684\u8cc7\u8a0a\u672a\u5217\u51fa\u5b8c\u6574\u5206\u6578\u548c\u6bd4\u8f03\u8868\u3002Bridge\u3001DROID\u3001Fractal \u4ee5\u53ca real-world inference \u660e\u78ba\u4e0d\u5728\u4eca\u6b21\u767c\u5e03\u7bc4\u570d\u5167\uff0c\u56e0\u6b64\u5b83\u8f03\u9069\u5408\u7814\u7a76\u5718\u968a\u6e2c\u8a66\u7d71\u4e00\u5f0f\u6a5f\u68b0\u4eba\u6a21\u578b\uff0c\u800c\u4e0d\u662f\u5373\u6642\u63a5\u99c1\u771f\u5be6\u6a5f\u68b0\u4eba\u7684\u73fe\u6210\u65b9\u6848\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/uniwam.github.io\/\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/UniWAM\/UniWAM\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a> \u00b7 <a href=\"https:\/\/www.modelscope.cn\/collections\/Kosmos524\/UniWAM\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>UniWAM \u4ee5 8B \u53c3\u6578\u6a21\u578b\u6574\u5408\u8a9e\u610f\u7406\u89e3\u3001\u8996\u89ba\u9810\u6e2c\u548c\u52d5\u4f5c\u751f\u6210\uff0c\u7784\u6e96\u6a5f\u68b0\u4eba\u64cd\u4f5c\u4e2d\u7406\u89e3\u8207\u57f7\u884c\u812b\u7bc0\u7684\u554f\u984c\u3002<\/p>\n","protected":false},"author":8,"featured_media":13120,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"wpai_generated_summary":"","wpai_meta_description":"","footnotes":""},"categories":[133,165,195,76,149,119,186,127,179,204,184,163,38,152,150,199],"tags":[],"class_list":["post-13121","post","type-post","status-publish","format-standard","hentry","category-133","category-165","category-195","category-76","category-149","category-119","category-186","category-127","category-nvidia","category-visionlanguageaction","category-robotic","category-163","category-38","category-python","category-150","category-dataset-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/13121","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=13121"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/13121\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/13120"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=13121"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=13121"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=13121"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}