{"id":12518,"date":"2026-09-22T18:53:05","date_gmt":"2026-09-22T10:53:05","guid":{"rendered":"https:\/\/infernews.com\/blog\/grounded-action-model-3d-grounding-as-a-foundation-for-robotics-code-coming-soon-2\/"},"modified":"2026-09-22T18:54:18","modified_gmt":"2026-09-22T10:54:18","slug":"grounded-action-model-3d-grounding-as-a-foundation-for-robotics-code-coming-soon-2","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/grounded-action-model-3d-grounding-as-a-foundation-for-robotics-code-coming-soon-2\/","title":{"rendered":"Grounded-Action-Model\uff1a\u8b93\u6a5f\u68b0\u4eba\u5148\u7406\u89e3\u7269\u4ef6\u4f4d\u7f6e\uff0c\u518d\u6c7a\u5b9a\u5982\u4f55\u52d5\u4f5c"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/teaser-7.jpg\" alt=\"A language instruction, a 2D point or a 2D box is resolved by a pretrained 3D grounding model into image tokens and dete\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u9762\u5c0d\u7269\u4ef6\u88ab\u79fb\u4f4d\u3001\u5834\u666f\u91cd\u65b0\u6392\u5217\uff0c\u6a5f\u68b0\u81c2\u672a\u5fc5\u53ea\u9700\u8981\u66f4\u9577\u7684\u6307\u4ee4\uff0c\u800c\u662f\u8981\u6e96\u78ba\u77e5\u9053\u76ee\u6a19\u5728\u54ea\u88cf\u3002Grounded Action Model\uff08GAM\uff09\u5c6c\u65bc\u6a5f\u68b0\u4eba\u57fa\u790e\u6a21\u578b\uff0c\u900f\u904e 3D grounding \u5c07\u8a9e\u8a00\u30012D \u9ede\u6216 2D \u6846\u9078\u8f49\u6210\u540c\u4e00\u5957\u4ee5\u7269\u4ef6\u70ba\u4e2d\u5fc3\u7684\u89c0\u5bdf\uff0c\u518d\u9810\u6e2c\u9023\u7e8c\u52d5\u4f5c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">GAM \u4f7f\u7528\u51cd\u7d50\u7684 promptable 3D detector\uff0c\u63d0\u53d6\u76ee\u6a19\u7269\u4ef6\u7684\u8996\u89ba\u7279\u5fb5\u548c metric geometry\uff0c\u4e26\u8207\u6a5f\u68b0\u4eba\u72c0\u614b\u6b77\u53f2\u4ea4\u7d66 multi-stream transformer\uff0c\u8f38\u51fa action chunks\u3002\u76f8\u8f03\u65bc\u7531 Vision-Language-Action Models\uff08VLAs\uff09\u6216 World-Action Models\uff08WAMs\uff09\u5f9e\u793a\u7bc4\u4e2d\u96b1\u6027\u5b78\u7fd2\u4f4d\u7f6e\u95dc\u4fc2\uff0cGAM \u628a 3D \u4f4d\u7f6e\u76f4\u63a5\u653e\u9032\u63a7\u5236\u6d41\u7a0b\uff0c\u4e5f\u53ef\u7531\u9ad8\u5c64 planner \u900f\u904e\u4e0d\u540c\u63d0\u793a\u65b9\u5f0f\u63a7\u5236\uff0c\u652f\u63f4\u8f03\u9577\u6d41\u7a0b\u548c\u9700\u8981\u8a18\u61b6\u7684\u64cd\u4f5c\u3002<\/p>\n\n\n\n<figure class=\"wp-block-image size-full\"><img data-dominant-color=\"d9d9d8\" data-has-transparency=\"false\" style=\"--dominant-color: #d9d9d8;\" loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"530\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/pasted-11f55e68968e.jpg\" alt=\"A language instruction, a 2D point or a 2D box is resolved by a pretrained 3D grounding model into image tokens and dete\" class=\"wp-image-12515 not-transparent\" srcset=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/pasted-11f55e68968e.jpg 1024w, https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/pasted-11f55e68968e-300x155.jpg 300w, https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/pasted-11f55e68968e-768x398.jpg 768w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u76ee\u524d\u53ef\u53c3\u8003\u7684\u7d50\u679c\u5305\u62ec\uff1a<br>\n&#8211; RoboTwin 2.0 \u7684 50 \u9805\u4efb\u52d9\u5e73\u5747\u6210\u529f\u7387\u70ba 55.3%\uff0c\u5834\u666f\u96a8\u6a5f\u5316\u6642\u70ba 47.6%\u3002<br>\n&#8211; LIBERO-PRO \u7684 16 \u7a2e\u64fe\u52d5\u8a2d\u5b9a\u5e73\u5747\u6210\u529f\u7387\u70ba 61%\uff0c\u76ee\u6a19\u91cd\u65b0\u653e\u7f6e\u6642\u6539\u5584\u8f03\u660e\u986f\u3002<br>\n&#8211; \u96d9\u81c2 YAM \u9762\u5c0d\u8996\u89ba\u8b8a\u5316\u4ecd\u6210\u529f 17\/20 \u6b21\uff1bFranka \u914d\u5408 Molmo2 planner\uff0c\u9577\u6d41\u7a0b\u4efb\u52d9\u7684 OOD step completion \u70ba 49.8%\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Grounded-Action-Model \u9069\u5408\u7814\u7a76\u6a5f\u68b0\u4eba\u64cd\u4f5c\u3001\u8996\u89ba\u8b8a\u5316\u548c\u9ad8\u4f4e\u5c64\u63a7\u5236\u5354\u4f5c\u7684\u5718\u968a\uff0c\u4f46\u76ee\u524d GitHub \u4ecd\u53ea\u6709\u8ad6\u6587\u8207\u9805\u76ee\u9801\u9762\uff0ctraining and inference code\u3001pretrained checkpoints \u53ca real-robot setup \u5c1a\u5f85\u516c\u958b\uff0c\u56e0\u6b64\u66ab\u6642\u4e0d\u80fd\u6309\u5132\u5b58\u5eab\u76f4\u63a5\u5b89\u88dd\u6216\u91cd\u73fe\u7d50\u679c\u3002\u6578\u64da\u4e3b\u8981\u4f86\u81ea\u6307\u5b9a\u57fa\u6e96\u548c\u5169\u6b3e\u6a5f\u68b0\u4eba\uff0c\u80fd\u5426\u5ef6\u4f38\u5230\u5176\u4ed6\u786c\u4ef6\u8207\u672a\u898b\u904e\u7684\u64cd\u4f5c\uff0c\u4ecd\u8981\u7b49\u5f85\u7a0b\u5f0f\u78bc\u53ca\u66f4\u591a\u6e2c\u8a66\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/grounded-action-model.github.io\/\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/GehaoZhang6\/Grounded-Action-Model\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Grounded Action Model \u628a\u8a9e\u8a00\u3001\u9ede\u548c\u6846\u9078\u63d0\u793a\u8f49\u6210\u540c\u4e00\u5957 3D \u7269\u4ef6\u8868\u793a\uff0c\u63d0\u5347\u6a5f\u68b0\u81c2\u9762\u5c0d\u5834\u666f\u8b8a\u5316\u6642\u7684\u7a69\u5b9a\u6027\u3002<\/p>\n","protected":false},"author":8,"featured_media":12517,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"wpai_generated_summary":"","footnotes":""},"categories":[133,76,149,119,127,173,204,210,184,132,150],"tags":[],"class_list":["post-12518","post","type-post","status-publish","format-standard","hentry","category-133","category-76","category-149","category-119","category-127","category-173","category-visionlanguageaction","category-worldactionmodel","category-robotic","category-3d","category-150"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/12518","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=12518"}],"version-history":[{"count":1,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/12518\/revisions"}],"predecessor-version":[{"id":12521,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/12518\/revisions\/12521"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/12517"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=12518"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=12518"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=12518"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}