{"id":9478,"date":"2026-06-21T16:52:13","date_gmt":"2026-06-21T08:52:13","guid":{"rendered":"https:\/\/infernews.com\/blog\/molmo-motion\/"},"modified":"2026-06-21T16:54:32","modified_gmt":"2026-06-21T08:54:32","slug":"molmo-motion","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/molmo-motion\/","title":{"rendered":"MolmoMotion \u628a\u8a9e\u8a00\u8b8a\u6210 3D \u52d5\u4f5c\u9810\u6e2c"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/06\/teaser-3.jpg\" alt=\"MolmoMotion teaser\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u73fe\u6709\u505a\u6cd5\u591a\u6578\u504f\u5411\u8ffd\u8e64\u5df2\u7d93\u767c\u751f\u7684\u79fb\u52d5\uff0c\u6216\u8005\u53ea\u5728 2D \u756b\u9762\u4f30\u8a08\u4e0b\u4e00\u6b65\u4f4d\u7f6e\uff1b\u4f5c\u8005\u8a8d\u70ba\u9019\u7a2e retrospective \u7bc4\u5f0f\u96e3\u4ee5\u652f\u63f4\u6a5f\u68b0\u4eba\u898f\u5283\u8207\u53ef\u63a7\u5f71\u7247\u751f\u6210\uff0c\u6240\u4ee5\u63d0\u51fa MolmoMotion\u3001MolmoMotion-1M \u540c PointMotionBench\uff0c\u628a\u554f\u984c\u6539\u6210\u300c\u6839\u64da\u8a9e\u8a00\u6307\u4ee4\u9810\u6e2c 3D \u9ede\u4f4d\u672a\u4f86\u8ecc\u8de1\u300d\u3002\u9019\u4e0d\u662f\u55ae\u7d14\u770b\u5f71\u7247\u7406\u89e3\u5167\u5bb9\uff0c\u800c\u662f\u8981\u6a21\u578b\u6839\u64da\u77ed\u6bb5 RGB \u6b77\u53f2\u30012D query points\u3001\u521d\u59cb 3D \u4f4d\u7f6e\uff0c\u4ee5\u53ca\u6587\u5b57\u52d5\u4f5c\u63cf\u8ff0\uff0c\u9810\u6e2c\u4e4b\u5f8c\u7d04 2 \u79d2\u7684 3D movement\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">MolmoMotion \u672c\u8eab\u5c6c\u65bc<strong>\u6a21\u578b<\/strong>\uff0c\u66f4\u6e96\u78ba\u5730\u8aaa\u662f 4B vision-language model\uff0c\u8655\u7406\u7684\u662f\u7269\u4ef6\u4e0a\u6307\u5b9a\u9ede\u6703\u600e\u6a23\u79fb\u52d5\u7684\u9810\u6e2c\u554f\u984c\u3002\u5132\u5b58\u5eab\u76ee\u524d\u516c\u958b\u7684\u662f autoregressive (AR) variant\uff0c\u4e26\u63d0\u4f9b\u8a13\u7df4\u8cc7\u6599\u3001\u8a55\u6e2c\u96c6\u3001\u5df2\u91cb\u51fa\u6a21\u578b\uff0c\u4ee5\u53ca\u7531 pretrain \u5230 long-horizon finetune \u7684\u5169\u968e\u6bb5\u6d41\u7a0b\uff0c\u8f03\u9069\u5408\u7814\u7a76\u5718\u968a\u76f4\u63a5\u91cd\u73fe\u7d50\u679c\u6216\u6539\u6210\u81ea\u5bb6\u4efb\u52d9\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u548c\u540c\u985e\u65b9\u6cd5\u76f8\u6bd4\uff0c\u9019\u500b\u9805\u76ee\u7684\u53d6\u5411\u5f88\u9bae\u660e\uff1a\u5b83\u4e0d\u662f\u5148\u505a\u4e00\u822c\u5f71\u7247\u7406\u89e3\uff0c\u518d\u53e6\u5916\u63a5 motion head\uff0c\u800c\u662f\u628a language-guided 3D point trajectory forecasting \u7576\u6210\u6838\u5fc3\u4efb\u52d9\u3002\u4ee3\u50f9\u662f\u8f38\u5165\u8981\u6c42\u8f03\u591a\uff0c\u4f60\u8981\u6709 query points \u540c\u521d\u59cb 3D \u8cc7\u8a0a\uff1b\u56de\u5831\u5247\u662f\u8f38\u51fa\u66f4\u8cbc\u8fd1\u898f\u5283\u7528\u9014\uff0c\u7279\u5225\u9069\u5408\u8981\u9810\u6e2c\u300c\u7269\u4ef6\u5c07\u6703\u9ede\u6a23\u90c1\u300d\u800c\u4e0d\u662f\u53ea\u60f3\u5206\u985e\u5834\u666f\u7684\u4eba\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u53ef\u9810\u6e2c\u6700\u9577\u7d04 2 \u79d2\u672a\u4f86\u8ecc\u8de1\uff0c\u6587\u4ef6\u63d0\u5230 15 fps\u3001F=30 \u6216 F=32 \u7684\u8a2d\u5b9a<\/li>\n\n\n\n<li>\u8a55\u6e2c\u6307\u6a19\u5217\u51fa ADE\u3001FDE\u3001PWT\uff0c\u7126\u9ede\u653e\u5728\u8ecc\u8de1\u6e96\u78ba\u5ea6\u800c\u975e\u53ea\u770b\u756b\u9762\u76f8\u4f3c\u5ea6<\/li>\n\n\n\n<li>\u914d\u5957\u5305\u542b MolmoMotion-1M \u8a13\u7df4\u8cc7\u6599\u96c6\u8207 PointMotionBench \u8a55\u6e2c\u57fa\u6e96<\/li>\n\n\n\n<li>\u4f5c\u8005\u6307\u51fa\u5b78\u5230\u7684 motion prior \u53ef\u8f49\u79fb\u5230 robotics planning \u8207 motion-guided video generation<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u90e8\u7f72\u8207\u6e2c\u8a66\u7684\u7406\u89e3\u65b9\u5f0f\u76f8\u7576\u76f4\u63a5\uff1a\u5148\u4e0b\u8f09\u6a21\u578b\u3001\u8cc7\u6599\u96c6\u6216 benchmark\uff0c\u518d\u6309\u5132\u5b58\u5eab\u63d0\u4f9b\u7684 evaluation \u8207 training \u6d41\u7a0b\u57f7\u884c\uff1b\u82e5\u4e0d\u6253\u7b97\u91cd\u8a13\uff0c\u8f03\u5408\u7406\u662f\u5148\u62ff\u5df2\u91cb\u51fa\u6a21\u578b\u8dd1 PointMotionBench \u6216\u81ea\u5bb6\u6a23\u672c\uff0c\u770b\u770b\u8a9e\u8a00\u6307\u4ee4\u6539\u8b8a\u6642\uff0c3D \u9ede\u4f4d\u9810\u6e2c\u662f\u5426\u7a69\u5b9a\u3002\u73fe\u6709\u8cc7\u8a0a\u672a\u898b\u5b8c\u6574\u6548\u80fd\u6578\u5b57\u6458\u9304\uff0c\u4f46\u5b98\u65b9\u660e\u78ba\u8072\u7a31\u8868\u73fe\u660e\u986f\u512a\u65bc\u65e2\u6709 forecasting \u65b9\u6cd5\uff1b\u8f03\u4fdd\u5b88\u7684\u5224\u65b7\u662f\uff0c\u5b83\u6700\u9069\u5408 embodied AI\u3001robotics\u3001\u53ef\u63a7\u5f71\u7247\u751f\u6210\u7814\u7a76\u8005\uff0c\u4ee5\u53ca\u9700\u8981\u628a\u8a9e\u8a00\u610f\u5716\u8f49\u6210\u672a\u4f86\u904b\u52d5\u5047\u8a2d\u7684\u5718\u968a\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u76f8\u95dc\u6a21\u578b\u8207\u8cc7\u6e90\u5305\u62ec\uff1aMolmoMotion\u3001\u5176 autoregressive (AR) variant\u3001MolmoMotion-1M\u3001PointMotionBench\uff1b\u5e95\u5c64 backbone \u521d\u59cb\u5316\u8207\u5f9e\u96f6\u8a13\u7df4\u5165\u53e3\u4ea6\u6709\u63d0\u4f9b\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>GitHub\uff1a<\/strong> <a href=\"https:\/\/github.com\/allenai\/molmo-motion\" rel=\"noopener noreferrer\">https:\/\/github.com\/allenai\/molmo-motion<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u9805\u76ee\u4e3b\u9801\uff1a<\/strong> <a href=\"https:\/\/allenai.org\/blog\/molmo-motion\" rel=\"noopener noreferrer\">https:\/\/allenai.org\/blog\/molmo-motion<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u9019\u662f\u4e00\u500b\u6703\u6309\u6587\u5b57\u6307\u4ee4\u9810\u6e2c\u7269\u4ef6\u672a\u4f86\u79fb\u52d5\u8def\u5f91\u7684 4B \u8996\u89ba\u8a9e\u8a00\u6a21\u578b\uff0c\u4e5f\u9023\u540c\u8cc7\u6599\u96c6\u8207\u57fa\u6e96\u4e00\u4f75\u516c\u958b\u3002<\/p>\n","protected":false},"author":8,"featured_media":9477,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,31,132,162,158,119,150,76,127,149,141,184,199],"tags":[],"class_list":["post-9478","post","type-post","status-publish","format-standard","hentry","category-133","category-video","category-3d","category-ai-productions","category-ide","category-119","category-150","category-76","category-127","category-149","category-141","category-robotic","category-dataset-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9478","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=9478"}],"version-history":[{"count":1,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9478\/revisions"}],"predecessor-version":[{"id":9480,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9478\/revisions\/9480"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/9477"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=9478"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=9478"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=9478"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}