{"id":11779,"date":"2026-09-01T18:35:36","date_gmt":"2026-09-01T10:35:36","guid":{"rendered":"https:\/\/infernews.com\/blog\/lightnav-0\/"},"modified":"2026-09-01T18:35:36","modified_gmt":"2026-09-01T10:35:36","slug":"lightnav-0","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/lightnav-0\/","title":{"rendered":"LightNav-0\uff1a\u4e00\u53e5\u8aaa\u8a71\u9a45\u52d5\u56db\u7a2e\u6a5f\u68b0\u4eba\u96f6\u6a23\u672c\u5c0e\u822a"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/teaser.jpg\" alt=\"LightNav-0 overview: a simulation-based data engine, three-stage model training, zero-shot deployment onto four robot em\"><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u4f60\u53eb\u4e00\u90e8\u6a5f\u68b0\u4eba\u300c\u53bb\u5230\u516c\u5712\u5165\u53e3\u55f0\u5f35\u9577\u6905\u300d\uff0c\u4f62\u5c31\u8981\u81ea\u5df1\u6435\u5230\u3001\u907f\u958b\u969c\u7919\u3001\u4ef2\u8981\u8655\u7406\u81ea\u5df1\u5c0d\u773c\u7747\u5230\u5605\u756b\u9762\u2014\u2014LightNav-0 \u5c31\u4fc2\u91dd\u5c0d\u5462\u985e\u8de8\u5834\u666f\u3001\u8de8\u5f62\u614b\u5605\u5c0e\u822a\u9700\u6c42\u3002\u4f62\u5c6c\u65bc\u6a21\u578b\u985e\u9805\u76ee\uff0c\u6838\u5fc3\u60f3\u89e3\u6c7a\u5605\u554f\u984c\u4fc2\uff1a\u50b3\u7d71\u5c0e\u822a\u7cfb\u7d71\u6bcf\u63db\u4e00\u7a2e\u6a5f\u68b0\u4eba\u6216\u5834\u666f\u5c31\u8981\u91cd\u65b0\u8a13\u7df4\u6216\u52a0\u4efb\u52d9\u982d\uff0c\u96f6\u6a23\u672c\u6cdb\u5316\u80fd\u529b\u8584\u5f31\u3002LightNav-0 \u9078\u64c7\u5514\u52a0 waypoint predictor\u3001\u5514\u52a0\u4efb\u52d9\u5c08\u7528\u52d5\u4f5c\u982d\uff0c\u6de8\u4fc2\u64f4\u5c55\u8a5e\u8868\uff0c\u52a0\u5165\u96d9\u901a\u9053\u6307\u5411 token \u540c RVQ \u52d5\u4f5c token\uff0c\u7b49\u7a7a\u9593\u63a8\u7406\u540c\u52d5\u4f5c\u7de8\u78bc\u90fd\u76f4\u63a5\u7531 Qwen3-VL-4B-Instruct \u539f\u6709\u5605\u81ea\u56de\u6b78 LM head \u89e3\u78bc\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u8981\u505a\u5230\u5462\u9ede\uff0c\u8a13\u7df4\u6578\u64da\u4fc2\u95dc\u9375\u3002\u9805\u76ee\u7528\u5497\u4e00\u5957 Real2Sim2Real \u6578\u64da\u5f15\u64ce\uff0c\u5c07 2,000 \u5e7e\u500b\u4e92\u806f\u7db2\u6536\u96c6\u5605\u771f\u5be6\u5834\u666f\u8f49\u6210\u53ef\u91cd\u8907\u4f7f\u7528\u5605\u6a21\u64ec\u74b0\u5883\uff0c\u7522\u751f 4,000 \u5e7e\u500b\u9418\u5605\u8996\u89ba\u8a9e\u8a00\u52d5\u4f5c\u7d93\u9a57\uff0c\u518d\u5206\u4e09\u968e\u6bb5\u8a13\u7df4\uff1aEmbodied Reasoning \u4e2d\u8a13\u7df4\u3001Embodied SFT\u3001Online RL\u3002\u5462\u7a2e\u505a\u6cd5\u7e5e\u904e\u5497\u771f\u5be6\u9059\u64cd\u4f5c\u6536\u96c6\u901f\u5ea6\u6162\u3001\u6210\u672c\u9ad8\u5605\u74f6\u9838\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u540c\u5178\u578b\u505a\u6cd5\u6bd4\uff0c\u5dee\u7570\u5728\u65bc\u7528\u6a21\u64ec\u5408\u6210\u4ee3\u7406\u771f\u5be6\u7d93\u9a57\uff0c\u518d\u900f\u904e\u7d71\u4e00 token \u4ecb\u9762\u8b93\u6307\u4ee4\u8ddf\u96a8\u3001\u958b\u653e\u8a5e\u5f59\u7269\u4ef6\u5c0e\u822a\u3001\u8996\u89ba\u8ffd\u8e64\u5171\u4eab\u540c\u4e00\u6a21\u578b\uff0c\u90e8\u7f72\u6642\u76f4\u63a5\u96f6\u6a23\u672c\u843d\u5730\u5230\u56db\u7a2e\u6a5f\u68b0\u4eba\u5f62\u614b\uff0c\u5514\u9700\u8981\u9010\u500b\u5fae\u8abf\u3002\u7d50\u679c\u65b9\u9762\uff0c\u5b98\u65b9\u6307 LightNav-0 \u55ba\u5341\u500b\u6a21\u64ec\u8a2d\u5b9a\u4e0a\u53d6\u5f97 state-of-the-art\uff0c\u4e26\u55ba\u4eba\u5f62\u3001\u56db\u8db3\u3001\u7121\u4eba\u6a5f\u540c\u8f2a\u5f0f\u6a5f\u68b0\u4eba\u4e0a\u5b8c\u6210\u771f\u5be6\u9077\u79fb\u6e2c\u8a66\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u505a\u6a5f\u68b0\u4eba\u901a\u7528\u5c0e\u822a\u7814\u7a76\u5605\u5718\u968a\u3001\u8981\u505a\u5177\u8eab AI \u843d\u5730\u5605\u521d\u5275\uff0c\u6216\u8005\u95dc\u6ce8 Physical AGI \u8def\u7dda\u5605\u7814\u7a76\u8005\uff0c\u6703\u6700\u5bb9\u6613\u5f9e\u4e2d\u53d7\u60e0\u3002\u5c0d\u4e00\u822c\u958b\u767c\u8005\u569f\u8b1b\uff0c\u7406\u89e3\u5165\u53e3\u4fc2 Hugging Face \u4e0a\u5605\u6a21\u578b\u6b0a\u91cd\u3001\u8ad6\u6587\u540c\u9805\u76ee\u9801\u63d0\u4f9b\u5605\u6a21\u64ec\u8a55\u6e2c\u7d50\u679c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u91cd\u9ede\u6458\u8981\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>LightNav-0 \u4fc2\u4ee5 Qwen3-VL-4B-Instruct \u70ba\u9aa8\u5e79\u5605\u901a\u7528\u5c0e\u822a\u6a21\u578b\uff0c\u7121\u4efb\u52d9\u5c08\u7528\u9810\u6e2c\u982d<\/li>\n<li>\u7528\u7d71\u4e00 token \u4ecb\u9762\u540c\u6642\u8986\u84cb\u6307\u4ee4\u8ddf\u96a8\u3001\u958b\u653e\u8a5e\u5f59\u7269\u4ef6\u5c0e\u822a\u540c\u8996\u89ba\u8ffd\u8e64<\/li>\n<li>Real2Sim2Real \u6578\u64da\u5f15\u64ce\u5c07 2,000+ \u771f\u5be6\u5834\u666f\u8f49\u5316\u70ba 4,000+ \u5c0f\u6642\u8a13\u7df4\u7d93\u9a57<\/li>\n<li>\u4e09\u968e\u6bb5\u8a13\u7df4\u6d41\u7a0b\uff1aEmbodied Reasoning \u4e2d\u8a13\u7df4\u3001Embodied SFT\u3001Online RL<\/li>\n<li>\u96f6\u6a23\u672c\u9077\u79fb\u81f3\u4eba\u5f62\u3001\u56db\u8db3\u3001\u8f2a\u5f0f\u53ca\u7a7a\u4e2d\u6a5f\u68b0\u4eba\uff0c\u55ba\u5341\u500b\u6a21\u64ec\u57fa\u6e96\u5ba3\u7a31\u9054\u5230 SOTA<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/www.lightorigins.com\/en\/blog\/lightnav-0\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/lightorigins\/LightNav-0\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>LightNav-0 \u662f\u4e00\u500b\u901a\u7528\u5c0e\u822a\u6a21\u578b\uff0c\u628a\u9810\u8a13\u7df4\u8996\u89ba\u8a9e\u8a00\u6a21\u578b Qwen3-VL \u7684\u7a7a\u9593\u8a8d\u77e5\u80fd\u529b\u5c0d\u9f4a\u5230\u5c0e\u822a\u4efb\u52d9\uff0c\u7121\u9700\u4efb\u52d9\u5c08\u7528\u982d\u5373\u53ef\u540c\u6642\u8655\u7406\u6307\u4ee4\u8ddf\u96a8\u3001\u958b\u653e\u8a5e\u5f59\u7269\u4ef6\u5c0e\u822a\u8207\u8996\u89ba\u8ffd\u8e64\uff0c\u4e26\u96f6\u6a23\u672c\u9077\u79fb\u5230\u4eba\u5f62\u3001\u56db\u8db3\u3001\u8f2a\u5f0f\u53ca\u7a7a\u4e2d\u6a5f\u68b0\u4eba\u3002<\/p>\n","protected":false},"author":8,"featured_media":11778,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"wpai_generated_summary":"","footnotes":""},"categories":[133,185,119,149,184,193,210],"tags":[],"class_list":["post-11779","post","type-post","status-publish","format-standard","hentry","category-133","category-qwen","category-119","category-149","category-robotic","category-agi","category-worldactionmodel"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11779","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=11779"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11779\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/11778"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=11779"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=11779"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=11779"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}