
{"id":9401,"date":"2026-06-21T00:40:06","date_gmt":"2026-06-20T16:40:06","guid":{"rendered":"https:\/\/infernews.com\/blog\/implementation-of-paper-quot-playful-agentic-robot-learning-quot\/"},"modified":"2026-06-21T02:07:35","modified_gmt":"2026-06-20T18:07:35","slug":"implementation-of-paper-quot-playful-agentic-robot-learning-quot","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/implementation-of-paper-quot-playful-agentic-robot-learning-quot\/","title":{"rendered":"RATs \u7528\u591a\u4ee3\u7406\u73a9\u51fa\u6a5f\u68b0\u4eba\u6280\u80fd\u5eab"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/06\/rats-teaser.jpg\" alt=\"RATs pipeline overview \u2014 click to play the video\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u73fe\u6709\u6a5f\u68b0\u4eba\u4ee3\u7406\u5f88\u591a\u6642\u4ecd\u7136\u6cbf\u7528 task-driven \u8def\u7dda\uff1a\u5148\u6536\u5230\u660e\u78ba\u6307\u4ee4\uff0c\u518d\u900f\u904e Code-as-Policy \u7522\u751f\u53ef\u57f7\u884c\u7a0b\u5f0f\u4f86\u5b8c\u6210\u4efb\u52d9\u3002RATs \u5247\u6279\u8a55\u9019\u7a2e\u505a\u6cd5\u592a\u4f9d\u8cf4\u5916\u90e8\u4efb\u52d9\uff0c\u4ee4\u53ef\u91cd\u7528\u6280\u80fd\u53ea\u6703\u5728\u88ab\u8981\u6c42\u6642\u624d\u51fa\u73fe\uff0c\u6240\u4ee5\u5b83\u63d0\u51fa\u4e00\u500b\u591a\u4ee3\u7406 Code-as-Policy \u7cfb\u7d71\uff0c\u5148\u7528 free-form play \u81ea\u884c\u767c\u660e\u7df4\u7fd2\u76ee\u6a19\uff0c\u518d\u628a\u6210\u529f\u884c\u70ba\u6574\u7406\u6210\u6280\u80fd\u5eab\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u9019\u500b\u9805\u76ee\u5c6c\u65bc<strong>\u6a5f\u68b0\u4eba\u5b78\u7fd2\u6846\u67b6<\/strong>\uff0c\u8981\u89e3\u6c7a\u7684\u662f\u6a5f\u68b0\u4eba\u4ee3\u7406\u9047\u5230\u65b0\u4efb\u52d9\u6642\uff0c\u6b20\u7f3a\u53ef\u76f4\u63a5\u8abf\u7528\u7684\u9577\u671f\u6280\u80fd\u7d2f\u7a4d\u3002RATs \u5206\u6210 Play \u8207 Evaluation \u5169\u6bb5\uff1a\u524d\u8005\u7531 proposer\u3001planner\u3001policy-writer\u3001verifier\u3001failure-diagnoser \u5e7e\u500b LLM \u4ee3\u7406\u5354\u4f5c\uff0c\u5f8c\u8005\u628a\u5df2\u51cd\u7d50\u7684\u6280\u80fd\u7576\u6210 planner context \u91cd\u7528\uff0c\u800c\u4e14\u5f37\u8abf <strong>no gradients\u3001no RL<\/strong>\uff0c\u4e3b\u8981\u9760 structured natural-language feedback \u8207 code reuse \u5b78\u7fd2\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5982\u679c\u4f60\u60f3\u8a66\u9019\u500b\u9805\u76ee\uff0c\u8f03\u9069\u5408\u628a\u5b83\u7576\u6210\u7814\u7a76\u578b\u7cfb\u7d71\u4f86\u8dd1 benchmark\uff0c\u800c\u4e0d\u662f\u5373\u88dd\u5373\u7528\u5c0f\u5de5\u5177\u3002\u74b0\u5883\u8981\u6c42\u5305\u62ec Python 3.10\u3001CUDA-capable GPU\uff0c\u4e26\u727d\u6d89 LIBERO-PRO\u3001MolmoSpaces\u3001Robosuite \u53ca\u771f\u5be6 Franka Panda \u6d41\u7a0b\uff1b\u6bd4\u8f03\u5408\u7406\u7684\u6e2c\u8a66\u6b21\u5e8f\uff0c\u662f\u5148\u770b Play \u968e\u6bb5\u600e\u6a23\u751f\u6210\u6280\u80fd\uff0c\u518d\u6aa2\u67e5 Evaluation \u968e\u6bb5\u5c0d held-out tasks \u6709\u6c92\u6709\u6539\u5584\u3002<\/p>\n\n\n<div\n    \tclass=\"align wp-block-vpb-video\"    id='vpbpVideoPlayer-1'\n    data-attributes='{&quot;source&quot;:&quot;https:\\\/\\\/playful-rats.github.io\\\/assets\\\/videos\\\/pipeline-overview-v9.mp4&quot;,&quot;repeat&quot;:true,&quot;autoplay&quot;:true,&quot;muted&quot;:true,&quot;align&quot;:&quot;&quot;,&quot;poster&quot;:&quot;&quot;,&quot;controls&quot;:{&quot;play-large&quot;:true,&quot;restart&quot;:false,&quot;rewind&quot;:true,&quot;play&quot;:true,&quot;fast-forward&quot;:true,&quot;progress&quot;:true,&quot;current-time&quot;:true,&quot;duration&quot;:false,&quot;mute&quot;:true,&quot;volume&quot;:true,&quot;pip&quot;:false,&quot;airplay&quot;:false,&quot;settings&quot;:true,&quot;download&quot;:false,&quot;fullscreen&quot;:true},&quot;width&quot;:&quot;100%&quot;,&quot;radius&quot;:&quot;0px&quot;,&quot;resetOnEnd&quot;:false,&quot;autoHideControl&quot;:true,&quot;isSetup&quot;:false}'\n><\/div>\n\n\n<p class=\"wp-block-paragraph\">\u5b83\u7684\u5275\u65b0\u9ede\uff0c\u5728\u65bc\u628a\u300c\u73a9\u300d\u6b63\u5f0f\u7d0d\u5165 lifelong robot skill learning\uff1a\u4e0d\u662f\u96a8\u6a5f\u63a2\u7d22\uff0c\u800c\u662f\u8b93\u4ee3\u7406\u81ea\u5df1\u63d0\u51fa\u53ef\u5b78\u7fd2\u4efb\u52d9\u3001\u9010\u6b65\u9a57\u8b49\u4e2d\u9593\u9032\u5ea6\u3001\u5931\u6557\u5f8c\u518d\u8a3a\u65b7\u91cd\u8a66\uff0c\u6700\u5f8c\u628a\u6210\u529f\u57f7\u884c\u84b8\u993e\u6210 reusable skill library\u3002\u9019\u4ee4\u6280\u80fd\u53ef\u5728\u8de8\u74b0\u5883\u60c5\u5883\u91cd\u7528\uff0c\u4e0d\u4e00\u5b9a\u7d81\u6b7b\u539f\u672c\u8a13\u7df4\u5834\u666f\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u8ad6\u6587\u7d66\u51fa\u7684\u7d50\u679c\u76f8\u7576\u5177\u9ad4\uff1a\u5728 LIBERO-PRO \u8207 MolmoSpaces\uff0cplay-learned skills \u76f8\u6bd4 no play \u8207 random-play baselines \u6709\u63d0\u5347\uff0c\u5c0d CaP-Agent0 \u5206\u5225\u9ad8\u51fa 20.6 \u548c 17.0 \u500b\u767e\u5206\u9ede\uff1b\u628a\u6280\u80fd\u76f4\u63a5\u6aa2\u7d22\u9032\u5176\u4ed6 inference-time Code-as-Policy agents \u7684 context\uff0c\u5c0d Robosuite \u8207\u771f\u5be6\u4e16\u754c transfer \u4ea6\u5206\u5225\u63d0\u5347 8.9 \u548c 8.8 \u9ede\u3002\u76f8\u95dc\u6a21\u578b\u8207\u57fa\u7dda\u4e3b\u8981\u5305\u62ec CaP-X\u3001CaP-Agent0\uff0c\u4ee5\u53ca\u6587\u4e2d\u4f7f\u7528\u7684 LLM agents \u5354\u4f5c\u6d41\u7a0b\uff1b\u82e5\u4f60\u95dc\u5fc3 agentic robotics\u3001\u6280\u80fd\u91cd\u7528\u8207\u771f\u6a5f\u8f49\u79fb\uff0c\u9019\u500b\u9805\u76ee\u5f88\u503c\u5f97\u7d30\u8b80\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u985e\u578b\u5b9a\u4f4d<\/strong>\uff1a\u591a\u4ee3\u7406\u6a5f\u68b0\u4eba\u5b78\u7fd2\u6846\u67b6\uff0c\u6838\u5fc3\u662f Code-as-Policy \u8207\u6280\u80fd\u5eab\u91cd\u7528<\/li>\n\n\n\n<li><strong>\u65b9\u6cd5\u91cd\u9ede<\/strong>\uff1a\u5148 Play \u81ea\u63d0\u4efb\u52d9\u5b78\u6280\u80fd\uff0c\u518d Evaluation \u628a\u6280\u80fd\u6ce8\u5165 planner context<\/li>\n\n\n\n<li><strong>\u6280\u8853\u53d6\u5411<\/strong>\uff1a\u4e0d\u9760 gradients \u6216 RL\uff0c\u4e3b\u8981\u4f9d\u8cf4\u81ea\u7136\u8a9e\u8a00\u56de\u994b\u3001\u7a0b\u5f0f\u4fee\u6b63\u8207 code reuse<\/li>\n\n\n\n<li><strong>\u9069\u5408\u5834\u666f<\/strong>\uff1a\u7814\u7a76 embodied agents\u3001robot skill library\u3001cross-environment transfer \u7684\u5718\u968a<\/li>\n\n\n\n<li><strong>\u5df2\u63d0\u5230\u7684\u76f8\u95dc\u7cfb\u7d71<\/strong>\uff1aCaP-X\u3001CaP-Agent0\u3001LIBERO-PRO\u3001MolmoSpaces\u3001Robosuite\u3001Franka Panda<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>GitHub\uff1a<\/strong> <a href=\"https:\/\/github.com\/Playful-RATs\/rats\" rel=\"noopener noreferrer\">https:\/\/github.com\/Playful-RATs\/rats<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u9805\u76ee\uff1a<\/strong> <a href=\"https:\/\/playful-rats.github.io\/\" rel=\"noopener\">https:\/\/playful-rats.github.io\/<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>RATs \u628a\u6a5f\u68b0\u4eba\u5b78\u7fd2\u62c6\u6210\u5148\u73a9\u3001\u5f8c\u89e3\u984c\u5169\u6bb5\u6d41\u7a0b\uff0c\u91cd\u9ede\u4e0d\u662f\u5fae\u8abf\uff0c\u800c\u662f\u9760 LLM \u4ee3\u7406\u5beb\u7a0b\u5f0f\u7d2f\u7a4d\u53ef\u91cd\u7528\u6280\u80fd\u3002<\/p>\n","protected":false},"author":8,"featured_media":9400,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,179,116,38,162,152,105,150,76,127,184,197,201],"tags":[],"class_list":["post-9401","post","type-post","status-publish","format-standard","hentry","category-133","category-nvidia","category-agentic","category-38","category-ai-productions","category-python","category-python-nlp","category-150","category-76","category-127","category-robotic","category-framework","category-skill-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9401","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=9401"}],"version-history":[{"count":4,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9401\/revisions"}],"predecessor-version":[{"id":9420,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/9401\/revisions\/9420"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/9400"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=9401"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=9401"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=9401"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}