[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-robotics":3},{"tag":4,"articles":10},{"id":5,"name":6,"slug":6,"article_count":7,"description_zh":8,"description_en":9},"bbb7896a-2a77-418e-b199-5614fb32b3d5","robotics",3,"機器人技術涵蓋感知、控制、模擬與規劃，重點在讓模型能在真實環境中穩定行動。這個標籤聚焦世界模型、階層式規劃、少樣本示範與可變形物體模擬等主題，反映如何降低算力成本並提升長時序控制與零樣本遷移能力。","Robotics here focuses on the methods that let machines act reliably in the physical world: perception, control, simulation, and planning. Topics include world models, hierarchical planning, sparse-demonstration learning, and deformable-object simulation, all aimed at better long-horizon control with less compute.",[11],{"id":12,"slug":13,"title":14,"summary":15,"category":16,"image_url":17,"cover_image":17,"language":18,"created_at":19},"b46591a1-fdf6-42dd-9d29-d472bf9feeb9","hierarchical-planning-latent-world-models-zh","階層式規劃讓世界模型更省算力","這篇論文用多時間尺度的潛在世界模型做階層式規劃，目標是讓長時序控制更穩、搜尋成本更低，還能做真實機器人的 zero-shot 控制。","research","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775455787320-5ntv.png","zh","2026-04-06T06:09:31.716627+00:00"]