[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-routing":3},{"tag":4,"articles":9},{"id":5,"name":6,"slug":6,"article_count":7,"description_zh":8,"description_en":8},"5e39410d-43ed-4038-898f-80eccafefe5d","routing",1,null,[10,19,27],{"id":11,"slug":12,"title":13,"summary":14,"category":15,"image_url":16,"cover_image":16,"language":17,"created_at":18},"5b27896f-ad48-4a9a-8b6e-823568d8c669","wei-shen-me-lu-you-cai-shi-mo-xing-fu-wu-de-zhen-zheng-ping-zh","為什麼路由才是模型服務的真正瓶頸","模型服務的主要限制不是推理本身，而是路由決策；誰、何時、送到哪個模型與副本，才決定延遲、成本與穩定性。","industry","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778278838578-nms7.png","zh","2026-05-08T22:20:22.020009+00:00",{"id":20,"slug":21,"title":22,"summary":23,"category":24,"image_url":25,"cover_image":25,"language":17,"created_at":26},"072a2114-1f7f-4d61-99f7-be82c686c286","unipool-shared-expert-pool-moe-zh","UniPool：共享 MoE 專家池","UniPool 把 MoE 的分層專家改成全域共享池，減少重複參數，並在五個 LLaMA 規模模型上改善驗證損失。","research","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778221269156-lam7.png","2026-05-08T06:20:40.070989+00:00",{"id":28,"slug":29,"title":30,"summary":31,"category":15,"image_url":32,"cover_image":32,"language":17,"created_at":33},"54b3fd97-c8e6-4b92-b87b-40913f024775","why-routing-belongs-at-the-center-of-model-serving-zh","為什麼 routing 應該放在 model serving 的中心","Routing 應該是 model serving 的單一入口，因為它能加快模型迭代，也能把服務層變成產品能力的一部分。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1777882252815-32q6.png","2026-05-04T08:10:33.607394+00:00"]