[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"topic-zh-brrl-ppo-bpo-gbpo-":3},{"cluster":4,"timeline":16},{"id":5,"slug":6,"title":7,"pinned":8,"status":9,"summary":10,"category":11,"language":12,"created_at":13,"merged_into":14,"article_count":15,"first_seen_at":13,"last_updated_at":13},"d68f0236-14c5-4881-bef2-4e02edc5c9e3","brrl-ppo-bpo-gbpo-","BRRL 取代 PPO 剪裁：BPO 與 GBPO 的穩定性升級",false,"active","BRRL 把 PPO 的剪裁目標改寫成有界比例框架，推出 BPO 與 GBPO，主打更穩定的更新與更清楚的理論基礎。","research","zh","2026-07-05T09:04:30.045021+00:00",null,1,[17],{"id":18,"slug":19,"title":7,"summary":10,"category":11,"image_url":20,"cover_image":20,"published_at":21,"is_canonical_seed":22},"7a04d752-3f1a-4df7-b7c5-8bcb1e69c565","bounded-ratio-reinforcement-learning-ppo-zh","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1776751794578-t5j7.png","2026-04-21T06:09:39.421+00:00",true]