Post-Training 范式

Mixed RL

Model fusion

• https://www.zhihu.com/search?type=content&q=model%20fusion
• https://www.zhihu.com/question/1953499475269101203/answer/1954651725777602425

美团：Model fusion DeepSeek/GLM: distill/Mixed RL MiMO: opd

参考：

• https://tokens-for-thoughts.notion.site/post-training-101#262b8b68a46d80ca97a0fd72fc4bdf61