Learning from Interaction Experience via Localized Context Distillation

Zhezheng Hao, Yongkang Yang, Xilin Xia +9

Learning from interaction experience is essential for enabling large language models to continually improve during deployment. Context distillation offers a promising path: privile…