Cheng Jin
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
Towards Architectural-Level Safety Control: Relocating Hazardous Knowledge into Removable Experts in Mixture-of-Experts Models
According to the abstract, Cheng Jin, Zhongtian Ma, Xiao Wang and colleagues propose REX, which adds new removable experts to a mixture-of-experts model and concentrates hazardous knowledge there, so excising the experts removes it. They contrast this with unlearning that suppresses knowledge still recoverable by fine-tuning.