Wei Qin
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
Revisiting Over-refusal in Safety-aligned Large Language Models: A Representation and Optimization Perspective
According to the abstract, Chao Chen, Pengyang Shao, Wei Qin and three co-authors study why safety-trained language models refuse harmless but risky-looking requests. They report that harmful and over-refused prompts overlap in deeper layers and share gradient directions, and propose a method named GOAL.