Finding Skill Neurons in Pre-trained Transformer-based Language Models

Wang, Xiaozhi, Wen, Kaiyue, Zhang, Zhengyan · arXiv · 2022