MBTSAD: Mitigating Backdoors in Language Models Based on Token Splitting and Attention Distillation.

Yidong Ding, Jiafei Niu, Ping Yi

Browse the full ICTAI paper archive.