Picture for Jiafei Niu

Jiafei Niu

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

Add code
Jan 06, 2025
Figure 1 for MBTSAD: Mitigating Backdoors in Language Models Based on Token Splitting and Attention Distillation
Figure 2 for MBTSAD: Mitigating Backdoors in Language Models Based on Token Splitting and Attention Distillation
Figure 3 for MBTSAD: Mitigating Backdoors in Language Models Based on Token Splitting and Attention Distillation
Figure 4 for MBTSAD: Mitigating Backdoors in Language Models Based on Token Splitting and Attention Distillation
Viaarxiv icon