xingyizhao / PURE

Code associated with ICML (2024). "Defense against Backdoor Attack on Pre-trained Language Models via Head Pruning and Attention Normalization"

4 stars 0 forks source link

readme

PURE

Code associated with Proceedings of the 41 st International Conference on Machine Learning. "Defense against Backdoor Attack on Pre-trained Language Models via Head Pruning and Attention Normalization"