TIDE: Making Task-Agnostic Backdoors Harder to Erase in Pre-trained Language Models.

Yan Meng, Zhigang Lu, Bing Li, Anan Du, Shuchao Pang

Browse the full ACISP paper archive.