Constrained Optimization with Dynamic Bound-scaling for Effective NLP Backdoor DefenseDownload PDFOpen Website

2022 (modified: 05 Nov 2022)ICML 2022Readers: Everyone
Abstract: Modern language models are vulnerable to backdoor attacks. An injected malicious token sequence (i.e., a trigger) can cause the compromised model to misbehave, raising security concerns. Trigger in...
0 Replies

Loading