Jin, H.; Li, Y.; Fan, H.; Shen, L.; Li, X.; Li, B. Uncovering and Aligning Anomalous Attention Heads to Defend Against NLP Backdoor Attacks. AAAI 2026, 40, 37472-37480.