Jiang, Wenzheng, Ke Liang, Xuankun Rong, Jingxuan Zhou, Zhengyi Zhong, Guancheng Wan, and Ji Wang. “PurMM: Attention-Guided Test-Time Backdoor Purification in Multimodal Large Language Models”. Proceedings of the AAAI Conference on Artificial Intelligence 40, no. 42 (March 14, 2026): 35562–35570. Accessed May 14, 2026. https://ojs.aaai.org/index.php/AAAI/article/view/40867.