Phi: Preference Hijacking in Multi-modal Large Language Models at Inference Time.

Yifan Lan, Yuanpu Cao, Weitong Zhang, Lu Lin, Jinghui Chen

Browse the full EMNLP paper archive.