Dissecting Misalignment of Multimodal Large Language Models via Influence Function

Open in new window