Leveraging Chat-Based Large Vision Language Models for Multimodal Out-Of-Context Detection