This research demonstrates that Large Visual-Language Models (LVLMs) can effectively detect fake news by analyzing both textual and visual content in context.

• LVLMs outperform traditional LLMs in multimodal fake news classification tasks
• In-context learning approach eliminates the need for expensive fine-tuning
• Models can analyze how images and text interact to identify misleading content
• Research provides a cost-effective security solution for combating misinformation

This advancement offers security professionals a powerful tool to automatically screen content across platforms, protecting users from visual-textual misinformation campaigns without requiring specialized model training.

Large Visual-Language Models Are Also Good Classifiers: A Study of In-Context Multimodal Fake News Detection