EXCLAIM is a novel explainable agentic system that detects Out-of-Context misinformation by analyzing inconsistencies between images and text.

• Utilizes a hierarchical retrieval framework to detect subtle misalignments between visual content and textual narratives
• Employs multiple specialist agents to analyze content from different perspectives
• Provides transparent explanations for its detection decisions, enhancing trust and understanding
• Demonstrates superior performance compared to existing systems, especially on challenging misinformation examples

This research addresses critical security challenges in our information ecosystem by providing powerful tools to combat sophisticated misinformation that pairs authentic images with misleading text, helping protect public discourse and decision-making.

EXCLAIM: An Explainable Cross-Modal Agentic System for Misinformation Detection with Hierarchical Retrieval