CRAVE is a novel framework that combines retrieval-augmented LLMs with clustering techniques to revolutionize fact-checking on social media platforms.

• Automatically retrieves and organizes multimodal evidence from diverse, often contradictory sources
• Clusters evidence into coherent narratives that represent different perspectives on a claim
• Employs an LLM-based judge to deliver fact-checking verdicts with supporting evidence summaries
• Creates a decision-support tool for professional fact-checkers to combat misinformation at scale

This research addresses critical security concerns by providing transparent, evidence-based verification of viral social media content, helping combat the spread of false information in digital ecosystems.

Fact-Checking with Contextual Narratives: Leveraging Retrieval-Augmented LLMs for Social Media Analysis