HintsOfTruth provides a breakthrough dataset combining 27K real and synthetic image/claim pairs to help automate the first critical step in fact-checking: identifying claims worth checking.

• First comprehensive multimodal dataset combining images and text for checkworthiness detection
• Includes both real-world and synthetic claims to improve robustness of detection methods
• Designed to address challenges across diverse content domains
• Creates a foundation for scaling fact-checking efforts through automation

This research directly enhances information security by providing tools to identify potentially harmful misinformation before it spreads, helping organizations and platforms to prioritize verification resources more effectively.

HintsOfTruth: A Multimodal Checkworthiness Detection Dataset with Real and Synthetic Claims