FakeReasoning introduces a novel approach to detect AI-generated images with higher accuracy and better generalization across different generation models.

• Creates a unified reasoning framework that detects forgeries and explains its decisions
• Achieves 86.3% accuracy on cross-model generalization tests
• Incorporates explicit logical reasoning chains rather than just highlighting suspicious areas
• Demonstrates enhanced performance on real-world scenarios where generation models are unknown

This research addresses critical security challenges by providing robust tools to identify AI-generated content that could be used for misinformation campaigns or fraud, helping organizations maintain digital trust.

FakeReasoning: Towards Generalizable Forgery Detection and Reasoning