This is a Plain English Papers summary of a research paper called AI Image Detector LEGION Spots Fakes with 89% Accuracy While Explaining Why They're Synthetic. If you like these kinds of analysis, you should join AImodels.fyi or follow us on Twitter.

LEGION: Learning to Ground and Explain for Synthetic Image Detection

Overview

  • LEGION combines synthetic image detection with explainable AI
  • Uses multimodal large language models to identify AI-generated images
  • Provides visual evidence (grounding) and natural language explanations
  • Achieves 89% detection accuracy on diverse synthetic image datasets
  • Offers detailed explanations that align with human rationales
  • Novel approach that balances detection performance with interpretability

Plain English Explanation

LEGION is a new system that spots fake images while explaining why they're fake. Think of it like a detective who not only identifies counterfeit money but also points out exactly what features gave it away.

Most current AI detectors just give you a yes/no answer about whether...

Click here to read the full summary of this paper