Tag
Researchers from Kennesaw State University investigate cross-prompt generalization in detecting AI-generated fake news using interpretable linguistic features (lexical diversity, readability, emotion). A random forest classifier trained on one prompting strategy and tested on another achieves AUC values of 0.988–1.000, suggesting these features capture stable, generalizable properties of AI-generated text.
AEyeDE is an attention-based attribution framework that uses a proxy Transformer model to extract attention maps from text and trains a lightweight CNN to distinguish human-written from AI-generated text, outperforming text-only baselines and showing robustness across settings.