Tag
This paper presents an adversarial methodology for creating and detecting AI-generated social bot content, curating a multilingual, cross-platform dataset of paired human and AI messages. Training on this adversarial data yields detection that significantly outperforms existing content-based bot detection models in real-world settings.