Adversarial Examples

Inputs specifically designed to fool AI models into making mistakes. Important for understanding model robustness and security.