Fetching the paper…

Confidence-Calibrated Adversarial Training: Generalizing to Unseen Attacks · Around