Trading inference-time compute for adversarial robustness

The article discusses research on trading computational resources during inference time to improve adversarial robustness in AI systems. This approach explores how allocating more compute power at inference can enhance model security against adversarial attacks.