Fetching the paper…

Robustifying Vision Transformer without Retraining from Scratch by Test-Time Class-Conditional Feature Alignment · Around