Code relative to "Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks"
Why do you think that https://github.com/searchableai/KitanaQA is a good alternative to auto-attack
Code relative to "Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks"
Why do you think that https://github.com/searchableai/KitanaQA is a good alternative to auto-attack