Fetching the paper…

The shape and simplicity biases of adversarially robust ImageNet-trained CNNs · Around