accuracy comparison number wrong?

[rgeirhos]

In https://distill.pub/2019/advex-bugs-discussion/response-2/ it is stated that "testing on the CIFAR test set incurs an accuracy of 23.5% (out of 88%)."

When compared to Table 1 of the original paper (https://arxiv.org/pdf/1905.02175.pdf), it should be "out of 63%" if I'm not mistaken: the 88% refers to the accuracy of a model trained on ImageNet_R rather than CIFAR-10.

Furthermore, at the bottom of the page, the github issue link leads to a 404 error: https://github.com/distillpub/post--ilyas-response-2/issues/new which could be fixed by updating the link and instead referring to https://github.com/distillpub/post--ilyas-responses/issues