Pure leakage in the test set

[FudaMingdaXsc]

It seems that you used the entire test set to train encode_text_learn, while also using the test set for testing, which is seriously inconsistent with the ZERO-SHOT mentioned in your paper.Can you explain this issue

[zqhang]

We conduct zero-shot performance evaluations on datasets other than MVTEC AD using the test set of MVTEC AD dataset. And, we assess the zero-shot performance on the MVTEC AD dataset using the test set of VisA dataset. This aligns with the approach presented in our paper. You can check the script train.sh and test.sh

[FudaMingdaXsc]

Thank you very much for your reply. Your explanation convinced me that your ZERO-SHOT approach is different from others' ZERO-SHOT approach. It's a great idea and worth following by others

[zqhang]

Thanks for your acknowledgment, I will close this issue.