reviewers ' comments in detail

We thank the reviewers [R1, R2, R3] for their valuable and constructive comments. We will follow R1's suggestion to clarify this Thank you for the suggestion. Our paper presents an empirical visualization of robust vs. non-robust NNs under adversarial attacks and WRN, ResNet50), which demonstrate that our TLA is not restricted to a single model architecture [R2]. Overall, we observe similar improvements using our proposed TLA method. We apologize for the confusion.