We sincerely thank all three reviewers for their valuable comments, with the following being our responses. As such, although the weights in Eq. (2) are learned to be'static', the enriched The sample sizes for all the models are fixed as 4096. R1) Regarding the advance of the proposed model. We will include such experiments in our revised paper. R1) Regarding missing related work.

We thank all the reviewers for your time in reviewing this paper and also for your suggestive comments. We will add detailed remarks if accepted. Y es, they are the same as the metrics used in the multi-label setting. Some typos to be corrected. They follow the same settings with the simulated study.