Query Regarding Multiphase Number in Training

[LeiJiang1]

Dear author, I have a question regarding the multiphase number used in your study. Since the multiphase number can be set differently during training and significantly impacts the test results, could you kindly clarify the specific multiphase number you used during training to achieve the results reported in your paper?

In my own experiments, I observed some intriguing behaviors:

When using multiphase = 2 during training, the test results with inference = 1 are suboptimal, whereas inference = 2 produces much better outcomes.
However, training with multiphase = 1 yields good results when testing with inference = 1. Yet, this seems to forgo the benefits of your proposed method.
I would greatly appreciate your insights into this matter. Thank you for your time, and I look forward to your response.