Some information about the variables in the code

[hoomegic]

There are many versions of the LoRD method in the ‘lord_train.py’, such as LoRD-III, LoRD-IV, LoRD-V, etc. What are the differences between these different versions? In addition, the vanilla method is not reflected in the paper. What is this method?

[liangzid]

They are different loss functions we have tested. You can use LoRD-VI, which can be seen as the final version of LoRD. vanilla denotes the MLE in the paper.

Thanks for your feedback. We will clean our repository recently~