Why add same noise to pairwise images?

[ChenDRAG]

Hi, thank you for the efforts in open-source the paper code. When looking into the code, I found it manually modified the noise added to the pairwise images to be the same.

DiffusionDPO/train.py

Line 1053 in f3ca6f2

if args.train_method == 'dpo': # make timesteps and noise same for pairs in DPO

I'm wondering why this needs to be done. If this is necessary by theory?

[LeonNerd]

From my understanding of the paper, the formula in the paper has reflected this point. From the point of view of training, at tempstep n, how should we judge which picture is good, that is, adding the same noise to the pair of pictures at time t