I remember that DDBM uses VQGAN. Why isn't it used here?

[inor1yq]

In your paper, it states alignment with BBDM. Why wasn‘t VQGAN employed?

[zhengkw18]

Could you specify where DDBM uses VQGAN? It should be a continuous-space diffusion bridge model and not rely on discrete tokenizers.

[yxc1234567539]

bbdm and ddbm