Hadn’t seen that before! Seems very in line with what with the broader points about regularization. In table 4 they show faster convergence in 200 epochs when used alongside REPA. I’d be curious to see if it ended up beating REPA by itself with full 800 epochs of training — or if something about this new latent space, leads to plateauing itself (learns faster but caps out on expressivity). We’ve seen that phenomena before in other situations (eg UNET learns faster than DiT because of convolutions, but stops learning beyond a certain point).
Comments
Nice summary! I missed the mention of EQ-VAE when it comes to generation quality. Tiny trick, huge impact! Have you tried it?
Hadn’t seen that before! Seems very in line with what with the broader points about regularization. In table 4 they show faster convergence in 200 epochs when used alongside REPA. I’d be curious to see if it ended up beating REPA by itself with full 800 epochs of training — or if something about this new latent space, leads to plateauing itself (learns faster but caps out on expressivity). We’ve seen that phenomena before in other situations (eg UNET learns faster than DiT because of convolutions, but stops learning beyond a certain point).