AI RESEARCH
Boosting Latent Diffusion Models via Disentangled Representation Alignment
arXiv CS.CV
•
ArXi:2601.05823v2 Announce Type: replace Latent Diffusion Models (LDMs) rely heavily on the compressed latent space provided by Variational Autoencoders (VAEs) for high-quality image generation. Recent studies have attempted to obtain generation-friendly VAEs by directly adopting alignment strategies from LDM