Replaying VAE-encoded latents instead of full images reduces catastrophic forgetting and mode collapse in sequential fine-tuning of text-to-image diffusion models.
Bermano, Gal Chechik, and Daniel Cohen-Or
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Mitigating Catastrophic Forgetting and Mode Collapse in Text-to-Image Diffusion via Latent Replay
Replaying VAE-encoded latents instead of full images reduces catastrophic forgetting and mode collapse in sequential fine-tuning of text-to-image diffusion models.