VAE-LFA suppresses semantic drift in multi-turn DiT image editing by low-pass filtering latent discrepancies and aligning low-frequency components to an EMA of previous rounds in VAE space.
The visual semantics depend strongly on the broader environment, background, or spatial composition rather than a single foreground object
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Why Do DiT Editors Drift? Plug-and-Play Low Frequency Alignment in VAE Latent Space
VAE-LFA suppresses semantic drift in multi-turn DiT image editing by low-pass filtering latent discrepancies and aligning low-frequency components to an EMA of previous rounds in VAE space.