StyleVAR performs controllable style transfer by autoregressively modeling VQ-VAE tokens with a transformer that blends style and content via scale-dependent cross-attention, trained in two stages with supervised fine-tuning and GRPO reinforcement learning, outperforming AdaIN on multiple metrics.
Title resolution pending
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
-
StyleVAR: Controllable Image Style Transfer via Visual Autoregressive Modeling
StyleVAR performs controllable style transfer by autoregressively modeling VQ-VAE tokens with a transformer that blends style and content via scale-dependent cross-attention, trained in two stages with supervised fine-tuning and GRPO reinforcement learning, outperforming AdaIN on multiple metrics.