Residual Decoder Adapter upgrades existing visual tokenizers for AR image models by adding a paired codebook and pixel-space residual branch, raising OCR accuracy on text rendering benchmarks from ~12-24% to ~36-58%.
Stylear: Customizing multimodal autoregressive model for style-aligned text-to-image genera- tion.arXiv preprint arXiv:2505.19874, 2025
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
-
Residual Decoder Adapter: ID-Preserving Tokenizer Adaption for Autoregressive Text Rendering
Residual Decoder Adapter upgrades existing visual tokenizers for AR image models by adding a paired codebook and pixel-space residual branch, raising OCR accuracy on text rendering benchmarks from ~12-24% to ~36-58%.