Research map / Image generation / Text-to-image diffusion
GlyphPrinter: Region-Grouped Direct Preference Optimization for Glyph-Accurate Visual Text Rendering
Xincheng Shuai, Ziye Li, Henghui Ding, Dacheng Tao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add R-GDPO preference stage after Stage 1 fine-tuning (ablation) | GlyphPrinter · GlyphAcc-Complex | Sentence accuracy: 0.6981 -> 0.8349 | Improved |
| add R-GDPO preference stage after Stage 1 fine-tuning (ablation) | GlyphPrinter · GlyphAcc-Multilingual | Sentence accuracy: 0.8023 -> 0.8932 | Improved |
| use R-GDPO instead of SFT on GlyphCorrector (ablation) | GlyphPrinter · GlyphAcc-Complex | Sentence accuracy: 0.7406 -> 0.8349 | Improved |
| use R-GDPO instead of Mask-SFT (correct regions only) (ablation) | GlyphPrinter · GlyphAcc-Complex | Sentence accuracy: 0.7547 -> 0.8349 | Improved |
| apply plain SFT in Stage 2 (vs Stage 1 only) (ablation) | GlyphPrinter · GlyphAcc-Complex | Sentence accuracy: 0.6981 -> 0.7406 | Improved |
| add Regional Reward Guidance at inference (ablation) | GlyphPrinter · GlyphAcc-Complex | Sentence accuracy: 0.8019 -> 0.8349 | Improved |
| add Regional Reward Guidance at inference (ablation) | GlyphPrinter · GlyphAcc-Multilingual | Sentence accuracy: 0.8772 -> 0.8932 | Improved |
| combine inter- and intra-sample losses (lambda_inter 0.7) vs inter only (ablation) | GlyphPrinter · GlyphAcc-Complex | Sentence accuracy: 0.8066 -> 0.8349 | Improved |
| combine inter- and intra-sample losses (lambda_inter 0.7) vs intra only (ablation) | GlyphPrinter · GlyphAcc-Multilingual | Sentence accuracy: 0.8458 -> 0.8932 | Improved |
| use GlyphPrinter vs Qwen-Image | GlyphPrinter vs Qwen-Image · Image aesthetic (user study) | Aes: 7.9143 vs 8.0749 | Mixed |
Closest papers by meaning
- GlyphAnchor: Enhancing Visual Text Rendering via Position-Anchored Glyph Priors (NeurIPS 2026 · similarity 0.90)
- GlyphMastero: A Glyph Encoder for High-Fidelity Scene Text Editing (CVPR 2025 · similarity 0.89)
- GlyphSR: A Simple Glyph-Aware Framework for Scene Text Image Super-Resolution (AAAI 2025 · similarity 0.88)
- VecGlypher: Unified Vector Glyph Generation with Language Models (CVPR 2026 · similarity 0.88)
- GlyphDraw2: Automatic Generation of Complex Glyph Posters with Diffusion Models and Large Language Models (AAAI 2025 · similarity 0.87)
- TextCtrl: Diffusion-based Scene Text Editing with Prior Guidance Control (NeurIPS 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
