Research map / Image generation / Text-to-image diffusion
Text-Conditioned Background Generation for Editable Multi-Layer Documents
Taewon Kang, Joseph K J, Christopher Tensmeyer, Jihyung Kil, Wanrong Zhu, Ming C Lin, Vlad Morariu
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 latent masking (ablation) | Ours · Synthetic doc benchmark | OCR accuracy: 0.9085 -> 0.9665 | Improved |
| add latent masking (ablation) | Ours · Synthetic doc benchmark | WCAG coverage (%): 99.67 -> 99.75 | No clear effect |
| add latent masking (ablation) | Ours · Synthetic doc benchmark | CLIP prompt score: 0.2542 -> 0.2374 | Hurt |
| add ARO (ablation) | Ours · Synthetic doc benchmark | WCAG coverage (%): 97.35 -> 99.75 | Improved |
| add ARO (ablation) | Ours · Synthetic doc benchmark | OCR accuracy: 0.9012 -> 0.9665 | Improved |
| add multi-page consistency (LLM summarization+instruction) (ablation) | Ours · Synthetic doc benchmark | CLIP MP consistency: 0.6420 -> 0.6955 | Improved |
| add multi-page consistency (ablation) | Ours · Synthetic doc benchmark | WCAG coverage (%): 99.69 -> 99.75 | No clear effect |
| ours vs BAGEL (latent masking + ARO + MPC) | Ours vs BAGEL · Synthetic doc benchmark | WCAG coverage (%): 66.98 -> 99.75 | Improved |
| ours vs GPT-5 | Ours vs GPT-5 · Synthetic doc benchmark | OCR accuracy: 0.5225 -> 0.9665 | Improved |
| ours vs GPT-5 | Ours vs GPT-5 · Synthetic doc benchmark | CLIP MP consistency: 0.6870 vs 0.6955 | No clear effect |
Closest papers by meaning
- TextCenGen: Attention-Guided Text-Centric Background Adaptation for Text-to-Image Generation (ICML 2025 · similarity 0.87)
- Style-Editor: Text-driven Object-centric Style Editing (CVPR 2025 · similarity 0.84)
- Energy-Guided Optimization for Personalized Image Editing with Pretrained Text-to-Image Diffusion Models (AAAI 2025 · similarity 0.84)
- Improving Editability in Image Generation with Layer-wise Memory (CVPR 2025 · similarity 0.84)
- AnyDoc: Enhancing Document Generation via Large-Scale HTML/CSS Data Synthesis and Height-Aware Reinforcement Optimization (CVPR 2026 · similarity 0.83)
- LayerCraft: Enhancing Text-to-Image Generation with CoT Reasoning and Layered Object Integration (NeurIPS 2025 · similarity 0.83)
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.
