Research map / Image generation / Editing and style transfer
WeEdit: A Dataset, Benchmark and Glyph-Guided Framework for Text-centric Image Editing
Hui Zhang, Juntao Liu, Zongkai Liu, liqiang niu, Fandong Meng, Zuxuan Wu, Yu-Gang Jiang
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 glyph-guided SFT on WeEdit data | Qwen-Image-Edit-2509 · WeEdit Bilingual Benchmark | Overall IA: 3.49 -> 6.99 | Improved |
| add RL stage with multi-dimensional rewards to SFT | WeEdit-SFT · WeEdit Bilingual Benchmark | Overall IA: 6.99 -> 7.47 | Improved |
| add RL stage with multi-dimensional rewards to SFT | WeEdit-SFT · WeEdit Bilingual Benchmark | Overall TC: 7.33 -> 8.19 | Improved |
| add full WeEdit (SFT+RL) to base model | Qwen-Image-Edit-2509 · WeEdit Multilingual Benchmark | Overall IA/TC/BP: +3.07/+1.28/+1.82 | Improved |
| add full WeEdit (SFT+RL) to base model | Qwen-Image-Edit-2509 · WeEdit Bilingual Benchmark | Overall IA/TC/BP: +3.98/+2.35/+2.21 | Improved |
| add full WeEdit to base on Translate and Rearrange | Qwen-Image-Edit-2509 · WeEdit Bilingual Benchmark | Translate IA: 1.26 -> 6.36 | Improved |
| add full WeEdit to base on Rearrange | Qwen-Image-Edit-2509 · WeEdit Bilingual Benchmark | Rearrange IA: 1.35 -> 7.82 | Improved |
| apply glyph guidance at inference only, without training (ablation) | Qwen-Image-Edit-2509 · WeEdit ablation | IA/TC/BP | No clear effect |
| add RL directly on SFT-only model (no glyph) (ablation) | WeEdit-SFT · WeEdit ablation | IA/TC/BP | No clear effect |
| add glyph guidance to SFT (ablation) | WeEdit-SFT · WeEdit ablation | IA/TC/BP | Improved |
| add reference image (RI) to the quality reward (ablation) | WeEdit-RL · WeEdit ablation | IA/TC/BP | Improved |
| use separate reward models per dimension instead of a single one (ablation) | WeEdit-RL · WeEdit ablation | IA/TC/BP | Improved |
Closest papers by meaning
- InsightEdit: Towards Better Instruction Following for Image Editing (CVPR 2025 · similarity 0.88)
- ImgEdit: A Unified Image Editing Dataset and Benchmark (NeurIPS 2025 · similarity 0.88)
- UltraEdit: Instruction-based Fine-Grained Image Editing at Scale (NeurIPS 2024 · similarity 0.88)
- PairEdit: Learning Semantic Variations for Exemplar-based Image Editing (NeurIPS 2025 · similarity 0.88)
- AnyEdit: Mastering Unified High-Quality Image Editing for Any Idea (CVPR 2025 · similarity 0.87)
- SteerVTE: Seamless Video Text Editing with Style and Glyph Control (NeurIPS 2026 · similarity 0.87)
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.
