Research map / Image generation / Editing and style transfer
Learning an Image Editing Model without Image Editing Pairs
Nupur Kumari, Sheng-Yu Wang, Nanxuan Zhao, Yotam Nitzan, Yuheng Li, Krishna Kumar Singh, Richard Zhang, Eli Shechtman, Jun-Yan Zhu, Xun Huang
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 VLM-based editing loss to DMD (ablation) | NP-Edit · GEdit-Bench (4-step) | VIEScore Overall (x10): 4.93 -> 6.10 | Improved |
| add DMD loss to VLM loss (ablation) | NP-Edit · GEdit-Bench (4-step) | VIEScore Overall (x10): 1.93 -> 6.10 | Improved |
| add identity-preservation question to VLM loss (ablation) | NP-Edit · GEdit-Bench (4-step) | VIEScore Overall (x10): 5.76 -> 6.10 | Improved |
| use binary Yes/No cross-entropy instead of full-vocabulary CE (ablation) | NP-Edit · GEdit-Bench (4-step) | VIEScore Overall (x10): 5.89 -> 6.10 | Improved |
| add LPIPS loss for input-edit alignment (ablation) | NP-Edit · GEdit-Bench (4-step) | VIEScore Overall (x10): 6.10 -> 4.52 | Hurt |
| add cycle loss for input-edit alignment (ablation) | NP-Edit · GEdit-Bench (4-step) | VIEScore Overall (x10): 6.10 -> 5.21 | Hurt |
| larger training dataset (number of reference images) (ablation) | NP-Edit · GEdit-Bench (4-step) | VIEScore | Improved |
| larger VLM backbone as reward (ablation) | NP-Edit · GEdit-Bench (4-step) | VIEScore | Improved |
| use VLM-gradient training instead of SFT+Flow-GRPO | NP-Edit · GEdit-Bench (4-step) | VIEScore | Improved |
| NP-Edit vs few-step baselines | NP-Edit 2B · GEdit-Bench (4-step) | VIEScore Overall (x10): 6.10 vs 6.06 (Qwen-Image-Edit 20B) | Improved |
Closest papers by meaning
- Instruct-CLIP: Improving Instruction-Guided Image Editing with Automated Data Refinement Using Contrastive Learning (CVPR 2025 · similarity 0.89)
- Exploring Low-Dimensional Subspace in Diffusion Models for Controllable Image Editing (NeurIPS 2024 · similarity 0.89)
- PairEdit: Learning Semantic Variations for Exemplar-based Image Editing (NeurIPS 2025 · similarity 0.88)
- RelationAdapter: Learning and Transferring Visual Relation with Diffusion Transformers (NeurIPS 2025 · similarity 0.88)
- ImgEdit: A Unified Image Editing Dataset and Benchmark (NeurIPS 2025 · similarity 0.88)
- Learning Where to Edit Vision Transformers (NeurIPS 2024 · similarity 0.88)
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.
