Research map / Image generation / Text-to-image diffusion
Deconstructing Guidance: A Semantic Hierarchy for Precise Diffusion Model Editing
Wootaek Jeong, Junghyo Sohn, Jee Seok Yoon, Heung-Il Suk
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 Prism-Edit guidance decomposition | PnP · Wild-TI2I background subset | DINO/SSIM: 1.2240 -> 1.7555 | Improved |
| add Prism-Edit guidance decomposition | DDIM inversion · Wild-TI2I background subset | DINO/SSIM: 0.9796 -> 1.0731 | Improved |
| add Prism-Edit guidance decomposition | LEDITS++ · Wild-TI2I background subset | DINO/SSIM: 1.0317 -> 1.2328 | Improved |
| add Prism-Edit guidance decomposition | DDIM inversion · Wild-TI2I object subset | SSIM: 0.4588 -> 0.6942 | Improved |
| add Prism-Edit guidance decomposition | PnP · Wild-TI2I object subset | SSIM: 0.5554 -> 0.6952 | Improved |
| add Prism-Edit guidance decomposition | DDPM inversion · Wild-TI2I object subset | SSIM: 0.7721 -> 0.7722 | No clear effect |
| add Prism-Edit guidance decomposition | DDIM inversion · Wild-TI2I object subset | CLIP score: 0.3101 -> 0.2997 | Hurt |
| add Prism-Edit guidance decomposition | LEDITS++ · ImageNet-R-TI2I | DINO/SSIM: 0.8426 -> 0.9982 | Improved |
| add Prism-Edit guidance decomposition | DDIM inversion · ImageNet-R-TI2I | CLIP score: 0.3242 -> 0.3199 | Hurt |
| add Prism-Edit guidance decomposition | DDIM inversion · CycleGAN Summer to Winter (valid-text) | LPIPS x100: 62.76 -> 28.39 | Improved |
| add Prism-Edit guidance decomposition | DDPM inversion · CycleGAN Summer to Winter (null-text) | LPIPS x100: 33.50 -> 33.42 | No clear effect |
| edit intermediate sigma intervals instead of the distribution tails of the semantic map (ablation) | Prism-Edit · Qualitative layer selection | semantic leakage | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LEDITS++ — LEDITS++: Limitless Image Editing using Text-to-Image Models (CVPR 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Towards Understanding Cross and Self-Attention in Stable Diffusion for Text-Guided Image Editing (CVPR 2024 · similarity 0.90)
- Drag Your Noise: Interactive Point-based Editing via Diffusion Semantic Propagation (CVPR 2024 · similarity 0.89)
- Exploring Low-Dimensional Subspace in Diffusion Models for Controllable Image Editing (NeurIPS 2024 · similarity 0.89)
- Localizing and Editing Knowledge In Text-to-Image Generative Models (ICLR 2024 · similarity 0.89)
- FunEditor: Achieving Complex Image Edits via Function Aggregation with Diffusion Models (AAAI 2025 · similarity 0.88)
- Image Editing As Programs with Diffusion Models (NeurIPS 2025 · 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.
