Research map / Image generation / Text-to-image diffusion
InverFill: One-Step Inversion for Enhanced Few-Step Diffusion Inpainting
Duc Vu, Kien Nguyen, Trong-Tung Nguyen, Ngan Nguyen, Phong Nguyen, Khoi Nguyen, Cuong Pham, Anh Tran
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 InverFill inversion noise | SANA-Sprint 0.6B (2 NFEs) blended sampling · BrushBench | ImageReward x10: 10.73 -> 11.76 | Improved |
| add InverFill inversion noise | SANA-Sprint 0.6B (2 NFEs) blended sampling · MagicBrush | ImageReward x10: 2.65 -> 3.07 | Improved |
| add InverFill inversion noise | SDXL Turbo (4 NFEs) blended sampling · BrushBench | CLIP similarity: 27.21 -> 27.74 | Improved |
| add InverFill to the specialized BrushNet pipeline | SDXL Turbo + BrushNet · BrushBench | ImageReward x10: 12.30 -> 12.34 | Improved |
| add InverFill to the specialized BrushNet pipeline | SDXL Turbo + BrushNet · MagicBrush | ImageReward x10: 4.31 -> 4.31 | No clear effect |
| add Re-Blending (ablation) | InverFill on SANA-Sprint 0.6B · BrushBench (SANA-Sprint 0.6B, 2 NFEs) | ImageReward x10: 7.93 -> 11.22 | Improved |
| add Gaussian regularization (ablation) | InverFill on SANA-Sprint 0.6B · BrushBench (SANA-Sprint 0.6B, 2 NFEs) | ImageReward x10: 11.22 -> 11.51 | Improved |
| add LADD adversarial loss (ablation) | InverFill on SANA-Sprint 0.6B · BrushBench (SANA-Sprint 0.6B, 2 NFEs) | ImageReward x10: 11.51 -> 11.76 | Improved |
| add Re-Blending (ablation) | InverFill on SDXL-Turbo · BrushBench (SDXL-Turbo, 4 NFEs) | ImageReward x10: 10.60 -> 11.29 | Improved |
| add Gaussian regularization (ablation) | InverFill on SDXL-Turbo · BrushBench (SDXL-Turbo, 4 NFEs) | ImageReward x10: 11.29 -> 12.10 | Improved |
| replace SDS regularization with Gaussian moment-matching loss (ablation) | InverFill on SANA-Sprint 0.6B · BrushBench (SANA-Sprint 0.6B, 2 NFEs) | ImageReward x10: 11.29 -> 11.51 | Improved |
| replace 50-step DDIM inversion (with blending) by one-step InverFill (ablation) | SDXL-Turbo blended sampling · BrushBench (SDXL-Turbo, 4 NFEs) | ImageReward x10: 12.07 -> 12.34 | Improved |
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.
- SANA-Sprint 0.6B — SANA-Sprint: One-Step Diffusion with Continuous-Time Consistency Distillation (ICCV 2025) · better on 1 of 1 shared result
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
- Image Inpainting via Tractable Steering of Diffusion Models (ICLR 2024 · similarity 0.90)
- RAD: Region-Aware Diffusion Models for Image Inpainting (CVPR 2025 · similarity 0.90)
- TurboFill: Adapting Few-step Text-to-image Model for Fast Image Inpainting (CVPR 2025 · similarity 0.89)
- Inversion-Free Image Editing with Language-Guided Diffusion Models (CVPR 2024 · similarity 0.88)
- One Stone with Two Birds: A Null-Text-Null Frequency-Aware Diffusion Models for Text-Guided Image Inpainting (NeurIPS 2025 · similarity 0.88)
- TD-Paint: Faster Diffusion Inpainting Through Time-Aware Pixel Conditioning (ICLR 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.
