Research map / Image generation / Fast diffusion sampling
Jumping the Landing Phase: Noise Variance Matching Enables Accurate Few-Step Inversion
Yang Luo, Zhineng Chen, Ya Gao, Xieping Gao, 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 NVM variance correction instead of global scaling (ablation) | DDIM inversion · HPDv2 (5-step) | CLIP-I: 77.45 -> 84.64 | Improved |
| add NVM variance correction to DDIM inversion (ablation) | DDIM inversion · HPDv2 (5-step) | CLIP-I: 67.18 -> 84.64 | Improved |
| add NVM variance correction to DDIM inversion (ablation) | DDIM inversion · HPDv2 (5-step) | LPIPS: 0.4687 -> 0.2159 | Improved |
| add global scaling of predicted noise to DDIM inversion (ablation) | DDIM inversion · HPDv2 (5-step) | CLIP-I: 67.18 -> 77.45 | Improved |
| add attention-guided spatial mask to NVM (ablation) | NVM · HPDv2 (5-step) | PSNR: 20.53 -> 21.04 | Improved |
| add attention-guided spatial mask to NVM (ablation) | NVM · HPDv2 (5-step) | LPIPS: 0.2159 -> 0.2692 | Hurt |
| add one ReNoise refinement iteration after NVM+mask (ablation) | NVM · HPDv2 (5-step) | CLIP-I: 84.48 -> 88.81 | Improved |
| add second ReNoise refinement iteration (ablation) | NVM · HPDv2 (5-step) | CLIP-I: 88.81 -> 93.55 | Improved |
| replace ReNoise with NVM | SDXL inversion · HPDv2 (5-step) | CLIP-I: 88.97 -> 93.55 | Improved |
| replace ReNoise with NVM | SDXL inversion · MS-COCO (5-step) | CLIP-I: 80.32 -> 90.27 | Improved |
| apply NVM instead of ReNoise on SD1.5 | SD1.5 inversion · HPDv2 (5-step) | CLIP-I: 85.77 -> 94.10 | Improved |
| apply NVM instead of ReNoise on SDXL Turbo | SDXL Turbo inversion · HPDv2 (5-step) | CLIP-I: 81.18 -> 80.35 | Mixed |
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.
- EDICT — EDICT: Exact Diffusion Inversion via Coupled Transformations (CVPR 2023) · better on 5 of 5 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
- EasyInv: Toward Fast and Better DDIM Inversion (ICML 2025 · similarity 0.89)
- Adapt and Diffuse: Sample-adaptive Reconstruction via Latent Diffusion Models (ICML 2024 · similarity 0.88)
- Divide-and-Conquer Posterior Sampling for Denoising Diffusion priors (NeurIPS 2024 · similarity 0.88)
- BELM: Bidirectional Explicit Linear Multi-step Sampler for Exact Inversion in Diffusion Models (NeurIPS 2024 · similarity 0.88)
- Provable Posterior Sampling with Denoising Oracles via Tilted Transport (NeurIPS 2024 · similarity 0.88)
- Improving Diffusion Inverse Problem Solving with Decoupled Noise Annealing (CVPR 2025 · 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.
