Research map / Image generation / Fast diffusion sampling
PnP-CM: Consistency Models as Plug-and-Play Priors for Inverse Problems
Merve Gulle, Junno Yun, Yasar Utku Alcalar, Mehmet Akcakaya
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 noise injection to PnP-CM (ablation) | PnP-CM · CelebA-HQ SR x4 | PSNR: 22.93 -> 24.25 | Improved |
| add momentum to PnP-CM (ablation) | PnP-CM · CelebA-HQ SR x4 | PSNR: 22.93 -> 25.97 | Improved |
| add both noise injection and momentum (ablation) | PnP-CM · CelebA-HQ SR x4 | PSNR: 22.93 -> 27.27 | Improved |
| add noise injection (ablation) | PnP-CM · CelebA-HQ Gaussian deblur | LPIPS: 0.427 -> 0.251 | Improved |
| add momentum without noise injection (ablation) | PnP-CM · CelebA-HQ Gaussian deblur | PSNR: 27.26 -> 26.59 | Hurt |
| add momentum without noise injection (ablation) | PnP-CM · CelebA-HQ inpainting | PSNR: 28.69 -> 28.72 | No clear effect |
| add momentum at low NFE (ablation) | PnP-CM · fastMRI PD R=8, NFE=4 | PSNR: 30.70 -> 32.81 | Improved |
| add momentum at high NFE (ablation) | PnP-CM · fastMRI PD R=8, NFE=20 | PSNR: 32.86 -> 32.81 | No clear effect |
| increase NFE from 2 to 8 (ablation) | PnP-CM · CelebA-HQ SR x4 (sigma 0.05) | PSNR: 26.99 -> 27.37 | Improved |
| halve measurement noise sigma from 0.05 to 0.025 (ablation) | PnP-CM · CelebA-HQ SR x4 (NFE=4) | PSNR: 27.27 -> 27.78 | Improved |
| perturb hyperparameters by up to 25% (ablation) | PnP-CM · CelebA-HQ Gaussian deblur | PSNR | No clear effect |
| use ADMM PnP with CM prior instead of CM4IR pseudo-inverse guidance | CM-based solver · fastMRI Coronal PD-FS R=4 | PSNR: 26.51 -> 32.90 | Improved |
Closest papers by meaning
- InverseBench: Benchmarking Plug-and-Play Diffusion Priors for Inverse Problems in Physical Sciences (ICLR 2025 · similarity 0.90)
- Provably Robust Score-Based Diffusion Posterior Sampling for Plug-and-Play Image Reconstruction (NeurIPS 2024 · similarity 0.88)
- SITCOM: Step-wise Triple-Consistent Diffusion Sampling For Inverse Problems (ICML 2025 · similarity 0.88)
- Prior Mismatch and Adaptation in PnP-ADMM with a Nonconvex Convergence Analysis (ICML 2024 · similarity 0.88)
- DMPlug: A Plug-in Method for Solving Inverse Problems with Diffusion Models (NeurIPS 2024 · similarity 0.88)
- Improving Diffusion Models for Inverse Problems Using Optimal Posterior Covariance (ICML 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.
