Research map / Image generation / Fast diffusion sampling
CoSIGN: Few-Step Guidance of ConSIstency Model to Solve General INverse Problems
Jiankun Zhao, Bowen Song, Liyue Shen
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 pretrained consistency model prior instead of end-to-end model without CM (ablation) | CoSIGN · LSUN bedroom block inpainting | FID: 42.68 -> 39.89 | Improved |
| raise ControlNet guidance scale from 0 to 1.0 (ablation) | CoSIGN · LSUN bedroom block inpainting | LPIPS: 0.729 -> 0.146 | Improved |
| add hard measurement constraint (1 NFE) (ablation) | CoSIGN · LDCT sparse-view CT | PSNR: 31.38 -> 33.41 | Improved |
| add hard measurement constraint (2 NFE) (ablation) | CoSIGN · LDCT sparse-view CT | PSNR: 31.80 -> 34.26 | Improved |
| use second sampling step (2 NFE instead of 1) (ablation) | CoSIGN · LDCT sparse-view CT | PSNR: 33.41 -> 34.26 | Improved |
| increase NFE beyond 2 (ablation) | CoSIGN · LDCT sparse-view CT | PSNR / SSIM | No clear effect |
| use CoSIGN instead of FBP-UNet on CT | CoSIGN · LDCT sparse-view CT (23 angles) | PSNR / SSIM: PSNR 31.47 -> 33.41 (1 NFE); SSIM 0.874 vs 0.836 | Mixed |
| use CoSIGN vs baselines on natural image LPIPS | CoSIGN · LSUN bedroom (inpainting, SR) | LPIPS: inpainting 0.137 vs CDDB 0.126; SR 0.217 vs SwinIR 0.200 | Mixed |
| apply trained model to out-of-domain 10-angle CT | CoSIGN · LDCT CT out-of-domain (10 angles) | PSNR: FBP-UNet 22.49 vs CoSIGN 28.33 | Improved |
Closest papers by meaning
- SITCOM: Step-wise Triple-Consistent Diffusion Sampling For Inverse Problems (ICML 2025 · similarity 0.91)
- Solving Inverse Problems with Latent Diffusion Models via Hard Data Consistency (ICLR 2024 · similarity 0.90)
- Improving Diffusion Models for Inverse Problems Using Optimal Posterior Covariance (ICML 2024 · similarity 0.89)
- Diffusion State-Guided Projected Gradient for Inverse Problems (ICLR 2025 · similarity 0.89)
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.89)
- Linearly Constrained Diffusion Implicit Models (NeurIPS 2025 · similarity 0.89)
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.
