Research map / Image generation / Fast diffusion sampling
Dual Ascent Diffusion for Inverse Problems
Minseo Kim, Axel Levy, Gordon Wetzstein
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 dual variable together with noising step (DDiff vs DDiff-HQS) (ablation) | DDiff · FFHQ 256 Super-Resolution 4x | LPIPS: 0.369 -> 0.199 | Improved |
| add dual variable without noising step (Diff-PnP-ADMM vs Diff-PnP-HQS) (ablation) | Diff-PnP-HQS · FFHQ 256 Gaussian Deblurring | PSNR: 5.82 -> 6.18 PSNR, LPIPS 0.858 -> 0.820 | Hurt |
| add dual variable alone without noising on super-resolution (ablation) | Diff-PnP-HQS · FFHQ 256 Super-Resolution 4x | LPIPS: 0.535 -> 0.572 | Hurt |
| add noising step (with dual variable) (ablation) | Diff-PnP-ADMM · FFHQ 256 Super-Resolution 4x | LPIPS: 0.572 -> 0.199 | Improved |
| replace DAPS-style MCMC with dual ascent at same NFEs | DDiff · FFHQ, LPIPS vs time | LPIPS and sampling time | Improved |
| use DDiff instead of DAPS at high measurement noise | DDiff · FFHQ phase retrieval, sigma=0.3 | LPIPS: LPIPS divided by 3 at sigma=0.3 | Improved |
| use DDiff instead of DAPS at very low noise | DDiff · FFHQ, sigma<0.05 | PSNR | Mixed |
| DDiff vs DAPS across tasks (Gaussian/motion deblurring, ImageNet phase retrieval) | DDiff · FFHQ and ImageNet | PSNR: Gaussian deblur FFHQ 28.87 vs 29.63 | Mixed |
| apply data consistency in pixel space instead of latent space (LatentDDiff) (ablation) | LatentDDiff · latent diffusion | reconstruction quality and cost | 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.
- DMPlug — DMPlug: A Plug-in Method for Solving Inverse Problems with Diffusion Models (NeurIPS 2024) · 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
- Integrating Intermediate Layer Optimization and Projected Gradient Descent for Solving Inverse Problems with Diffusion Models (ICML 2025 · similarity 0.90)
- Unleashing the Denoising Capability of Diffusion Prior for Solving Inverse Problems (NeurIPS 2024 · similarity 0.90)
- Improving Diffusion Inverse Problem Solving with Decoupled Noise Annealing (CVPR 2025 · similarity 0.89)
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.89)
- Improving Diffusion-based Inverse Algorithms under Few-Step Constraint via Linear Extrapolation (NeurIPS 2025 · similarity 0.89)
- Principled Probabilistic Imaging using Diffusion Models as Plug-and-Play Priors (NeurIPS 2024 · 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.
