Research map / Image generation / Fast diffusion sampling
Solving Inverse Problems with Flow-based Models via Model Predictive Control
George Webber, Alexander Denker, Riccardo Barbano, Andrew J. Reader
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 |
|---|---|---|---|
| Use MPC-Flow variants instead of global optimal control (FlowGrad, OC-Flow) | flow-based inverse solver · CelebA restoration tasks | PSNR/SSIM and runtime: MPC-Δt denoising PSNR 31.55 vs FlowGrad 26.07, OC-Flow 19.39 | Improved |
| Increase number of discretisation steps K in MPC-RHC (ablation) | MPC-RHC · OrganCMNIST computed tomography | PSNR / SSIM | Improved |
| Use single-step RHC (K=1) instead of K=3 (ablation) | MPC-RHC · CelebA super-resolution (x2) | PSNR: 31.39 -> 18.02 | Hurt |
| Use single-step RHC (K=1) instead of K=3 (speed) (ablation) | MPC-RHC · CelebA denoising | time per image (s): 175.20 -> 2.05 | Improved |
| Use MPC-Δt instead of K=3 RHC (non-linear deblurring) (ablation) | MPC-Flow · CelebA nonlinear deblurring | PSNR: 21.71 -> 25.44 | Improved |
| Add control regularisation (MPC-RHC K=1 vs FlowChef) for colourisation | FLUX.2 guidance · Image colourisation (FLUX.2) | PSNR: 21.3 -> 23.8 | Improved |
| Add control regularisation (MPC-RHC K=1 vs FlowChef) for style transfer | FLUX.2 guidance · Style transfer (45 images) | style-content trade-off | Improved |
| Use MPC-RHC (K=1) guidance vs unconditional generation (cost) | FLUX.2 colourisation · Image colourisation (FLUX.2) | time (s): 53 -> 101 | Hurt |
| Use PnP-Flow versus MPC-Δt on linear tasks | CelebA restoration · CelebA linear restoration | PSNR/SSIM: denoising PnP-Flow 32.45 vs MPC-Δt 31.55; super-resolution MPC-Δt 31.89 vs 31.49 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 3 on most shared results.
- PnP-Flow — PnP-Flow: Plug-and-Play Image Restoration with Flow Matching (ICLR 2025) · better on 3 of 6 shared results
- FlowChef — FlowChef: Steering of Rectified Flow Models for Controlled Generations (ICCV 2025) · better on 2 of 3 shared results
- D-Flow — DFlow: A Generative Model Combining Denoising AutoEncoder and Normalizing Flow for High Fidelity Waveform Generation (ICML 2024) · better on 0 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
- Flow Priors for Linear Inverse Problems via Iterative Corrupted Trajectory Matching (NeurIPS 2024 · similarity 0.90)
- Solving Inverse Problems with FLAIR (NeurIPS 2025 · similarity 0.90)
- Latent Refinement via Flow Matching for Training-free Linear Inverse Problem Solving (NeurIPS 2025 · similarity 0.90)
- Semantic Image Inversion and Editing using Rectified Stochastic Differential Equations (ICLR 2025 · similarity 0.88)
- Boosting Flow-based Generative Super-Resolution Models via Learned Prior (CVPR 2024 · similarity 0.88)
- PnP-Flow: Plug-and-Play Image Restoration with Flow Matching (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.
