Research map / Image generation / Fast diffusion sampling
Stochastic Optimal Control Sampling for Diffusion Inverse Problems
Jie Zhang, Youmei Qiu, Hanling Tian, Jingyuan Zhang, Xiang Yin, Xiaolin Huang
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 SOCS-Nonlinear-gamma instead of SOCS-Linear for 4x super-resolution on FFHQ (FID) (ablation) | SOCS · FFHQ 4x SR | FID: 72.99 -> 76.86 | Hurt |
| use nonlinear refinement instead of linear projection on nonlinear deblurring (ablation) | SOCS · FFHQ nonlinear deblur | PSNR: 21.73 -> 28.29 | Improved |
| use nonlinear refinement instead of linear projection on motion deblurring (ablation) | SOCS · FFHQ motion deblur | PSNR: 26.53 -> 29.74 | Improved |
| use finite terminal penalty gamma instead of gamma to infinity on box inpainting (ablation) | SOCS-Nonlinear · FFHQ box inpainting | FID: 55.26 -> 43.79 | Improved |
| use finite terminal penalty gamma instead of gamma to infinity on random inpainting (ablation) | SOCS-Nonlinear · FFHQ random inpainting | FID: 55.42 -> 52.73 | Improved |
| use finite terminal penalty gamma on high dynamic range reconstruction (ablation) | SOCS-Nonlinear · FFHQ high dynamic range | PSNR: 26.30 -> 24.31 | Hurt |
| use finite terminal penalty gamma on phase retrieval (ablation) | SOCS-Nonlinear · FFHQ phase retrieval | FID: 52.43 -> 44.86 | Improved |
| use SOCS-Nonlinear-gamma instead of strongest baseline DAPS on 4x SR | SOCS · ImageNet 4x SR | FID: 112.60 -> 93.77 | Improved |
| apply SOCS on SD3-medium instead of FlowDPS for x12 super-resolution | SD3-medium · FFHQ 768x768 SR x12 (Avgpool) | FID: 65.00 -> 53.61 | Improved |
| increase gamma towards infinity (ablation) | SOCS · multiple tasks | perceptual quality | Hurt |
| use very small gamma (~10^3) (ablation) | SOCS · super-resolution, phase retrieval | quality / consistency | Hurt |
| vary gamma within 10^5 to 10^9 (ablation) | SOCS · multiple tasks | metrics | No clear effect |
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.
- FlowDPS — FlowDPS : Flow-Driven Posterior Sampling for Inverse Problems (ICCV 2025) · better on 2 of 2 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
- A Variational Perspective on Solving Inverse Problems with Diffusion Models (ICLR 2024 · similarity 0.91)
- Arbitrary-steps Image Super-resolution via Diffusion Inversion (CVPR 2025 · similarity 0.90)
- CCS: Controllable and Constrained Sampling with Diffusion Models via Initial Noise Perturbation (NeurIPS 2025 · similarity 0.89)
- Solving Diffusion ODEs with Optimal Boundary Conditions for Better Image Super-Resolution (ICLR 2024 · similarity 0.89)
- Improving Diffusion Inverse Problem Solving with Decoupled Noise Annealing (CVPR 2025 · similarity 0.89)
- Analyzing and Guiding Zero-Shot Posterior Sampling in Diffusion Models (NeurIPS 2026 · 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.
