Research map / Image generation / Fast diffusion sampling
DC-Solver: Improving Predictor-Corrector Diffusion Sampler via Dynamic Compensation
Wenliang Zhao, Haolin Wang, Jie Zhou, Jiwen Lu
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 dynamic compensation ratios to UniPC (DC-Solver) (ablation) | UniPC · FFHQ unconditional NFE=5 | FID: 18.66 -> 10.38 | Improved |
| use constant compensation ratio 1.1 instead of dynamic ratios (ablation) | DC-Solver · FFHQ NFE=5 | FID: 10.38 -> 13.99 | Hurt |
| use constant compensation ratio 0.9 (ablation) | UniPC · FFHQ NFE=5 | FID: 18.66 -> 26.43 | Hurt |
| increase datapoints for ratio search from 10 to 30 (ablation) | DC-Solver · FFHQ NFE=5 | FID: 10.38 -> 10.93 | No clear effect |
| reduce datapoints from 10 to 5 (ablation) | DC-Solver · FFHQ NFE=5 | FID: 10.38 -> 12.39 | Hurt |
| use compensation order K=2 instead of K=1 or K=3 (ablation) | DC-Solver · FFHQ NFE=5 | FID: 12.70 (K=1), 11.63 (K=3) -> 10.38 | Improved |
| increase optimization iterations beyond 40 (ablation) | DC-Solver · FFHQ NFE=5 | FID: 10.38 -> 10.63 | No clear effect |
| add dynamic compensation to predictor-only DDIM | DDIM · FFHQ NFE=5 | FID: 57.92 -> 16.56 | Improved |
| add dynamic compensation to predictor-only DPM-Solver++ | DPM-Solver++ · FFHQ NFE=5 | FID: 27.80 -> 11.97 | Improved |
| predict compensation ratios by cascade polynomial regression for unseen CFG/NFE | DC-Solver · SD2.1 unseen CFG=6.0 NFE=14 | MSE: 0.304 -> 0.196 | Improved |
| apply DC-Solver instead of UniPC at inference | DC-Solver · SD2.1 NFE=5 | inference time (s): 1.533 -> 1.532 | 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.
- UniPC — UniPC: A Unified Predictor-Corrector Framework for Fast Sampling of Diffusion Models (NeurIPS 2023) · better on 9 of 9 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
- Accelerating Diffusion Sampling with Optimized Time Steps (CVPR 2024 · similarity 0.90)
- A Unified Sampling Framework for Solver Searching of Diffusion Probabilistic Models (ICLR 2024 · similarity 0.90)
- PFDiff: Training-Free Acceleration of Diffusion Models Combining Past and Future Scores (ICLR 2025 · similarity 0.89)
- Diffusion Sampling Correction via Approximately 10 Parameters (ICML 2025 · similarity 0.89)
- Informed Correctors for Discrete Diffusion Models (NeurIPS 2025 · similarity 0.89)
- Alleviating Exposure Bias in Diffusion Models through Sampling with Shifted Time Steps (ICLR 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.
