Research map / Image generation / Fast diffusion sampling
On Calibrating Diffusion Probabilistic Models
Tianyu Pang, Cheng Lu, Chao Du, Min Lin, Shuicheng YAN, Zhijie Deng
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 |
|---|---|---|---|
| subtract expected predicted noise (calibration) | 3-order DPM-Solver, NFE 15 · CIFAR-10 | FID: 4.61 -> 4.22 | Improved |
| subtract expected predicted noise (calibration) | 3-order DPM-Solver, NFE 20 · CIFAR-10 | FID: 3.89 -> 3.32 | Improved |
| calibration with 3-order solver at very low NFE | 3-order DPM-Solver, NFE 10 · CIFAR-10 | FID: 23.96 -> 53.50 | Hurt |
| calibration with 3-order solver at NFE 10 | 3-order DPM-Solver, NFE 10 · CelebA 64x64 | FID: 11.92 -> 35.47 | Hurt |
| calibration with 1-order solver | 1-order DPM-Solver, NFE 20 · CIFAR-10 | FID: 9.72 -> 8.86 | Improved |
| calibration with 2-order solver | 2-order DPM-Solver, NFE 20 · CIFAR-10 | FID: 4.14 -> 4.03 | Improved |
| calibration on CelebA | 3-order DPM-Solver, NFE 15 · CelebA 64x64 | FID: 3.91 -> 3.62 | Improved |
| calibration on CelebA | 3-order DPM-Solver, NFE 20 · CelebA 64x64 | FID: 2.84 -> 2.33 | Improved |
| calibration improves convergence of sampling | 3-order DPM-Solver · CIFAR-10 | FID at 1,000 NFE: 3.45 -> 3.32 | Improved |
| calibration on other metrics | 3-order DPM-Solver, NFE 20 · CIFAR-10 | sFID: 4.18 -> 4.14 | Improved |
| calibration on inception score | 3-order DPM-Solver, NFE 20 · CIFAR-10 | IS: 9.29 -> 9.38 | Improved |
| estimate calibration term from fewer training images (ablation) | calibrated DPM · CIFAR-10 | FID | Hurt |
Closest papers by meaning
- O(d/T) Convergence Theory for Diffusion Probabilistic Models under Minimal Assumptions (ICLR 2025 · similarity 0.89)
- Dimension-free Score Matching and Time Bootstrapping for Diffusion Models (NeurIPS 2025 · similarity 0.88)
- Beyond Scores: Proximal Diffusion Models (NeurIPS 2025 · similarity 0.88)
- On Inference Stability for Diffusion Models (AAAI 2024 · similarity 0.87)
- Distributional Diffusion Models with Scoring Rules (ICML 2025 · similarity 0.87)
- Convergence of Score-Based Discrete Diffusion Models: A Discrete-Time Analysis (ICLR 2025 · similarity 0.87)
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.
