Research map / Image generation / Fast diffusion sampling
gDDIM: Generalized denoising diffusion implicit models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 gDDIM instead of probability-flow RK45 | DDPM · CIFAR10, NFE=50 | FID: 6.62 -> 2.59 | Helped |
| Use gDDIM instead of probability-flow RK45 | BDM · CIFAR10, NFE=50 | FID: 7.12 -> 2.49 | Helped |
| Use gDDIM instead of probability-flow RK45 | CLD · CIFAR10, NFE=50 | FID: 31.7 -> 2.26 | Helped |
| Use the gDDIM parameterization K=R rather than K=L (ablation) | CLD · CIFAR10, NFE=50 | FID: 3.31 -> 2.26 | Helped |
| Use the gDDIM parameterization K=R rather than K=L (ablation) | CLD · CIFAR10, NFE=20 | FID: 368 -> 3.90 | Helped |
| Increase lambda from 0.0 to 0.1 (ablation) | gDDIM · CIFAR10, NFE=50 | FID: 5.17 -> 5.51 | Hurt |
| Increase lambda from 0.0 to 0.3 (ablation) | gDDIM · CIFAR10, NFE=50 | FID: 5.17 -> 12.13 | Hurt |
| Increase lambda from 0.0 to 0.5 (ablation) | gDDIM · CIFAR10, NFE=50 | FID: 5.17 -> 33 | Hurt |
| Increase lambda from 0.0 to 0.7 (ablation) | gDDIM · CIFAR10, NFE=50 | FID: 5.17 -> 41 | Hurt |
| Increase lambda from 0.0 to 1.0 (ablation) | gDDIM · CIFAR10, NFE=50 | FID: 5.17 -> 49 | Hurt |
| Use stochastic gDDIM rather than Euler-Maruyama at lambda 0.1 | CLD · CIFAR10, NFE=50, lambda=0.1 | FID: 168 -> 5.51 | Helped |
Closest papers by meaning
- Diffusion Bridge Implicit Models (ICLR 2025 · similarity 0.91)
- Interpreting and Improving Diffusion Models from an Optimization Perspective (ICML 2024 · similarity 0.90)
- Directly Denoising Diffusion Models (ICML 2024 · similarity 0.89)
- Denoising Diffusion Step-aware Models (ICLR 2024 · similarity 0.89)
- O(d/T) Convergence Theory for Diffusion Probabilistic Models under Minimal Assumptions (ICLR 2025 · similarity 0.89)
- Score-Optimal Diffusion Schedules (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.