Research map / Image generation / Fast diffusion sampling
Denoising Diffusion Probabilistic 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 |
|---|---|---|---|
| train with simplified unweighted epsilon objective instead of true variational bound (ablation) | DDPM · CIFAR10 | FID | Helped |
| train with simplified objective instead of true variational bound (ablation) | DDPM · CIFAR10 | NLL (bits/dim) | Hurt |
| predict epsilon instead of posterior mean, with simplified objective (ablation) | DDPM · CIFAR10 | FID | Helped |
| predict epsilon instead of posterior mean, with true variational bound (ablation) | DDPM · CIFAR10 | FID | No clear effect |
| learn reverse process variances instead of fixing them (ablation) | DDPM · CIFAR10 | FID | Hurt |
| sigma_t^2 = beta_t vs tilde beta_t (ablation) | DDPM · CIFAR10 | sample quality | No clear effect |
| predict x0 instead of epsilon (ablation) | DDPM · CIFAR10 | sample quality | Hurt |
| add dropout 0.1 (ablation) | DDPM · CIFAR10 | sample quality | Helped |
| add random horizontal flips (ablation) | DDPM · CIFAR10 | sample quality | Helped |
| scale up model to ~256M params (ablation) | DDPM · LSUN Bedroom | FID: 6.36 -> 4.90 | Helped |
Closest papers by meaning
- Assessing the quality of denoising diffusion models in Wasserstein distance: noisy score and optimal bounds (NeurIPS 2025 · similarity 0.91)
- Efficient Denoising Diffusion via Probabilistic Masking (ICML 2024 · similarity 0.90)
- Denoising Diffusion Bridge Models (ICLR 2024 · similarity 0.90)
- UDPM: Upsampling Diffusion Probabilistic Models (NeurIPS 2024 · similarity 0.90)
- O(d/T) Convergence Theory for Diffusion Probabilistic Models under Minimal Assumptions (ICLR 2025 · similarity 0.90)
- EVODiff: Entropy-aware Variance Optimized Diffusion Inference (NeurIPS 2025 · 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.