Research map / Image generation / Fast diffusion sampling
Score-based Generative Models with Lévy Processes
Eunbi Yoon, Keehun Park, Sungwoong Kim, Sungbin Lim
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 Levy noise (LIM) instead of Brownian noise | diffusion model · Imbalanced CIFAR10 | FID: 62.62 -> 21.07 | Improved |
| use Levy noise (LIM) instead of Brownian noise | diffusion model · Imbalanced CIFAR10 | Recall: 0.5002 -> 0.5549 | Improved |
| use Levy noise (LIM) instead of Brownian noise | diffusion model · Imbalanced CIFAR10 | MMD: 0.01396 -> 0.00416 | Improved |
| use Levy noise (LIM) instead of Brownian noise | diffusion model · synthetic 10:1 Gaussian mixture | FID: 8.312 -> 0.663 | Improved |
| use Levy noise (LIM) instead of Brownian noise | diffusion model · synthetic 10:1 Gaussian mixture | MMD: 0.02 -> 0.026 | Hurt |
| use Levy noise (LIM) instead of Brownian noise | diffusion model · CIFAR10LT conditional | Recall: 0.5796 -> 0.5686 | Mixed |
| use Levy noise (LIM) instead of Brownian noise | diffusion model · CIFAR10LT conditional | FID: 22.00 -> 15.594 | Mixed |
| use Levy noise (LIM) instead of Brownian noise | diffusion model · CIFAR10 conditional | FID: 16.48 -> 18.24 | Mixed |
| use Levy noise (LIM) instead of Brownian noise | diffusion model · CelebA-HQ imputation | LPIPS: 0.0132 -> 0.0197 | Improved |
| use Levy noise with 500 NFE vs diffusion with 1000 NFE | DDPM architecture · CelebA-HQ 256x256 | FID: 11.87 -> 7.76 | Improved |
| use Levy noise (LIM) instead of Brownian noise | ADM architecture · ImageNet 64x64 | FID: 14.23 -> 12.97 | Improved |
| change alpha from 1.8 to 1.5 (ablation) | LIM-DDPM · CelebA 64x64 | FID: 3.21 -> 1.57 | Improved |
Closest papers by meaning
- Heavy-Tailed Diffusion with Denoising Levy Probabilistic Models (ICLR 2025 · similarity 0.91)
- Generative Fractional Diffusion Models (NeurIPS 2024 · similarity 0.91)
- Training Diffusion-based Generative Models with Limited Data (ICML 2025 · similarity 0.89)
- Towards Non-Asymptotic Convergence for Diffusion-Based Generative Models (ICLR 2024 · similarity 0.88)
- O(d/T) Convergence Theory for Diffusion Probabilistic Models under Minimal Assumptions (ICLR 2025 · similarity 0.88)
- Evaluating the design space of diffusion-based generative models (NeurIPS 2024 · similarity 0.88)
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.
