Research map / Image generation / Fast diffusion sampling
GENIE: Higher-Order Denoising Diffusion Solvers
Tim Dockhorn, Arash Vahdat, Karsten Kreis
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 |
|---|---|---|---|
| replace DDIM (Euler) with second-order GENIE solver | DDIM ODE sampler · CIFAR-10 unconditional NFEs=10 | FID: 11.2 -> 5.28 | Improved |
| replace DDIM with GENIE | DDIM ODE sampler · CIFAR-10 unconditional NFEs=20 | FID: 5.87 -> 3.94 | Improved |
| replace DDIM with GENIE | DDIM ODE sampler · Cats upsampler NFEs=5 | FID: 9.47 -> 5.53 | Improved |
| replace multistep S-PNDM with GENIE | S-PNDM · CIFAR-10 unconditional NFEs=10 | FID: 10.3 -> 5.28 | Improved |
| add mixed network parameterization for the distilled head (ablation) | GENIE head · CIFAR-10 unconditional NFEs=10 | FID: 6.32 -> 6.04 | Improved |
| add distillation loss weighting g_d(t)=gamma_t^2 (ablation) | GENIE head · CIFAR-10 unconditional NFEs=10 | FID: 7.45 -> 6.04 | Improved |
| double head depth to two residual blocks (ablation) | GENIE head · CIFAR-10 unconditional NFEs=20 | FID: 3.94 -> 4.05 | No clear effect |
| replace distilled head with exact autodiff derivatives | GENIE · CIFAR-10 unconditional | FID | No clear effect |
| increase NFE | GENIE vs DDIM gap · CIFAR-10 unconditional | FID: 11.2 -> 5.28 at 10 NFEs; 5.16 -> 3.64 at 25 | Mixed |
| DDIM ODE vs LSUN datasets at NFEs=5 | GENIE · LSUN Bedrooms/Church | FID | Hurt |
Later papers that compared with this method
7 later accepted papers report Genie2 (16M) as a baseline in their results tables; 6 of them report a better number on most of the results they share with it.
- Protein Autoregressive Modeling via Multiscale Structure Generation (ICML 2026 · better on 1 of 1 shared result)
- Proteina: Scaling Flow-based Protein Structure Generative Models (ICLR 2025 · better on 3 of 3 shared results)
- Observation-Guided Diffusion Probabilistic Models (CVPR 2024 · better on 2 of 2 shared results)
- Bespoke Solvers for Generative Flow Models (ICLR 2024 · better on 2 of 2 shared results)
- Efficient Denoising Diffusion via Probabilistic Masking (ICML 2024 · better on 2 of 2 shared results)
- PLANNER: Generating Diversified Paragraph via Latent Language Diffusion Model (NeurIPS 2023 · better on 2 of 4 shared results)
- AR-Diffusion: Auto-Regressive Diffusion Model for Text Generation (NeurIPS 2023 · better on 2 of 2 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
- Think while You Generate: Discrete Diffusion with Planned Denoising (ICLR 2025 · similarity 0.90)
- EVODiff: Entropy-aware Variance Optimized Diffusion Inference (NeurIPS 2025 · similarity 0.90)
- Denoising Diffusion Bridge Models (ICLR 2024 · similarity 0.90)
- Distilling ODE Solvers of Diffusion Models into Smaller Steps (CVPR 2024 · similarity 0.89)
- Diffusion Bridge Implicit Models (ICLR 2025 · similarity 0.89)
- Interpreting and Improving Diffusion Models from an Optimization Perspective (ICML 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.
