Research map / Image generation / Fast diffusion sampling
CHORDS: Diffusion Sampling Accelerator with Multi-core Hierarchical ODE Solvers
Jiaqi Han, Haotian Ye, Puheng Li, Minkai Xu, James Zou, Stefano Ermon
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 theory-derived initialization sequence instead of uniform (ablation) | CHORDS · HunyuanVideo, 8 cores | speedup: 2.6 -> 2.9 | Improved |
| use theory-derived initialization sequence instead of uniform (ablation) | CHORDS · HunyuanVideo, 6 cores | speedup: 2.1 -> 2.5 | Improved |
| use theory-derived initialization sequence instead of uniform (ablation) | CHORDS · HunyuanVideo, 4 cores | speedup: 1.8 -> 2.1 | Improved |
| use theory-derived initialization sequence instead of uniform (ablation) | CHORDS · Flux, 8 cores | speedup: 2.2 -> 2.4 | Improved |
| use theory-derived initialization sequence instead of uniform (ablation) | CHORDS · HunyuanVideo, 8 cores | VBench quality: 84.0% -> 84.1% | No clear effect |
| increase diffusion steps from 50 to 100 | CHORDS · HunyuanVideo, 8 cores | speedup: 2.9 -> 3.6 | Improved |
| increase diffusion steps from 50 to 100 | CHORDS · HunyuanVideo, 8 cores | VBench quality: 84.1% -> 84.6% | Improved |
| increase diffusion steps from 50 to 100 | CHORDS · HunyuanVideo, 8 cores | latent RMSE: 0.068 -> 0.076 | Hurt |
| increase number of cores from 4 to 8 | CHORDS · HunyuanVideo | speedup: 2.1 -> 2.9 | Improved |
| use CHORDS instead of SRDS with four cores | SRDS · video diffusion models | speedup: 1.3x on average -> ~2x | Improved |
| use CHORDS instead of ParaDIGMS | ParaDIGMS · HunyuanVideo, 8 cores | latent RMSE: 0.202 -> 0.068 | Improved |
| use CHORDS parallel sampling instead of sequential | Sequential solver · HunyuanVideo, 8 cores | VBench quality: 84.4% -> 84.1% | No clear effect |
Later papers that compared with this method
1 later accepted paper reports CHORDS as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Robust Parallel Diffusion Sampling via Dynamic Jacobian Bandwidth (ICML 2026 · 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
- Accelerating Diffusion Models with Parallel Sampling: Inference at Sub-Linear Time Complexity (NeurIPS 2024 · similarity 0.92)
- A Simple Early Exiting Framework for Accelerated Sampling in Diffusion Models (ICML 2024 · similarity 0.92)
- Accelerating Parallel Sampling of Diffusion Models (ICML 2024 · similarity 0.92)
- Simple and Fast Distillation of Diffusion Models (NeurIPS 2024 · similarity 0.91)
- TADA: Improved Diffusion Sampling with Training-free Augmented DynAmics (NeurIPS 2025 · similarity 0.91)
- Tortoise and Hare Guidance: Accelerating Diffusion Model Inference with Multirate Integration (NeurIPS 2025 · similarity 0.91)
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.
