Research map / Image generation / Fast diffusion sampling
Accelerating Diffusion Models via Equal-Risk Caching
Can Zhang, Liangshun Zou
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 probe length l'=5 instead of l'=1 (ablation) | ERC · FLUX.1 dev B=15 | PSNR: 25.2465 -> 25.4227 | Improved |
| use probe length l'=5 instead of l'=11 (ablation) | ERC · FLUX.1 dev B=26 | PSNR: 29.5831 -> 31.5304 | Improved |
| use probe length l'=5 instead of l'=9 (ablation) | ERC · FLUX.1 dev B=15 | PSNR: 24.3353 -> 25.4227 | Improved |
| increase offline prompts from 1 to 10 (ablation) | ERC · FLUX.1 dev B=26 | LPIPS: 0.1080 -> 0.0816 | Improved |
| increase offline prompts from 10 to 50 (ablation) | ERC · FLUX.1 dev B=26 | PSNR: 31.5304 -> 31.6257 | No clear effect |
| increase offline prompts from 10 to 50 (ablation) | ERC · FLUX.1 dev B=15 | PSNR: 25.4227 -> 25.4993 | No clear effect |
| use equal-risk schedule instead of LeMiCa schedule | ERC · FLUX.1 dev B=26 | PSNR: 30.1025 -> 31.5304 | Improved |
| use equal-risk schedule instead of LeMiCa schedule | ERC · FLUX.1 dev B=15 | LPIPS: 0.1857 -> 0.1954 | Hurt |
| use equal-risk schedule instead of LeMiCa schedule | ERC · Qwen-Image B=17 | PSNR: 27.9708 -> 27.6859 | Hurt |
| add DCTA to ERC | ERC · Qwen-Image B=17 | LPIPS: 0.1268 -> 0.0842 | Improved |
| add DCTA to ERC | ERC · Wan2.1 B=17 | VBench (%): 74.08 -> 74.80 | Improved |
| use ERC instead of LeMiCa on OOD prompts | ERC · FLUX.1 dev GenAI B=26 | PSNR: 29.9245 -> 31.9375 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DiCache — DiCache: Let Diffusion Model Determine Its Own Cache (ICLR 2026) · better on 1 of 1 shared result
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
- Cache Me if You Can: Accelerating Diffusion Models through Block Caching (CVPR 2024 · similarity 0.92)
- CacheQuant: Comprehensively Accelerated Diffusion Models (CVPR 2025 · similarity 0.91)
- Timestep Embedding Tells: It's Time to Cache for Video Diffusion Model (CVPR 2025 · similarity 0.91)
- DeepCache: Accelerating Diffusion Models for Free (CVPR 2024 · similarity 0.90)
- FasterCache: Training-Free Video Diffusion Model Acceleration with High Quality (ICLR 2025 · similarity 0.89)
- Learning-to-Cache: Accelerating Diffusion Transformer via Layer Caching (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.
