Research map / Image generation / Fast diffusion sampling
ResCa: Residual Caching for Diffusion Transformers Acceleration
Haipeng Fang, Yu Li, Fan Tang, Yixing Lu, Juan Cao, Sheng Tang
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 trajectory-based clustering instead of feature-based clustering (ablation) | ResCa · FLUX (N=8) | image quality | Improved |
| use K=16 or K=32 clusters (ablation) | ResCa · FLUX (N=8, O=1) | image quality vs cost | Improved |
| raise residual order from 1 to 2 (ablation) | ResCa-IT · DiT-XL/2 ImageNet | FID: 2.62 -> 2.57 | Improved |
| raise residual order from 2 to 3 or 4 (ablation) | ResCa-IT · DiT-XL/2 ImageNet | FID: 2.57 -> 2.58 | No clear effect |
| use implicit Euler instead of ClusCa mixed first-order prediction | ResCa-IE · DiT-XL/2 ImageNet (N=5) | FID: 2.65 -> 2.62 | Improved |
| use implicit Euler instead of ClusCa | ResCa-IE · FLUX DrawBench (N=5) | Image Reward: 0.9825 -> 0.9958 | Improved |
| use implicit Euler instead of TaylorSeer forecasting | ResCa-IE · DiT-XL/2 ImageNet | FID: 2.49 at 3.23x vs 2.49 at 2.77x | Improved |
| use BDF2 implicit update | ResCa-IB · FLUX DrawBench (N=7) | Image Reward: 0.9737 -> 0.9889 | Improved |
Closest papers by meaning
- HarmoniCa: Harmonizing Training and Inference for Better Feature Caching in Diffusion Transformer Acceleration (ICML 2025 · similarity 0.91)
- Accelerating Diffusion Transformer via Increment-Calibrated Caching with Channel-Aware Singular Value Decomposition (CVPR 2025 · similarity 0.91)
- Learning-to-Cache: Accelerating Diffusion Transformer via Layer Caching (NeurIPS 2024 · similarity 0.90)
- Accelerating Diffusion Transformers with Token-wise Feature Caching (ICLR 2025 · similarity 0.90)
- Cache Me if You Can: Accelerating Diffusion Models through Block Caching (CVPR 2024 · similarity 0.90)
- DeepCache: Accelerating Diffusion Models for Free (CVPR 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.
