cs.CVNov 18, 2026

Branch Caching After Step Distillation in Four- and Eight-Step Image Generators

Authors: Mingyu Liang, Dingkun Xu

Organizations: School of Computer Science, Nanjing University, Nanjing, China.

Abstract

Can a distilled image generator save time by reusing features between its few remaining denoising steps? We study a fixed branch cache that replaces selected attention and MLP evaluations while preserving the student's noise schedule. One full trajectory and one raw-cache trajectory from the same initial noise supply a scalar correction for each branch type and reuse position. At 102421024^2, one internal reuse reduces four-step Qwen-Image-Lightning denoising time from 1.279 to 0.977 seconds. Three reuses yield 1.51×1.51\times speedup on eight-step Lightning and 1.53×1.53\times on eight-step Hyper-FLUX. We evaluate image quality on all 200 DrawBench prompts. With the reuse positions held fixed, correction improves mean same-seed image fidelity over raw reuse by 0.66 and 0.87 dB on the two Lightning students. On Hyper-FLUX the PSNR difference is 0.01 dB, while LPIPS improves from 0.1625 to 0.1546. These results measure the effect of branch caching and correction while retaining each student's original sampling grid.

Figures & tables

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. FeatFix: Reuse What You Verify through Local Exact-Feature Correction for Faster Cached Diffusion Inference

    Jul 30, 2026Hanshuai Cui, Zhiqing Tang, Zhi Yao +3Diffusion-Based Image GenerationDiffusion Model Caching

  2. ReCache: Learning Budget-Aware Caching Schedules for Diffusion Models via REINFORCE

    Jun 4, 2026Mishan Aliev, Eva Neudachina, Ilya Bykov +4Diffusion Model CachingDiffusion Model Inference Acceleration

  3. PoSTCache: From Forecasting to Postdictive Smoothing for Revisable Diffusion Caching

    Oct 19, 2026Liwei Lan, Hui Chen, Jungong Han +1Diffusion Model CachingDiffusion Model Inference Acceleration