arXiv：cs.LG· Qinhan Lyu, Jusheng Zhang, Zhixing Zhang, Hao Liu, Zijie Liang, Jin Jiang, Keze Wang·· 11 小时前
CacheScope：面向无训练扩散加速的精确计算分配方法
CacheScope: Exact-Compute Allocation for Training-Free Diffusion Acceleration
AI 导读
CacheScope 通过在线识别高风险特征复用步骤并选择性回退到完整去噪计算，实现无训练扩散 Transformer 推理加速，且不改变底层缓存重建器，无需额外训练、梯度或辅助网络。
正文
Abstract:Training-free feature reuse accelerates diffusion transformer inference by replacing full denoising computation at selected steps with cached approximations. Existing methods mainly improve the reconstruction of skipped features, but we find that once a stable first-order reconstructor is used, increasing reconstruction complexity brings limited gains, whereas the placement of exact-computation steps has a much larger effect on fidelity. More importantly, absolute shallow-probe variation is jointly affected by input-trajectory displacement and network response strength, and therefore does not directly reflect cache risk. We address this confounding by normalizing probe-output variation with the corresponding input displacement, yielding a directional gain per unit trajectory movement. Based on this signal, we propose CacheScope, which identifies high-risk reuse steps online and selectively falls back to full denoising computation while leaving the underlying cache reconstructor unchanged. CacheScope requires no additional training, gradients, auxiliary networks, or prompt-specific offline schedule optimization. Across FLUX.1-dev, Wan2.1-T2V-1.3B, and Open-Sora 1.2, CacheScope achieves up to 3.57x acceleration and reduces median LPIPS by up to 70.1% over the strongest probe-based baseline.
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.12530 [cs.LG] |
| (or arXiv:2610.12530v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.12530 arXiv-issued DOI via DataCite |
Submission history
From: Qinhan Lyu [view email]
[v1]
Thu, 8 Oct 2026 10:19:34 UTC (2,655 KB)
来源：arXiv：cs.LG · arxiv.org