CacheScope is a training‑free method that accelerates diffusion transformer inference by selectively replacing full denoising steps with cached approximations. It identifies high‑risk reuse steps online by normalizing probe‑output variation with input displacement, allowing the system to fall back to exact computation only when necessary. The approach achieves up to 3.57× speed‑up and reduces median LPIPS by up to 70.1% on several state‑of‑the‑art models without requiring additional training or offline scheduling.
arXiv:2607. 29398v1 Announce Type: new Abstract: Diffusion models have revolutionized generative tasks but incur high latency due to iterative denoising.
arXiv:2610.10366v1 Announce Type: new Abstract: Feature caching accelerates diffusion sampling by replacing expensive network evaluations with predictions from previously computed activations. Howeve...
arXiv:2606. 26778v1 Announce Type: cross Abstract: Diffusion Transformers (DiTs) have driven substantial progress in image and video generation but suffer from prohibitive computational costs.
arXiv:2607. 27842v1 Announce Type: cross Abstract: Diffusion models are widely used to generate high-quality images and videos, but their iterative denoising process remains computationally intensive.
Diffusion models have achieved remarkable success in image and video generation, yet the high computational cost of iterative sampling remains a critical bottleneck for practical deployment. Feature c...
arXiv:2602. 13357v3 Announce Type: replace-cross Abstract: Diffusion Transformers (DiTs) achieve state-of-the-art performance in high-fidelity image and video generation but suffer from expensive inference due to their iterative denoising structure.
arXiv:2603.01623v2 Announce Type: replace Abstract: Diffusion models have become the dominant tool for high-fidelity image and video generation, yet are critically bottlenecked by their inference spe...
arXiv:2610.10457v1 Announce Type: new Abstract: Diffusion Transformers (DiTs) achieve remarkable performance in video synthesis, but their iterative denoising process suffers from high inference late...
arXiv:2602.24208v2 Announce Type: replace-cross Abstract: Diffusion models achieve state-of-the-art video generation quality, but their inference remains expensive due to the large number of sequenti...
The paper introduces TRACK, a training‑free trajectory routing method that accelerates video diffusion by selectively switching between large and small models during denoising steps. A calibration process generates a disagreement score map, guiding the selection of the appropriate model at each step to maintain quality while reducing computational cost. Experiments on Wan 2.1, Cosmos 3, TurboDiffusion, and FastVideo show speedups ranging from 1.95× to 2.73× with comparable quality and diversity.
arXiv:2606. 31026v1 Announce Type: cross Abstract: We propose OTCache, a training-free framework for accelerating diffusion sampling via caching schedule prediction.
The paper introduces a cache‑aware post‑training framework for Mixture‑of‑Experts (MoE) models that jointly adapts the MoE backbone and lightweight auxiliary cache routers while keeping the native Top‑K expert‑selection rule. Two modes are proposed: Temporal Router, which predicts same‑layer reuse and retains experts for future tokens, and Spatio‑Temporal Router, which adds a Spatio Router that refines the temporal cache using the causal predecessor’s hidden state. Experiments on Qwen3 and GPT‑OSS across GSM8K, MATH, and CommonsenseQA show that Temporal Router improves cache hit rates and reduces expert‑weight traffic, while Spatio‑Temporal Router achieves the best load‑adjusted efficiency, outperforming strong prefetching baselines.