Abstract
Diffusion-based world models enable high-quality interactive environment generation but suffer from substantial inference overhead due to repeated Transformer evaluations during denoising. Existing caching methods mainly exploit temporal redundancy at the feature or token level, leaving the underlying mathematical structure of diffusion features largely unexplored. In this work, we reveal that world-model features exhibit highly stable singular subspaces across nearby denoising steps, while their singular values follow predictable evolution patterns. Building on this observation, we propose SpectralCache, a training-free spectral caching framework that reuses stable singular subspaces and estimates only low-dimensional singular values through linear extrapolation. We further exploit the spectral consistency between neighboring full-computation features to skip selected expensive backbone evaluations via singular value scaling. Extensive experiments on representative world models demonstrate that SpectralCache consistently improves inference efficiency while preserving generation quality. On HunyuanWorld-Voyager-13B, SpectralCache achieves 5.22x acceleration while maintaining a WorldScore of 65.90 for static scenes, substantially outperforming existing training-free caching methods in inference efficiency.
Keywords
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Mi, Z., Zhao, P., Hu, Z., Yu, X., Wang, Y., Zhang, G. L., & Huang, S. (2026). SpectralCache: Accelerating Diffusion-Based World Models via Spectral Feature Caching. https://omanscience.com/en/articles/spectralcache-accelerating-diffusion-based-world-models-via-spectral-feature-caching
MLA 9
Mi, Zhendong, et al. "SpectralCache: Accelerating Diffusion-Based World Models via Spectral Feature Caching." https://omanscience.com/en/articles/spectralcache-accelerating-diffusion-based-world-models-via-spectral-feature-caching.
Chicago (author–date)
Mi, Zhendong, Pu Zhao, Ziyu Hu, Xiaodong Yu, Yanzhi Wang, Grace Li Zhang, and Shaoyi Huang. 2026. "SpectralCache: Accelerating Diffusion-Based World Models via Spectral Feature Caching." https://omanscience.com/en/articles/spectralcache-accelerating-diffusion-based-world-models-via-spectral-feature-caching.
Harvard
Mi, Z., Zhao, P., Hu, Z., Yu, X., Wang, Y., Zhang, G. L. and Huang, S. (2026) 'SpectralCache: Accelerating Diffusion-Based World Models via Spectral Feature Caching', Available at: https://omanscience.com/en/articles/spectralcache-accelerating-diffusion-based-world-models-via-spectral-feature-caching.
Vancouver
Mi Z, Zhao P, Hu Z, Yu X, Wang Y, Zhang GL, et al. SpectralCache: Accelerating Diffusion-Based World Models via Spectral Feature Caching. https://omanscience.com/en/articles/spectralcache-accelerating-diffusion-based-world-models-via-spectral-feature-caching
IEEE
Z. Mi, P. Zhao, Z. Hu, X. Yu, Y. Wang, G. L. Zhang, and S. Huang, "SpectralCache: Accelerating Diffusion-Based World Models via Spectral Feature Caching," https://omanscience.com/en/articles/spectralcache-accelerating-diffusion-based-world-models-via-spectral-feature-caching.