Frequency-Step Cache: Separating Synchronization and Repair for Efficient Autoregressive Video Diffusion
cs.CV, cs.LG
Submitted: 2026-10-11
Updated: 2026-10-11
Comments: 30 pages, 11 figures, 18 tables
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Sources
- SkyReels-V2: Infinite-length Film Generative Model
- $\Delta$-DiT: A Training-Free Acceleration Method Tailored for Diffusion Transformers
- Not All Frames Deserve Full Computation: Accelerating Autoregressive Video Generation via Selective Computation and Predictive Extrapolation
- Frequency-Aware Error-Bounded Caching for Accelerating Diffusion Transformers
- FreqCa: Accelerating Diffusion Models via Frequency-Aware Caching
- MAGI-1: Autoregressive Video Generation at Scale
- FORA: Fast-Forward Caching in Diffusion Transformer Acceleration
- Matrix-Game 3.0: Real-Time and Streaming Interactive World Model with Long-Horizon Memory
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models