NaviCache: 视频生成的测试时自校准缓存
NaviCache: Test-Time Self-Calibration Caching for Video Generation
AI总结：
针对视频扩散模型计算成本高的问题，提出NaviCache方法，将特征演化重构思为惯性导航系统问题，通过双状态估计架构自适应跟踪特征变化比和潜在漂移，实现有界误差的计算跳过，在多个模型上取得优异性能。
摘要：
视频扩散模型（VDMs）受到巨大计算成本的限制。虽然基于离线校准的加速方法存在校准数据依赖、校准时间过长以及对分布偏移敏感的问题，但离线无校准方法消除了这些障碍。然而，由于它们依赖于瞬时零阶近似，其中输入和输出差异之间的映射实时变化，因此容易受到观测噪声的影响，并忽略了扩散轨迹内的内在动量。在本文中，我们提出NaviCache，一种即插即用的测试时自校准方法，将特征演化重新构想为惯性导航系统（INS）问题。NaviCache通过建模输入和输出变化之间的相对耦合，弥合了基本领域差距和扩散的非平稳特性。我们引入了一种双状态估计架构，通过专门的初始对准阶段初始化，自适应地跟踪特征变化比及其潜在漂移。通过将时间相关的噪声调度与不确定性感知的测量更新机制相结合，NaviCache为有界误差的计算跳过提供了理论基础的机制。在HunyuanVideo、Wan和Open-Sora系列上的大量实验表明，NaviCache在计算跳过时表现出更准确的误差判断，并实现了出色的综合性能。
英文摘要：
Video Diffusion Models (VDMs) is constrained by immense computational costs. While offline calibration-based acceleration suffers from calibration data dependency, prohibitive calibration duration, and susceptibility to distribution shifts, offline calibration-free methods eliminate these hurdles. However, since they rely on instantaneous zero-order approximations where the mapping between input and output differences varies in real-time, they are susceptible to observational noise and ignore the intrinsic momentum within the diffusion trajectory. In this paper, we propose NaviCache, a plug-and-play test-time self-calibration method re-conceptualizing feature evolution as an Inertial Navigation System (INS) problem. NaviCache bridges the fundamental domain gap and the non-stationary nature of diffusion by modeling the relative coupling between input and output variations. We introduce a dual-state estimation architecture that adaptively tracks the feature change ratio and its latent drift, initialized via a specialized Initial Alignment phase. By integrating a time-dependent noise schedule with an uncertainty-aware Measurement Update mechanism, NaviCache provides a theoretically grounded mechanism for error-bounded computation skipping. Extensive experiments on the HunyuanVideo, Wan, and Open-Sora series demonstrate that NaviCache exhibits more accurate error judgment for computation skipping and achieves outstanding comprehensive performance.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
