TetherCache通过分层缓存与动态对齐，解决长视频生成中的质量退化问题。
TetherCache: Stabilizing Autoregressive Long-Form Video Generation with Gated Recall and Trusted Alignment
- 分三区管理缓存，用门控机制选择关键历史帧以保持多样性。
- 在240秒生成中，质量退化从7.84降至1.33，语义评分显著提升。
- 无需训练、即插即用，适合追求稳定长视频生成的研究者与开发者。
自回归视频扩散模型通过依赖已生成内容来自然实现流式与可变长度视频生成。然而，将此类模型扩展至分钟级生成仍具挑战：有限的键值缓存预算导致无法保留完整历史，而反复依赖自生成帧会引发上下文分布漂移，随时间累积造成视觉伪影、质量下降和时序漂移。本文提出TetherCache，一种无需训练、即插即用的缓存管理策略，用于实现抗漂移的长视频生成。TetherCache将缓存分为锚点、记忆和近期区域，并引入两项互补机制。首先，GRAB（门控回忆与注意力多样性平衡）通过结合基于注意力的相关性与时间多样性，选取长程记忆帧，在固定缓存预算下保留信息丰富且多样化的历史上下文。其次，TAME（通过记忆编辑实现可信对齐）通过将新召回的记忆项统计特性对齐至可信上下文分布，减轻因历史特征漂移带来的污染。基于Self-Forcing，TetherCache在VBench-Long数据集上持续提升30秒、60秒及240秒生成质量。尤其在240秒生成中，整体评分与语义评分大幅提升，质量漂移由7.84降至1.33，验证了其在长时序自回归视频扩散中的有效性。
原文摘要 · Abstract (English)
Autoregressive video diffusion models provide a natural formulation for streaming and variable-length video generation by conditioning newly generated frames on previously generated content. However, extending these models to minute-level generation remains challenging: the limited KV-cache budget prevents the model from retaining the full history, while repeatedly conditioning on self-generated frames induces a context distribution shift that accumulates over time, leading to visual artifacts, quality degradation, and temporal drift. In this paper, we propose TetherCache, a training-free and plug-and-play cache management strategy for drift-resistant long video generation. TetherCache organizes the cache into sink, memory, and recent regions, and introduces two complementary mechanisms. First, GRAB (Gated Recall with Attention-Diversity Balancing) selects long-range memory frames using a gated score that combines attention-based relevance with temporal diversity, preserving informative yet diverse historical context under a fixed cache budget. Second, TAME (Trusted Alignment via Memory Editing) lightly edits newly recalled memory tokens by aligning their statistics to a trusted context distribution, reducing the pollution caused by drifted historical features. Built on Self-Forcing, TetherCache consistently improves long-video generation quality on VBench-Long across 30s, 60s, and 240s settings. In particular, for 240s generation, it substantially improves overall and semantic scores while reducing quality drift from 7.84 to 1.33, demonstrating its effectiveness for stable long-horizon autoregressive video diffusion.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。