论文概要
研究领域: CV 作者: Zihan Su, Teng Hu, Jiangning Zhang, Ruiyan Wang, Ran Yi 发布时间: 2026-07-13 arXiv: 2607.11836
中文摘要
自回归扩散模型实现了高质量视频生成，但其序列本质固有地遭受误差累积。在长时程视频合成中，微小预测偏差随时间复合，不可避免地导致无约束的生成漂移、结构崩溃和严重视觉退化。本文提出Cycle-World，一种为稳定且时间一致的长视频生成设计的新框架。该方法通过强制训练和推理阶段的时间可逆性来解决误差漂移。理论上证明前向生成漂移可被循环一致性目标严格瓶颈。训练时，集成高效的反向预测模型将因果约束隐式嵌入前向生成器，迫使其产生可逆序列，紧密贴合自然视频流形。推理时，将该冻结反向模型重新用作运行时校正器。通过基于梯度的循环引导，它迭代优化生成潜在表征，在累积误差被提交到历史上下文前主动抑制。在VBench基准上的大量实验表明Cycle-World的双阶段协同显著缓解误差漂移，在60秒合成中达到最优整体生成质量和长时程时间一致性。
原文摘要
Autoregressive diffusion models have enabled high-quality video generation, yet their sequential nature inherently suffers from error accumulation. In long-horizon video synthesis, minor prediction deviations compound over time, inevitably leading to unconstrained generative drift, structural collapse, and severe visual degradation. To address this, we propose Cycle-World, a novel framework designed for stable and temporally consistent long-video generation. Our approach tackles error drift by enforcing strict temporal reversibility across both the training and inference phases. Theoretically, we demonstrate that forward generative drift can be strictly bottlenecked by a cycle-consistency objective. During training, we integrate an efficient reverse-prediction model to implicitly embed cau…
— 自动采集于 2026-07-15
#论文 #arXiv #CV #小凯