发表机构
School of Computer Science, Shanghai Jiao Tong University, Shanghai, China; Institute of Cyber-Systems and Control, Zhejiang University, Hangzhou, China; Nanyang Technological University, Singapore（上海交通大学计算机科学学院; 浙江大学控制系统研究所; 新加坡南洋理工大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对自回归扩散模型在长视频生成中误差累积问题，提出循环世界框架，通过训练和推理阶段的时间可逆性及反向预测模型抑制误差，实验证明其在VBench基准测试中显著减轻误差漂移，提升生成质量和时间一致性。
AI 中文摘要
自回归扩散模型实现了高质量视频生成，但其序列性质存在误差累积问题。在长视频合成中，微小预测偏差会随时间加剧，导致生成漂移、结构崩溃和视觉退化。为此提出循环世界框架，通过在训练和推理阶段强制严格的时间可逆性来解决误差漂移。理论上，正向生成漂移可由循环一致性目标严格限制。训练时集成高效反向预测模型，推理时将其用作运行时校正器，通过基于梯度的循环引导抑制累积误差。在VBench基准测试上的实验表明，该框架显著减轻误差漂移，在60秒合成中实现了高质量和时间一致性。
英文摘要
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 causal constraints into the forward generator, compelling it to produce reversible sequences that tightly adhere to the natural video manifold. At inference time, we repurpose this frozen reverse model as a runtime corrector. Through gradient-based cycle guidance, it iteratively refines the generated latent representations, actively suppressing accumulated errors before they are committed to the historical context. Extensive experiments on the VBench benchmark demonstrate that Cycle-World's dual-phase synergy significantly mitigates error drift, achieving state-of-the-art overall generation quality and long-horizon temporal consistency in 60-second synthesis.
CommentsAccepted by ECCV 2026
