这篇论文揭示了视频世界模型长视频生成不稳定的根本原因——表示维度塌缩，并提出了一个简单有效的正则化方法，在VBench上大幅超过Diffusion Forcing。
视频扩散世界模型在自动驾驶、机器人等任务中生成长视频时，滑动窗口自回归推理会导致误差累积，使帧质量随时间下降。研究发现，复合误差与隐藏表示的维度塌缩紧密相关，有效秩在生成漂移开始时急剧下降。单纯扩大训练数据无法提升模型抗误差漂移能力。提出视频表示正则化，通过轻量训练约束稳定潜在表示，抑制迭代误差累积。相比Diffusion Forcing方法，在VBench的美学质量指标上从38.65提升至55.56，成像质量从44.37提升至72.08。
Mitigating Compounding Error via Video Representation Regularization
Video diffusion-based world models enable long autoregressive video generation for robotics, autonomous driving and simulation tasks, yet sliding-window autoregressive inference suffers from severe error accumulation that degrades frame quality over time. Although this phenomenon has been widely observed, the underlying mechanism of compounding error and how to achieve stable long-horizon generation remain largely unresolved. In this paper, we investigate the internal representation dynamics of video world models and discover that compounding error is tightly coupled with dimensional collapse of hidden representations. Specifically, the effective rank of model representations sharply decreases at the onset of generation drift, revealing a strong connection between representational degradation and long-term rollout instability. Furthermore, we find that pure training data scaling fails to boost model resistance to error drift, contradicting mainstream scaling paradigms. To address this problem, we propose video representation regularization, a lightweight training constraint that stabilizes latent representations and suppresses iterative error accumulation. Compared with Diffusion Forcing, our method achieves improvements from 38.65 to 55.56 and from 44.37 to 72.08 on the Aesthetic Quality and Imaging Quality metrics of VBench. Our work establishes the first connection between autoregressive video drifting and model internal representations, adopts erank as a quantitative metric for error accumulation, reveals counterintuitive scaling limitations for video world models, and presents a simple yet effective regularization strategy to improve long video generation robustness.