arXiv cs.AI· Jianhui Zhang, Zuyao Lin, Pengxin Xu, Jingdong Wang, Xingyu Chen·· 3 天前AI 评分26
4LC：面向动态世界建模高效条件生成的 4D 潜在缓存
4LC: 4D Latent Cache for Efficient Conditioning in Dynamic World Modeling
AI 导读
研究者提出 4LC，一种带复合视觉表征与灵活时空访问接口的 4D 潜在缓存，可直接复用静态与动态内容的视觉表征作为生成条件，避免逐次 RGB 渲染与重编码。4LC 将 VAE、DINO、V-JEPA、VGGT 等多个视觉编码器的表征绑定到共享几何与运动上，并支持在任意指定时间与视角读取缓存。
正文 · 原文
Abstract:For video generation in dynamic world modeling, observed static and dynamic content should be efficiently reused to provide conditions for repeated generation at different times and viewpoints. We introduce 4LC, a 4D latent cache with composite visual representations and a flexible spatiotemporal access interface. As a latent cache, 4LC directly reuses visual representations of static and dynamic content to provide generation conditions, avoiding per-request RGB rendering and re-encoding. As a composite cache, 4LC binds visual representations from multiple visual encoders (e.g., VAE, DINO, V-JEPA, VGGT) to shared geometry and motion, allowing the composite representations to describe the dynamic world with rich information. Besides, we design a flexible spatiotemporal interface, which directly reads the cache at independently specified times and viewpoints and combines readouts across observations, providing a common conditioning format for diverse downstream tasks. Building on 4LC, we develop 4LC-Video to address multiple video generation tasks through different configurations of source observations, spatiotemporal requests, and historical reuse, including camera control, temporal control, novel-view synthesis, scene revisits, and long-term generation. Our experiments show more efficient condition acquisition, effective spatiotemporal controllability, and scene consistency across tasks using a single 4LC-Video model without task-specific fine-tuning.
| Comments: | Project page: this https URL |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.17625 [cs.CV] |
| (or arXiv:2610.17625v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.17625 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jianhui Zhang [view email]
[v1]
Tue, 13 Oct 2026 02:47:58 UTC (11,757 KB)
来源：arXiv cs.AI · arxiv.org