面向实体中心 4D 场景生成的接地潜变量
Grounded Latents for Entity-Centric 4D Scene Generation
用实体级 3D 潜变量替代稠密体素表示，把场景拆成语义局部区域再逐实体扩散生成，并沿时间传播潜变量得到 4D 一致性；对做驾驶场景生成、可控仿真数据合成的人，这是一条绕开体素开销、又能精细控制单个车辆/行人的可复用路线。
原文依据 a framework for generative modeling in a compact, entity-centric latent space, where each grounded 3D latent represents a semantically meaningful local region
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论尽管近期工作已探索了 3D 或 4D 驾驶场景的生成建模，但多数方法在稠密体素表示上运行，计算开销大，且难以保持时序或结构一致性。这些方法往往产生模糊或粘连的实体（即汽车、卡车、行人），并且缺乏对单个场景元素的细粒度控制。我们提出 LatentWorld，一个在紧凑的、以实体为中心的潜空间中进行生成建模的框架，其中每个接地 3D 潜变量表示场景中一个语义上有意义的局部区域。这一表述使得在保留几何细节的同时，能够对前景和背景元素进行精确且一致的控制。我们进一步将该表示扩展到 4D：以生成的 3D 场景为条件，为自车与动态参与者学习一个运动扩散模型，并让接地潜变量随时间传播。我们的框架生成物理一致且时序连贯的 4D 场景，支持可控且逼真的生成。
英文原文摘要 1056 字符
Although recent work has explored generative modeling of 3D or 4D driving scenes, most approaches operate on dense voxel-based representations, which are computationally expensive and struggle to maintain temporal or structural consistency. These methods often produce blurred or merged entities (i.e., cars, trucks, pedestrians) and lack fine-grained control over individual scene elements. We propose LatentWorld, a framework for generative modeling in a compact, entity-centric latent space, where each grounded 3D latent represents a semantically meaningful local region of the scene. This formulation enables precise, consistent control of both foreground and background elements while preserving geometric detail. We further extend this representation to 4D by learning a motion diffusion model for both ego and dynamic actors, conditioned on the generated 3D scene, and by propagating the grounded latents through time. Our framework produces physically consistent and temporally coherent 4D scenes, supporting controllable and realistic generation.