[论文] WorldAlign: Decoupled 4D Reward for World-Consistent Video Generation

论文概要

研究领域: CV
作者: Jing He, Kaixin Ding, Xingye Tian, Guibao Shen, Wenhang Ge, Xin Tao, Pengfei Wan, Ying-Cong Chen
发布时间: 2026-10-08
arXiv: 2610.12382

中文摘要

忠实的视觉世界模拟要求生成视频保持 4D 世界一致性，包括静态一致性和动态一致性。几何感知后训练是提升世界一致性的有前景路径，但现有方法多依赖静态场景假设；即便支持动态场景的方法也难以提供可靠的静态一致性反馈，动态一致性则常被忽视或评估不足。为此我们提出 WorldAlign——解耦的 4D 奖励框架，在语义层面分离静态区域和动态主体，让各自与适配其假设的世界先验对齐。静态区域通过语义引导的掩码重投影与几何世界先验对齐，实现更可靠的静态一致性评估；辅助的相机运动奖励抑制近似静态的退化解。动态主体则使用强视觉-语言模型（VLM）作为动态世界先验，基于样本特定清单构建 VLM-as-judge 奖励，评估动态性、物理合理性、形状和纹理一致性。该解耦设计无需人工偏好标注即可进行更有效的在线后训练。在两个预训练图生视频模型 Wan2.1 和 Wan2.2 上，WorldAlign 联合提升静态与动态一致性，且不抑制整体或主体运动。

原文摘要

Faithful visual world simulation requires generated videos to maintain 4D world consistency, encompassing both static and dynamic consistency. Static consistency requires coherent 3D structure in static environments across viewpoints, while dynamic consistency requires plausible subject motion and consistent appearance over time. Geometry-aware post-training offers a promising way to improve world consistency. However, existing methods often rely on a static-scene assumption. Even those that accommodate dynamic scenes struggle to provide reliable static-consistency feedback, while dynamic consistency is often overlooked or inadequately assessed. To address these limitations, we introduce WorldAlign, a decoupled 4D reward framework that semantically separates static regions and dynamic subj...

自动采集于 2026-10-12

#论文 #arXiv #CV #小凯