MEGA Hub

Orbia: Do Generated Videos Form a Consistent 3D World?

Authors

Do you know Dongyue Lu?You can claim authorship or link another user.Do you know Guian Fang?You can claim authorship or link another user.Do you know Hur Lim?You can claim authorship or link another user.Do you know Yonggang Wu?You can claim authorship or link another user.Do you know Yanrong Wang?You can claim authorship or link another user.Do you know Hongshan Chen?You can claim authorship or link another user.Do you know Shengju Qian?You can claim authorship or link another user.Do you know Yue Liao?You can claim authorship or link another user.Do you know Wei Chow?You can claim authorship or link another user.Do you know Lingdong Kong?You can claim authorship or link another user.Do you know Tianxin Huang?You can claim authorship or link another user.Do you know Yu Yang?You can claim authorship or link another user.Do you know Hongji Yang?You can claim authorship or link another user.Do you know Junchao Huang?You can claim authorship or link another user.Do you know Ao Liang?You can claim authorship or link another user.Do you know Yihao Wang?You can claim authorship or link another user.Do you know Zihan Wang?You can claim authorship or link another user.Do you know Rong Li?You can claim authorship or link another user.Do you know Hanlin Chen?You can claim authorship or link another user.Do you know Xin Wang?You can claim authorship or link another user.Do you know Wei Tsang Ooi?You can claim authorship or link another user.Do you know Mike Zheng Shou?You can claim authorship or link another user.Do you know Shuicheng Yan?You can claim authorship or link another user.

Abstract

Video generators serve as interactive worlds for exploration and simulation, yet realistic and visually appealing frames do not necessarily depict a consistent 3D world. Existing benchmarks assess selected geometric properties or use vision-language models to judge memory and consistency, leaving systematic 3D evaluation underexplored. We introduce Orbia, the first benchmark dedicated to a systematic and comprehensive evaluation of 3D consistency in interactive world models. Through controlled exploration and revisit, we organize the evaluation into three dimensions: input preservation, generated-content persistence, and 3D self-consistency. For input preservation, registered RGB-D references and annotated anchors support appearance, geometry, and spatial stability measurements. For generated-content persistence, we compare revisits to regions absent from the input image with their first generated observations to test whether models preserve what they create. For 3D self-consistency, cross-view depth and feature correspondence, regional scale agreement, and held-out Gaussian rendering test whether generated views describe the same 3D scene. Minute-long exploration and revisits further test long-term consistency. Our construction and annotation pipeline brings together public datasets, self-captured scenes, and simulation, yielding 720 instances across indoor and outdoor environments and six exploration-and-revisit protocols. Evaluation spans 27 text-, action-, and camera-conditioned models through native-interface adapters. Our Orbia benchmark provides a systematic basis for diagnosing 3D inconsistency and developing more persistent video world models.

Community

00