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

arXiv 2026 43.3 method

TLDR

Orbia is a benchmark for systematically evaluating 3D consistency in interactive world models via controlled exploration, revisits, and metrics across 720 instances.

Reasoning

The paper addresses a clear gap in evaluating 3D consistency of video-based world models, with a structured benchmark and diverse data sources. Its main limitation is that no experimental results or baseline comparisons are visible in the abstract, so the benchmark's utility and findings are not yet demonstrated.

Read-first score

Read-first score 43.3, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 51.

Recency 5%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 15%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=benchmark,dataset,evaluation,experiment,metric

Topical relevance 25%

72.9

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 15%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=dataset

Graph prestige 15%

5.7

PageRank-style prestige over the local citation graph built from available OpenAlex references. local_pagerank=0.0016

Citation impact 15%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 10%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 654.

Keyword Scores

world model

interactive world model

generative world model

video world model

world simulator

world dynamics prediction

model-based reinforcement learning world model

Deep Analysis

Innovations

Introduces Orbia, the first benchmark dedicated to systematic and comprehensive evaluation of 3D consistency in interactive world models.
Organizes evaluation into three dimensions: input preservation, generated-content persistence, and 3D self-consistency using controlled exploration and revisit.
Combines public datasets, self-captured scenes, and simulation to yield 720 instances across indoor and outdoor environments with six exploration-and-revisit protocols.
Evaluates 27 text-, action-, and camera-conditioned models through native-interface adapters.

Methodology

Orbia evaluates video world models through controlled exploration and revisit across three dimensions: input preservation using registered RGB-D references and annotated anchors; generated-content persistence by comparing revisits to regions absent from the input with first generated observations; and 3D self-consistency using cross-view depth and feature correspondence, regional scale agreement, and held-out Gaussian rendering. Minute-long exploration and revisits test long-term consistency across 720 indoor and outdoor instances.

Key Results

The abstract does not report quantitative experimental outcomes; it states that the benchmark provides a systematic basis for diagnosing 3D inconsistency and evaluating 27 models across six protocols.

Resources

Paper