WorldLens: Full-Spectrum Evaluations of Driving World Models in Real World

TL;DR

WorldLens introduces a comprehensive benchmark, dataset, and evaluation model to assess the realism, physics, and behavior of generative driving world models, addressing a key gap in embodied AI evaluation.

Contribution

It provides the first unified framework for evaluating geometric, physical, and behavioral fidelity of driving world models, including a large annotated dataset and an explainable scoring agent.

Findings

No existing model excels across all aspects of realism and behavior.

Textures often violate physics, while geometry-stable models lack behavioral fidelity.

The benchmark and dataset enable standardized, human-aligned evaluation of world models.

Abstract

Generative world models are reshaping embodied AI, enabling agents to synthesize realistic 4D driving environments that look convincing but often fail physically or behaviorally. Despite rapid progress, the field still lacks a unified way to assess whether generated worlds preserve geometry, obey physics, or support reliable control. We introduce WorldLens, a full-spectrum benchmark evaluating how well a model builds, understands, and behaves within its generated world. It spans five aspects -- Generation, Reconstruction, Action-Following, Downstream Task, and Human Preference -- jointly covering visual realism, geometric consistency, physical plausibility, and functional reliability. Across these dimensions, no existing world model excels universally: those with strong textures often violate physics, while geometry-stable ones lack behavioral fidelity. To align objective metrics with human judgment, we further construct WorldLens-26K, a large-scale dataset of human-annotated videos with numerical scores and textual rationales, and develop WorldLens-Agent, an evaluation model distilled from these annotations to enable scalable, explainable scoring. Together, the benchmark, dataset, and agent form a unified ecosystem for measuring world fidelity -- standardizing how future models are judged not only by how real they look, but by how real they behave.