Paper 2605.25874
WBench: A Comprehensive Multi-turn Benchmark for Interactive Video World Model Evaluation
- Published
- May 2026
- Research lab
- Independent
- Citations
- 7
- GitHub
- 175 stars
01 In brief
Summary
WBENCH is a comprehensive multi-turn benchmark for evaluating interactive video world models across five dimensions: video quality, setting adherence, interaction adherence, consistency, and physics compliance.
It contains 289 test cases and 1,058 interaction turns, covering diverse scenes, styles, subjects, and both first- and third-person perspectives.
The benchmark includes four interaction types: navigation, subject action, event editing, and perspective switching.
Navigation is unified across text, 6-DoF pose, and discrete-action control.
Evaluation uses 22 automatic sub-metrics combining specialist vision models and large multimodal models, all validated against human judgments.
Experiments on 20 state-of-the-art models reveal that no single model performs strongly across all dimensions.
Key findings include: navigation is largely independent of other dimensions; camera control does not guarantee perspective consistency; physical correctness correlates with rendering quality rather than control; benchmark difficulty is structured by perspective, scene type, and subject category; and navigation degrades most over turns.
The benchmark provides diagnostic insights into model strengths and weaknesses.
Code and data are available at https://github.com/meituan-longcat/WBench.
02 From the paper
Abstract
Interactive world models are advancing rapidly, yet existing benchmarks cover only part of the required competencies, leaving no unified standard for systematic evaluation. To fill this gap, we introduce WBench, a comprehensive multi-turn benchmark for interactive world model evaluation along five dimensions, namely video quality, setting adherence, interaction adherence, consistency, and physics compliance. WBench contains 289 test cases and 1,058 interaction turns, where each case specifies a world setting and a multi-turn interaction sequence, covering diverse scenes, styles, subjects, and both first- and third-person perspectives, together with four interaction types, including navigation, subject action, event editing, and perspective switching. For navigation, WBench unifies text, 6-DoF pose, and discrete-action control, enabling evaluation of models with different native input interfaces. Evaluation uses 22 automatic sub-metrics that combine specialist vision models with large multimodal models, and all metrics are validated against human judgments. Across 20 state-of-the-art models, we find that no single model performs strongly across all dimensions. We provide detailed diagnostic insights into the characteristic strengths, weaknesses, and open challenges of each model. Code and data are available at https://github.com/meituan-longcat/WBench.