iWorld-Bench: A Benchmark for Interactive World Models with a Unified Action Generation Framework

TL;DR

iWorld-Bench introduces a comprehensive benchmark with a large dataset and unified evaluation framework for assessing interactive world models' perception, memory, and action capabilities in diverse scenarios.

Contribution

The paper presents iWorld-Bench, a new large-scale dataset and a unified action generation framework for evaluating interactive world models across multiple tasks.

Findings

Evaluated 14 world models revealing key limitations.

Constructed a dataset with 330k video clips and 4.9k test samples.

Provided insights for future development of interactive world models.

Abstract

Achieving Artificial General Intelligence (AGI) requires agents that learn and interact adaptively, with interactive world models providing scalable environments for perception, reasoning, and action. Yet current research still lacks large-scale datasets and unified benchmarks to evaluate their physical interaction capabilities. To address this, we propose iWorld-Bench, a comprehensive benchmark for training and testing world models on interaction-related abilities such as distance perception and memory. We construct a diverse dataset with 330k video clips and select 2.1k high-quality samples covering varied perspectives, weather, and scenes. As existing world models differ in interaction modalities, we introduce an Action Generation Framework to unify evaluation and design six task types, generating 4.9k test samples. These tasks jointly assess model performance across visual generation, trajectory following, and memory. Evaluating 14 representative world models, we identify key limitations and provide insights for future research. The iWorld-Bench model leaderboard is publicly available at iWorld-Bench.com.