A Benchmark for Interactive World Models with a Unified Action Generation Framework
For researchers developing interactive world models for AGI, this provides a much-needed large-scale benchmark and unified evaluation framework, though it is an incremental contribution building on existing benchmarks.
The paper introduces iWorld-Bench, a benchmark with 330k video clips and 2.1k high-quality samples for evaluating interactive world models on physical interaction tasks. It proposes an Action Generation Framework to unify evaluation across 6 task types, generating 4.9k test samples, and identifies key limitations in 14 existing models.
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.