Looped World Models
2026-06-16 • Machine Learning
Machine LearningArtificial IntelligenceComputation and LanguageComputer Vision and Pattern Recognition
AI summaryⓘ
The authors introduce Looped World Models (LoopWM) to improve how machines predict future states in simulations. Their model repeatedly updates its internal understanding of the environment using a shared transformer module, making it much more efficient in terms of parameters than traditional methods. This approach adjusts how much computation it uses based on the difficulty of each prediction step. The authors suggest that this new way of increasing the internal processing depth, rather than just making models bigger, could advance world modeling research.
world modelstransformerlatent statesparameter efficiencyadaptive computationsimulationiterative refinementdeep learning
Authors
Hongyuan Adam Lu, Z. L. Victor Wei, Qun Zhang, Jinrui Zeng, Bowen Cao, Lingwei Meng, Mocheng Li, Zezhong Wang, Haonan Yin, Naifu Xue, Minyu Chen, Cenyuan Zhang, Zefan Zhang, Hao Wei, Jiawei Zhou, Haoran Xu, Hao Yang, Ronglai Zuo, Tongda Xu, Yonghao Li, Jian Chen, Hebin Wang, Zeyu Gao, Yang Li, Wei Zhao, Qimin Zhong, Siqi Liu, Yumeng Zhang, Leyan Cui, Zhangyu Wang, Wai Lam
Abstract
Current world models face a fundamental tension: faithful long-horizon simulation demands deep computation, but deeper models are expensive to deploy and prone to compounding errors. We resolve this by introducing Looped World Models (LoopWM), which are the first looped architectures for world modelling. Our method iteratively refines latent environment states through a parameter-shared transformer block. This yield up to 100x parameter efficiency over conventional approaches with adaptive computation that automatically scales depth to match the complexity of each prediction step. Orthogonal to scaling model size and training data, LoopWM establishes iterative latent depth as a new scaling axis for world simulation, which might significantly push the community forward.