This is not a typical survey of world models; it is a guide for those who want to build worlds. We do not aim to catalog every paper that has ever mentioned a ``world model". Instead, we follow one clear road: from early masked models that unified representation learning across modalities, to unified architectures that share a single paradigm, then to interactive generative models that close the action-perception loop, and finally to memory-augmented systems that sustain consistent worlds over time. We bypass loosely related branches to focus on the core: the generative heart, the interactive loop, and the memory system. We show that this is the most promising path towards true world models.
@article{arxiv.2510.20668,
title = {From Masks to Worlds: A Hitchhiker's Guide to World Models},
author = {Jinbin Bai and Yu Lei and Hecong Wu and Yuchen Zhu and Shufan Li and Yi Xin and Xiangtai Li and Molei Tao and Aditya Grover and Ming-Hsuan Yang},
journal= {arXiv preprint arXiv:2510.20668},
year = {2025}
}