OverSeer world model adds fixed auxiliary cameras to keep multi-agent video consistent
A new arXiv preprint (2610.28554, submitted 20 Oct 2026) from Runjia Li and co-authors introduces OverSeer, a multi-agent video world model that augments each agent's egocentric view with auxiliary camera views. The authors argue that fixing these auxiliary views in space makes them a proxy for a world-centric representation, and report that they substantially help the model maintain consistency; a delta-VAE encoder that stores only changes relative to a static reference image cuts the tokens needed to process them. Experiments in Minecraft and Doom are said to show improved world and agent consistency, with qualitative results on real-world multi-camera captures, though all results are self-reported and not yet peer-reviewed.
The paper offers a simple, compression-backed way to tackle cross-agent consistency, a core obstacle to extending video world models from single agents to teams.
Computer Science > Computer Vision and Pattern Recognition
arXiv:2610.28554 (cs)
[Submitted on 20 Oct 2026]
Title:OverSeer: Multi-Agent World Models Beyond What Agents See
Authors:Runjia Li, Kam Woh Ng, Minghao Chen, Junlin Han, Jianyuan Wang, Zhaochong An, Juan-Manuel Perez-Rua, Andrea Vedaldi, Shikun Liu
View a PDF of the paper titled OverSeer: Multi-Agent World Models Beyond What Agents See, by Runjia Li and 8 other authors