Newest papers

OverSeer: Multi-Agent World Models Beyond What Agents See

Rui Li, Kam Woh Ng, Minghao Chen, Junlin Han, Jianyuan Wang, Zhaochong An, Juan-Manuel Pérez-Rúa, Andrea Vedaldi, Shikun Liu · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28554 · cited by 0 works (OpenAlex) · open-access PDF

French

Malgré le succès des modèles mondiaux fixes, les étendre à plusieurs agents est difficile car il nécessite de générer plusieurs flux vidéo spécifiques aux agents qui sont cohérents dans le temps et entre les agents. Pour résoudre ce fossé, nous introduisons OverSeer, un modèle mondial multi-agent qui augmente les vues sur les agents égoïstes avec des vues sur les caméras auxiliaires. Nous suggérons que, en fixant les vues auxiliaires dans l'espace, elles agissent comme un proxy pour une représentation centrée sur le monde: Contrairement aux vues des agents, qui peuvent changer complètement en raison d'un simple changement de position des agents, nous introduisons des changements dans les vues auxiliaires directement sur les changements localisés dans l'état du monde. Nous

Machine translation (Meta M2M100); the English original follows.

Abstract

Despite the success of video world models, extending them to multiple agents is difficult because it requires generating several agent-specific video streams that are consistent in time and between agents. To address this gap, we introduce OverSeer, a multi-agent world model that augments egocentric agent views with auxiliary camera views. We suggest that, by fixing the auxiliary views in space, they act as a proxy for a world-centric representation: Unlike the agents' views, which can change entirely due to a simple change in agent position, changes in the auxiliary views directly map to localized changes in the world state. We show that these fixed views substantially help the model to maintain consistency. We also show that, due to their redundant and semi-static nature, the auxiliary views can be efficiently compressed with a delta-VAE that encodes only the changes relative to a static reference image, reducing the number of tokens required to process them. Experiments in Minecraft and Doom demonstrate improved world and agent consistency, with qualitative results on real-world multi-camera captures indicating that the approach is not tied to game engines. Our compression analysis further shows that delta-VAE substantially reduces the computational overhead of auxiliary-view generation. We conclude that these fixed views provide a simple and efficient way to improve multi-agent world models.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Simple (philosophy), Process (computing), Overhead (engineering), Real world data, Proxy (statistics)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange