PAWA: A multi-agent panoramic action-conditioned world model dataset and method
What happened
PAWA is an action-conditioned world modeling study for multiplayer Minecraft scenarios. It introduces a synchronized POV and panoramic video dataset, a collection pipeline, and a prediction framework. Author Siqi Yang says the framework uses a two-tower structure to jointly predict two video streams, and cross-attention to align tokens facing the same direction in the world coordinate system. According to the paper, PAWA's POV prediction quality outperforms existing multiplayer world models and can generate high-fidelity panoramic predictions. These conclusions come from the paper's own reporting and have not been independently verified. The dataset and collection pipeline are open-sourced to support follow-up research on multiplayer world models.
AI-written from the coverage · updated 2026-10-15
Coverage timeline
Follow the coverage to see the story from different sides.
- arXiv Game AIPickPAWA: Panoramic action-conditioned world modeling for multiple agents
PAWA presents a synchronized POV and panoramic video dataset and framework for action-conditioned multi-agent world modeling in multiplayer Minecraft scenarios. The model uses a two-tower structure to jointly predict two video streams, and cross-attention aligns tokens facing the same direction in a shared world coordinate system. The paper reports that its POV prediction quality outperforms existing multiplayer world models while also generating high-fidelity panoramic predictions. The dataset and collection pipeline are open-sourced.