The Global Neural World Model: Spatially Grounded Discrete Topologies for Action-Conditioned Planning

TL;DR

The paper introduces GNWM, a neural framework that creates stable, topologically grounded discrete maps for action-conditioned planning, enabling better environment modeling and causal discovery.

Contribution

It presents a novel self-stabilizing architecture that enforces topological quantization and translational equivariance without pixel-level reconstruction, enhancing environment understanding.

Findings

Prevents manifold drift during autoregressive rollouts.

Learns generalized transition dynamics through maximum entropy exploration.

Acts as both a spatial physics simulator and a causal discovery model.

Abstract

We present the Global Neural World Model (GNWM), a self-stabilizing framework that achieves topological quantization through balanced continuous entropy constraints. Operating as a continuous, action-conditioned Joint-Embedding Predictive Architecture (JEPA), the GNWM maps environments onto a discrete 2D grid, enforcing translational equivariance without pixel-level reconstruction. Our results show this architecture prevents manifold drift during autoregressive rollouts by using grid ``snapping'' as a native error-correction mechanism. Furthermore, by training via maximum entropy exploration (random walks), the model learns generalized transition dynamics rather than memorizing specific expert trajectories. We validate the GNWM across passive observation, active agent control, and abstract sequence regimes, demonstrating its capacity to act not just as a spatial physics simulator, but as a causal discovery model capable of organizing continuous, predictable concepts into structured topological maps.