Composition of Memory Experts for Diffusion World Models

ArXi:2605.18813v1 Announce Type: cross World models aim to predict plausible futures consistent with past observations, a capability central to planning and decision-making in reinforcement learning. Yet, existing architectures face a fundamental memory trade-off: transformers preserve local detail but are bottlenecked by quadratic attention, while recurrent and state-space models scale efficiently but compress history at the cost of fidelity.