Graph World Models for Constrained Epidemic Policy Planning
Read the original on arXiv AI →Graph World Models for Constrained Epidemic Policy Planning introduces EpiMind, a graph-factored recurrent state-space model that generates joint policy-conditioned rollouts from regional latent beliefs. The framework uses graph-temporal ADMM to optimize regional interventions, enforce shared-resource feasibility, and evaluate temporal specifications under the learned model. EpiMind achieves a 29% reduction in admission RMSE compared to graph-free dynamics, plans within 1–5% of the best feasible constant policy while guaranteeing shared-budget feasibility, and outperforms all deployable baselines across three resource budgets in real-context evaluation.
Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.