AAMAS 2026 Blue Sky Award Winner: Foundation World Models for Agents in Changing Environments

Summary

Florent Delgrange won the Best Blue Sky Paper Award at AAMAS 2026 for his work 'Foundation World Models for Agents that Learn, Verify, and Adapt Reliably Beyond Static Environments.' The paper proposes a research agenda that merges reinforcement learning and reactive synthesis into a single loop: a foundation world model is learned from experience but structured for a verifier to reason about. As the agent learns a policy, it updates the model, measures the reliability of its abstraction, and checks whether the policy still satisfies its specification. The verifier's feedback can reject unsafe updates, request data from uncertain regions, or trigger model revision. This matters because real environments change, and in multi-agent systems, every adapting agent alters the environment perceived by others. Reliability cannot be a one-time certificate at deployment; it must be maintained as the agent continues learning.