Federated inference and belief sharing

This paper formulates federated inference and belief sharing as a principled approach to distributed intelligence. By extending Active Inference to multi-agent settings, agents maintain local generative models while sharing beliefs through message passing to achieve collective inference. The framework addresses how agents can coordinate without sharing raw observations, preserving privacy while...