نظرة عامة
Anima investigates the intersection of Active Inference and current LLM-based AI — examining how Active Inference models of perception, policy, and action relate to LLM behavior, and how the framework can guide AI design for coherent, aligned behavior. It explores whether a full Active Inference pipeline — predictive processing, allostasis, a Markov blanket, and policy selection — can give an artificial agent genuine behavioral continuity and endogenous initiative between interactions.