Homeostatic Continual Learning Framework to Prevent Catastrophic Forgetting

This method proposes an agentic approach to continual learning by identifying environmental outliers when output outliers are detected. The framework aims to build world models by factorizing objects into features and mapping abstracted concepts to intents.