The Sleep Framework for LLM Continual Learning

The Sleep framework provides a method for Large Language Models to retain knowledge during continual learning. It aims to mitigate catastrophic forgetting by simulating sleep-like processes to consolidate new information without degrading previous capabilities.