Language Models Need Sleep: Learning to Self-Modify and Consolidate Memories
2026-06-02 • Machine Learning
Machine LearningArtificial Intelligence
AI summaryⓘ
The authors propose a new 'Sleep' method to help machine learning models learn continuously, similar to how humans learn over time. Their approach has two parts: first, 'Memory Consolidation,' where smaller models teach bigger ones to keep important knowledge; second, 'Dreaming,' where the model practices on made-up data to get better on its own. They show that this process helps models perform better on tasks requiring long-term learning and adapting without needing human help.
continual learningknowledge distillationreinforcement learninglarge language modelsmemory consolidationcurriculum generationon-policy distillationself-improvementfew-shot generalizationsynthetic data
Authors
Ali Behrouz, Farnoosh Hashemi, Vahab Mirrokni
Abstract
The past few decades have witnessed significant advances in the design of machine learning algorithms, from early studies on task-specific shallow models to more general deep Large Language Models (LLMs). Despite showing promising results in tasks that require instant prediction or in-context learning, existing models lack the ability to continually learn and effectively transfer their temporal in-context knowledge to their long-term parameters. Inspired by human learning process, we introduce a ''Sleep'' paradigm that allows the models to continually learn, distill their short-term fragile memories into stable long-term knowledge with replay, and recursively improve themselves with ''Dreaming'' process. In more detail, sleep consists of two stages: (1) Memory Consolidation: an upward distillation process, called Knowledge Seeding, where the memories of a smaller-self are distilled into a larger network to provide more capacity while preserving the knowledge. As a proof of concept, we present a new Generalized Distillation process for {Knowledge Seeding} (i.e., the combination of on-policy distillation with Reinforcement Learning (RL)-based imitation learning); (2) Dreaming: a self-improvement phase, where the model uses RL to generate a curriculum of synthetic data to rehearse new knowledge and refine existing capabilities without human supervision. Our experiments on long-horizon, continual learning, knowledge incorporation, and few-shot generalization tasks support the importance of the sleep stage.