Learning, Fast and Slow: Towards LLMs That Adapt Continually

TL;DR

This paper introduces a fast-slow learning framework for large language models that combines parameter stability with rapid task adaptation, improving continual learning and reducing catastrophic forgetting.

Contribution

It proposes Fast-Slow Training (FST), a novel method that uses fast context-based weights alongside slow parameter weights, enhancing sample efficiency and task adaptability.

Findings

FST is up to 3x more sample-efficient than RL-based training.

FST models retain closer proximity to the base model, reducing catastrophic forgetting.

FST enables better adaptation to new tasks in continual learning scenarios.

Abstract

Large language models (LLMs) are trained for downstream tasks by updating their parameters (e.g., via RL). However, updating parameters forces them to absorb task-specific information, which can result in catastrophic forgetting and loss of plasticity. In contrast, in-context learning with fixed LLM parameters can cheaply and rapidly adapt to task-specific requirements (e.g., prompt optimization), but cannot by itself typically match the performance gains available through updating LLM parameters. There is no good reason for restricting learning to being in-context or in-weights. Moreover, humans also likely learn at different time scales (e.g., System 1 vs 2). To this end, we introduce a fast-slow learning framework for LLMs, with model parameters as "slow" weights and optimized context as "fast" weights. These fast "weights" can learn from textual feedback to absorb the task-specific information, while allowing slow weights to stay closer to the base model and persist general reasoning behaviors. Fast-Slow Training (FST) is up to 3x more sample-efficient than only slow learning (RL) across reasoning tasks, while consistently reaching a higher performance asymptote. Moreover, FST-trained models remain closer to the base LLM (up to 70% less KL divergence), resulting in less catastrophic forgetting than RL-training. This reduced drift also preserves plasticity: after training on one task, FST trained models adapt more effectively to a subsequent task than parameter-only trained models. In continual learning scenarios, where task domains change on the fly, FST continues to acquire each new task while parameter-only RL stalls.