STABLE: Gated Continual Learning for Large Language Models

William Hoy, Nurcin Celik

Large language models (LLMs) increasingly require mechanisms for continual adaptation without full retraining. However, sequential updates can lead to catastrophic forgetting, wher…