Why LLMs Forget Earlier Tasks in Long Prompts

Original: Understanding Generalization and Forgetting in In-Context Continual Learning

Source: arxiv.org ↗