Efficient LLM Context Distillation

This paper specifically investigates context distillation a method that extends the utility of task-specific examples by internalizing them, thus augmenting the example set accessible for model inference.