Less is More: Selective Reflection for Compatible and Efficient Knowledge Distillation in Large Language Models

Open in new window