Model Distillation

You need GPT-4-level quality but at 1/100th the cost. You use GPT-4 to generate training data, then fine-tune a small model to mimic it. What is this process? Flag format: CONGRESS{process_in_snake_case}