← Quit0 / 0 correct
LLMs & AIHard1 of 10

Why can 'fine-tuning on our 200 FAQ answers' make a model worse at answering new questions about the product?