Fine-tuning
Beginner
Giving an already trained AI model extra practice on a narrower task.
Novice
Training an already trained model further on a smaller, targeted set of examples, such as instruction-and-answer pairs or hardware-coding problems.
Expert
Supervised fine-tuning (SFT) on instruction data turns a base model into an assistant. Too many passes over a narrow set overfit: the model grows more confident and less varied, which can raise pass@1 while lowering pass@10.