Data Preparation

How to Fine-Tune Models

A glimpse of Data Preparation

An instruction-following model learns from the responses you mark as desirable. A neat JSON file is only the container. The training example must also say who is speaking, what answer belongs to that input, and which tokens should contribute to the loss.

Consider a small routing task. The user writes “package late”; the desired assistant answer is delivery.

Continue learning

Pro unlocks guided learning, complete notes, practical walkthroughs, exercises, and every AI Research lesson.