Instruction tuning
Also: supervised fine-tuning, SFT
Training a base model on examples of instructions paired with good responses, so it answers rather than continuing the text. It is the step that turns a text predictor into something that behaves like an assistant.
A definition is not an understanding. This is taught properly in Supervised fine-tuning, hands on, part of Alignment & RLHF.
Learn this properlyTerms from the same topic