All terms
Training
Instruction Tuning
Fine-tuning a model on instruction-response pairs so it follows natural-language requests.
Definition
Instruction tuning is a supervised fine-tuning stage that trains a pretrained model on a wide variety of tasks written as natural-language instructions paired with responses. It teaches the model to follow user requests it has not seen before, rather than just predict the next token, turning a raw language model into a usable assistant. It is typically followed by preference learning and is a key reason chat models feel general-purpose.