Instruction Tuning

A fine-tuning approach where models are trained on (instruction, response) pairs — the technique that transformed base LLMs into useful assistants.