Instruction Tuning

A fine-tuning process where a language model is trained on instruction-response pairs to follow user instructions more reliably. Instruction-tuned models are far more useful as assistants than base models trained only on text prediction.