Instruction tuning
Also: Supervised Fine-Tuning · SFT · Post-Training
Post-training with examples of instructions and good answers so that a model follows requests reliably.
After pre-training, a model can continue text but does not automatically understand that it should answer a question. Instruction tuning shows it thousands of examples such as “Summarise this text” together with a model answer.
Together with RLHF and other methods, it forms what is called post-training, which turns a raw model into a useful assistant.
General information, not legal or professional advice.