Generative AI & LLMs
Instruction tuning
Training a model on instruction-and-response examples so it follows user requests better.
Example
Reviewers provide instruction-and-answer examples so the model learns the expected response style.
Why it matters
It helps turn a general text-producing system into a more useful task assistant.
You'll hear it
“It now follows the request instead of just continuing the text.”
Can you control it?
Set by the team
The people building or running the AI choose this setup. An everyday user generally needs their help to change how this part works.
Common questions
- Does it learn only the facts in the example answers?
- No. It can also learn patterns in how requests should be handled and replies organized.
- Can poor example answers teach bad habits?
- Yes. Repeated mistakes or unsuitable styles can influence the behavior learned from them.
- Does better instruction-following guarantee safe behavior?
- No. A system can follow a harmful request well unless other protections work.