Fine-tuning
Also called fine-tuned
Extra training that adjusts an existing model for a style, format or task, using a smaller set of examples.
After a model's main training, it can be trained again on a smaller, focused set of examples. That's fine-tuning. Companies use it to give a model a house style, a fixed output format, or a narrow job like sorting support tickets. The chatbots you use were themselves fine-tuned, with human feedback, to behave like helpful assistants.
Fine-tuning is mainly a way to change how a model responds. To make an AI answer from your own documents, the usual approach is different: give it the documents at question time (retrieval) rather than retrain it, because documents change, and retraining is slow and costly.
For example
A law firm might fine-tune a model to always answer in its memo format, and still give it this year's case files at question time so it answers from them.
Put it to use
Where you'll meet fine-tuning in ChatGPT, Gemini and Claude, how to use it there, and the mix-up most people make.
Daily Bytes Plus is in the Android app.