What is fine-tuning?
Fine-tuning is a second, smaller round of training that changes how a model behaves, its tone, formatting, or manners, rather than teaching it new facts about the world.
Once a model has gone through its main training, a company can run a further, much smaller training pass on a curated set of examples to change its behavior: making it better at following instructions, adjusting its tone, or teaching it to format answers a certain way. That process is fine-tuning. It reshapes how the model responds, it doesn't meaningfully expand what the model knows about current events or a specific business.
This is the detail people get wrong: fine-tuning is not how a company gets an AI assistant to answer accurately about today's product catalog or this week's pricing. Retraining a model every time a product page changes isn't practical. The usual way to give an AI assistant current, specific facts is to connect it to that information at the moment of the question, through a tool it can call or a retrieval step, not by baking the facts into the model itself.
Related
- Large language modelA large language model is a program trained on huge amounts of text to predict what word comes next, and that single skill is what lets it write, summarize, and answer questions.
- Training dataTraining data is the text a language model was shown while it was being built, and once training finishes it is compressed into the model's internal patterns rather than kept as documents it can look back up.
- Retrieval-augmented generation (RAG)Retrieval-augmented generation is the technique of retrieving relevant documents and giving them to a model as context before it answers.
- Tool useTool use is a model's ability to call an outside function mid-answer, such as a search, a calculator, or a code runner, and fold whatever it gets back into its response.
Want to know where you actually stand on this? Run a free visibility check or try the free tools.