Fine-tuning
Fine-tuning is the process of continuing the training of a pre-trained AI model (for example an LLM) on a dataset specific to a domain or task, in order to specialise its behaviour without starting from scratch.
Fine-tuning is the process of continuing the training of a pre-trained AI model (for example an LLM) on a dataset specific to a domain or task, in order to specialise its behaviour without starting from scratch.
Several variants exist: classic supervised fine-tuning, RLHF (Reinforcement Learning from Human Feedback), DPO (Direct Preference Optimisation) and parameter-efficient techniques such as LoRA and QLoRA, which only update a small part of the model.
In 2026, fine-tuning is still useful for niche cases (style, tone, domain vocabulary), but it is often superseded by RAG and good prompting, which are simpler to keep up to date.