Fine-tuning
Quick Answer
Fine-tuning is the process of further training a pre-trained model on a smaller, task-specific dataset so that it specialises in a particular domain, tone or set of behaviours. It sits between using a base model as-is and training a new model from scratch.
In Depth
What Fine-tuning really means
Modern fine-tuning approaches include full fine-tuning (updating all parameters), parameter-efficient methods such as LoRA (which update only a small set of adapters), and instruction tuning (teaching the model to follow a specific style of instructions).
Fine-tuning is powerful but not always the right answer. For most knowledge-grounding use cases, RAG is cheaper, faster to iterate on and easier to govern. Fine-tuning shines when you need to change a model's style, format or behaviour rather than its underlying facts.
Why It Matters
Business relevance for UK organisations
UK businesses should fine-tune when they need consistent output structure, a specific brand voice, or compliance with domain-specific language that a generic model cannot reliably produce.
Real-world example
How this shows up in practice
A Manchester retailer fine-tuned an LLM on its previous marketing copy and brand voice guidelines, achieving outputs that required 74% less editorial review.
Related Terms
Continue exploring
Large Language Model (LLM)
A Large Language Model (LLM) is a type of neural network trained on vast quantities of text to understand and generate human language. LLMs power chatbots, copilots, content generators and many modern AI features across consumer and business software.
BasicsTraining Data
Training data is the dataset used to teach a machine learning model the patterns it needs to perform its task. The quality, quantity, diversity and recency of training data directly determine how accurate and fair the resulting model will be.
TechnicalRAG (Retrieval Augmented Generation)
Retrieval Augmented Generation (RAG) is an architecture that combines a language model with an external knowledge source. Before generating an answer, the system retrieves relevant documents and feeds them to the model as context, dramatically reducing hallucinations and keeping answers current.
AdvancedTransfer Learning
Transfer learning is the practice of taking a model trained for one task and adapting it to a related task, saving time, data and compute. Fine-tuning a pre-trained language model on your own data is the most common contemporary example.