Fine-tuning
In one sentence
Training an already pre-trained foundation model further on a company's own dataset so it adapts to a specific task or writing style.
Fine-tuning is well-suited for repetitive tasks with fixed formats and styles, such as extracting invoice data into a standardized structure. It requires hundreds or thousands of high-quality sample pairs, as well as ongoing technical maintenance.
If the goal is to keep the AI updated with your latest company information, a knowledge base is a better fit: when information changes, simply update the knowledge base without retraining the model.
Example
A company fine-tunes a model using 2,000 curated customer service Q&As so its tone and formatting match those of human agents.
Related terms
- Pre-training
- The stage where a large language model learns language and general knowledge from massive volumes of public text, determining its baseline capabilities and knowledge cutoff date.
- Knowledge Base
- A repository of company documents and institutional knowledge that AI can search and reference, enabling it to answer questions and produce content grounded in internal facts.
- RAG (Retrieval-Augmented Generation)
- A method where AI retrieves relevant information from a knowledge base first, then generates an answer based on those materials.
