SECAI Core · phase 2 of 15
Fine tuning
Training a neural network model with a smaller dataset tailored for a specific task
The Explain card
- Plain English
- Fine tuning takes an existing trained model and continues training it on a smaller dataset built for a specific task, adapting its behaviour without starting over.
- Example
- A company fine-tunes a general LLM on its own support tickets so it answers in house style and knows the product names.
- Why it matters
- Fine tuning can erode the safety training the base model had, and the fine-tuning dataset is a poisoning target. A few hundred crafted examples can plant a hidden trigger behaviour.
- Hook
- Teaching an old model new tricks, including the tricks you did not intend.
Where it sits in the deck
Phase 2: How Models Are Built: Architectures and Learning Mechanics
With the paradigms named, zoom in on the architectural building blocks that turn data into trained artifacts.
