Definition
Fine-tuning in the context of Large Language Models (LLMs) refers to the process of adapting a pre-trained model to perform specialised tasks or excel in specific domains by training it further on a smaller, task-oriented dataset.
This technique updates the model’s parameters to align its outputs with targeted requirements, bridging the gap between general-purpose capabilities and domain-specific expertise.
The process typically involves structured steps like dataset preparation, model initialisation, and iterative training to optimise performance for the target use case.
Key aspects of fine-tuning include:
Adaptation
Starting with a model trained on vast general data, then refining it using curated datasets (e.g., medical reports, legal documents) to handle niche terminology and context.
Efficiency
Leveraging existing model knowledge rather than training from scratch, reducing computational costs compared to full pre-training.
Methods
Ranges from full fine-tuning (updating all parameters) to parameter-efficient approaches like LoRA (Low-Rank Adaptation), which modifies only subsets of weights.
Applications
Enables customisation for tasks like medical report generation, legal analysis, and controlled conversational tones while maintaining compliance with data regulations.
