fine-tuning

Syllable breakdown

Dictionary meaning

the process of taking a pre-trained machine learning model and continuing its training on a smaller, more specialised dataset in order to adapt its behaviour to a specific task, domain, or communication style; fine-tuning is far less computationally expensive than training a model from scratch and allows organisations to customise general-purpose LLMs for applications such as customer service, legal document review, or medical diagnosis support; parameter-efficient fine-tuning methods such as LoRA (Low-Rank Adaptation) have further reduced the computational cost of customisation, enabling fine-tuning on consumer-grade hardware