Definition
The process of updating a pre-trained model's parameters on a domain-specific dataset to refine its behavior, style, or task-specific performance. In RAG pipelines, it is typically used to optimize the 'Reader' model for better context adherence or to adapt 'Embedding' models for improved retrieval accuracy in niche domains.
Changes the model's internal weights (knowledge), whereas RAG provides external context via the prompt (memory).
"A general practitioner doctor undergoing a three-year specialized residency to become a cardiologist."
Conceptual Overview
The process of updating a pre-trained model's parameters on a domain-specific dataset to refine its behavior, style, or task-specific performance. In RAG pipelines, it is typically used to optimize the 'Reader' model for better context adherence or to adapt 'Embedding' models for improved retrieval accuracy in niche domains.
Disambiguation
Changes the model's internal weights (knowledge), whereas RAG provides external context via the prompt (memory).
Visual Analog
A general practitioner doctor undergoing a three-year specialized residency to become a cardiologist.