An ML team is choosing between fine-tuning a foundation model and building a RAG solution to improve an LLM's responses about specialized medical terminology. Which consideration most strongly favors fine-tuning over RAG?

Question

Accepted Answer

B. The medical terminology follows consistent, stable patterns that are unlikely to change frequently.. Fine-tuning is best when target knowledge follows stable, consistent patterns (like terminology, style, or domain-specific reasoning) that can be baked into model weights. If the medical terminology is stable, fine-tuning can teach the model that vocabulary without requiring runtime retrieval overhead on every inference call.

An ML team is choosing between fine-tuning a foundation model and building a RAG solution to improve an LLM's responses about specialized medical terminology. Which consideration most strongly favors fine-tuning over RAG?

Related Questions

Discussion