Which technique is used to supply domain-specific knowledge to an LLM?

Question

Options

• ADomain-adaptation training
• BPrompt template expansion
• CRetrieval-Augmented Generation
• DFine-tuning the model on general data

• A
  84% (32)
• B
  3% (1)
• C
  8% (3)
• D
  5% (2)

Explanation

Domain-adaptation training (A) is the technique of continuing to train a pre-trained LLM on domain-specific corpora (medical texts, legal documents, code, etc.), embedding that knowledge directly into the model's weights so it becomes intrinsically "aware" of the domain.

Why the distractors are wrong:

• B. Prompt template expansion structures how inputs are formatted but adds no new knowledge to the model - it just shapes what the model already knows.
• C. Retrieval-Augmented Generation (RAG) supplies domain knowledge at inference time via external retrieval, rather than training it into the model itself - it's a runtime augmentation, not a training technique.
• D. Fine-tuning on general data trains on broad, non-domain content, which improves general capability but actively avoids supplying domain-specific knowledge.

Memory tip: Think of domain-adaptation training as teaching the model a specialty - like sending a doctor to medical school. RAG, by contrast, is like giving that doctor a reference book to consult during a consultation; the book isn't "in" the doctor, it's looked up on demand. The exam question asks what gets the knowledge into the LLM, which points to training (A).

No community discussion yet for this question.