EasyDeepLearn

Should you fine-tune the LLM for RAG or improve retrieval first?

medium

Answer

  • Almost always improve retrieval first — it's cheaper, more auditable, and 80% of RAG problems are retrieval problems (wrong chunks, missing chunks, poor recall).
  • Fine-tune when: (1) the LLM refuses / hallucinates despite good retrieval; (2) the domain has heavy jargon the base LLM stumbles on; (3) you need a specific citation / format style.
  • Debug order: retrieval eval → prompt tuning → reranking → fine-tune generator only if the previous fail.
Check yourself — multiple choice
  • Always fine-tune first
  • Improve retrieval first (80% of RAG issues are retrieval); fine-tune LLM only if retrieval + prompt + rerank aren't enough
  • Only fine-tune
  • Random order

Debug order: retrieval → prompt → rerank → LLM fine-tune.

#rag#fine-tuning

Practise LLMs & GenAI

214 interview questions in this topic.

Related questions