Why must you score retrieval and generation as two separate layers?
LLM & AILevel: IntermediateRAG (Retrieval-Augmented Generation)
Why must you score retrieval and generation as two separate layers?
Hints
- Identify the stage: pretraining, adaptation, generation or evaluation.
- Name the inputs, constraints and a case where the approach fails.
Answer
A RAG answer can fail in two independent places, and the final answer looks the same either way. Retrieval might fetch the wrong chunks. Or retrieval was fine but the generator misused good chunks. A single accuracy number blends the two, so you cannot tell which to fix, and you end up tuning chunking when the model was hallucinating, or swapping models when the search was broken.
Source and licence: Skillumen — Awesome AI Engineer Interview Questions · CC0-1.0