RAG Interview Questions, issue 19, Jul 23, 2026
The Concatenation Trap
Senior ML Engineer interview at Google, and the interviewer asks:
“Your conversational RAG has five query rewrite strategies. Why not just run all five and concatenate the outputs?”
Don’t say: “More context is always better, so combining them improves recall.”
Why stacking query rewrite strategies silently averages away user intent, and how conditional routing boosts precision without the massive inference tax.
The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.