RAG Interview Questions, issue 6, Jul 10, 2026
The Reformulation Trap
Senior ML Engineer interview at Anthropic, and the interviewer asks:
“Your RAG system passes the raw user query straight to the retriever. Recall looks fine in your evals, but it collapses on real traffic, ambiguous and multi-hop questions especially. Before you touch the retriever, what does your query layer actually need to decide?”
Don’t say: “I’d add a query rewriter to clean up the input.”
Why a one-size-fits-all query cleaner quietly shreds multi-hop questions into noise, and the adaptive upstream classifier you actually need to diagnose query shape.
The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.