GO TO A SECTION

↑ ↓ to navigate · Enter to openExplore all sections →
FDEInterviews logo
← RAG & Agent System Design / 68
hardGleanHarveyHebbia

Retrieval metrics are green, the correct passage is provably in the context, and the answer is still wrong. Where do you look?

Recall@10 is 0.94 and climbing, and answer accuracy is going down. The two facts are consistent, and reconciling them is the whole question: every passage you add to raise recall also adds a distractor the generator has to survive.

Updated Sep 2026 · Grounded in real Forward Deployed Engineer interview loops and written to a senior-engineer editorial bar.

Recall@10 is 0.94 and climbing, and answer accuracy is going down. The two facts are consistent, and reconciling them is the whole question: every passage you add to raise recall also adds a distractor the generator has to survive.

20 answers per topic instead of 10, plus saved progress and bookmarks · no cardor unlock all 545 remaining answers · ₹2,000 / $25

Every answer, concept and course, every paper breakdown, all published video lessons, hands-on FDE Lab missions, Premium PDF guides and companion files, the full practice-test bank and work-sample downloads. Referral Premium excludes guide PDFs and their companion files.

6 months · One payment · No auto-renewal

Study alongside free video lessons.

Keep practising

Next in this trackAfter an index rebuild, a set of documents stopped appearing in any search result. Nothing errored and the count is right. Find them.PremiumNext in this trackYou added a cross-encoder reranker. p95 latency roughly doubled and answer quality did not move. Explain what happened and fix it.PremiumNext in this trackA maintenance assistant cites the right equipment family but the wrong manual revision. Fix the retrieval contract.Premium
FEDITOR'S NOTE

The question is engineered so that every retrieval instinct is a dead end, and the candidates who keep proposing retrieval fixes after being told the gold passage is in the context are failing the actual test. What earns the round is naming an experiment that isolates the generator, and the cleanest one is a sweep over k. Watching recall climb while end-to-end accuracy peaks and falls is the moment the room decides whether you have run one of these systems or read about them.

DISCUSSION · 0

No comments yet — be the first to share your approach.