Maximise recall — stay under the 194ms latency SLA.

Tune the retriever. More top-k and deeper rerank lift recall but cost latency — and chunk size has a sweet spot (too small fragments context, too large dilutes it). Cranking everything blows the budget.

Practice — no rating pressure. Get as close to the optimum as you can; each play teaches the pattern.

Recall

71.8%

Latency

129 / 194ms

Chunk size (tokens)

Top-k retrieved

Rerank depth

The RAG Pipeline — practice the decision, scored against the optimum

What does the The RAG Pipeline puzzle teach?

You can't max top-k and rerank for free — a latency budget forces the tradeoff, and chunk size has a sweet spot.

Why do interviews ask about this?

Tuning retrieval (chunking, top-k, rerank) is the core RAG-systems interview — know which knob trades quality for cost.

How is it scored?

The engine computes the provably-optimal answer for the exact instance you played (dynamic programming, shortest-path, or exhaustive search — never a heuristic) and scores your attempt as a percentage of it. No login needed to see your score.

← Arena

Practice0:00

The RAG Pipeline · D4

Maximise recall — stay under the 194ms latency SLA.

Practice — no rating pressure. Get as close to the optimum as you can; each play teaches the pattern.

Recall

71.8%

Latency

129 / 194ms

Chunk size (tokens)

Top-k retrieved

Rerank depth