🔎Top-k Retrieval
Also known as Top-k Results, Retrieval Depth, Top-k Chunks
Returning only the k best matches from a search. You choose the k.
Think of it like
It is like taking the top five books off a sorted pile. You pick how many to carry.
Example
Emma's bot used k = 3 and often missed facts. With k = 20 and a reranker that keeps the best 5, her correct answers rose.
Why it matters
Definition
Common mistake
Key takeaways
- 1You pick k. How many matches come back.
- 2Too few or too many. Missed facts or extra noise.
- 3Wide then narrow. Fetch many, rerank, keep few.