Retrieval Recall@Kmedium

Retrieval Recall@K

medium RAG & Retrieval 15 min

NumPy allowed · plain Python earns 10 bonus points

A query retrieves context

Embedding similarity ranks the documents before generation.

Open lesson →

01query

02embed

03retrieve

04context

What is RAG?→ embed →top context

retrieval.md44%

rag-systems.md33%

agents.md24%

Cosine-like similarity scores ranked before context assembly

Implement Recall@K for the retrieval stage specifically, evaluating whether the retriever found the known-relevant documents at all (a prerequisite for the generator to ever use them).

Task

Implement retrieval_recall_at_k(retrieved_ids, relevant_ids, k). Return |set(retrieved_ids[:k]) intersect relevant_ids| / len(relevant_ids). Return 0.0 if relevant_ids is empty.

Function Signature

retrieval_recall_at_k(retrieved_ids: list[str], relevant_ids: set, k: int) -> float

Examples

Example 1: full recall
Input: retrieved_ids = ['a', 'b', 'c']
relevant_ids = ['a', 'b']
k = 3
Output: 1

Example 2: partial recall
Input: retrieved_ids = ['a', 'x', 'y']
relevant_ids = ['a', 'b']
k = 3
Output: 0.5

Example 3: k truncates before relevant item found
Input: retrieved_ids = ['x', 'y', 'a']
relevant_ids = ['a']
k = 2
Output: 0

Constraints

k >= 1

Python3Saved ✓

def retrieval_recall_at_k(retrieved_ids, relevant_ids, k):

# Your implementation here

pass

Case 1: full recall

Input: retrieved_ids = ['a', 'b', 'c']
relevant_ids = ['a', 'b']
k = 3

Expected: 1

Case 2: partial recall

Input: retrieved_ids = ['a', 'x', 'y']
relevant_ids = ['a', 'b']
k = 3

Expected: 0.5

Case 3: k truncates before relevant item found

Input: retrieved_ids = ['x', 'y', 'a']
relevant_ids = ['a']
k = 2

Expected: 0