Recall@k is the share of relevant passages found among the top k results; precision@k is the share of relevant ones among those k results. These metrics evaluate search independently of how the answer is written.
Example
Out of 2 expected passages, 1 is in the top 5: recall@5 = 0.5.