Mean Reciprocal Rank, or MRR, measures how high the first relevant result appears in the retrieved ranking. For each query, the reciprocal rank is 1 divided by the position of the first relevant result. If the first relevant result is ranked second, the reciprocal rank is 0.5.
MRR is useful when the first good result matters most. It helps evaluate whether users or downstream models see useful context early, not just whether relevant context appears somewhere in the list.