Collected sources and patterns will appear here. Add from search, explore, or the patterns library.
Tuple[Qrels, RetrievalResults, List[int]] -> IRMetrics
Compare retrieved query results against ground-truth relevance annotations to compute rank-aware metrics (NDCG, MAP, MRR, Recall, Precision) at specific depth cutoffs.
Problem it solves
Assessing retriever performance requires standard, reproducible implementations of rank-based evaluation metrics.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.