Information retrieval evaluation is the systematic process of measuring and assessing how effectively an information retrieval system satisfies user queries by finding and ranking relevant data from a collection. This evaluation typically relies on standardized test collections comprising sample queries, document corpora, and relevance assessments to quantify system performance. Key evaluation criteria include retrieval accuracy, ranking quality, and coverage, which are traditionally quantified using metrics such as precision, recall, mean average precision, and normalized discounted cumulative gain. Beyond manual human relevance labeling, modern information retrieval evaluation encompasses automated assessment techniques, user satisfaction studies, interactive online experimentation, and performance benchmarks for complex search architectures such as retrieval-augmented generation pipelines.
