Maximum Marginal Relevance
A ranking method that balances relevance to the query against redundancy with already selected results to encourage diversity.
Information-Retrieval Context
Maximum marginal relevance selects results by balancing similarity to the query with redundancy relative to items already chosen. In RAG pipelines it can help assemble context from distinct pieces of evidence instead of repeatedly selecting near-duplicate passages.
Ranking Boundary
MMR does not create missing candidates or improve first-stage recall. It changes selection order within the candidate set according to a relevance-versus-diversity trade-off.