Each retriever produces scores or ranked candidates using different evidence, such as exact terms and learned meaning. A fusion method normalizes and merges those rankings before optional filtering or reranking.
Hybrid search combines lexical matching with semantic or vector retrieval.
Each retriever produces scores or ranked candidates using different evidence, such as exact terms and learned meaning. A fusion method normalizes and merges those rankings before optional filtering or reranking.