Asymmetric model support: Optimizing semantic search for queries and documents

OpenSearch 3.5 introduces support for asymmetric embedding models, which enhance semantic search relevance by applying distinct encoding strategies to short queries and long, information-rich documents. Benchmarks demonstrate that these models significantly outperform traditional symmetric approaches, especially when there is a large disparity between query and passage lengths. The article provides a detailed, step-by-step guide for implementing this technology using the new semantic field type and remote connectors.