Dense Retrieval overview

Dense Retrieval as a cross-cutting topic — similarity search using dense vector representations of queries and documents. Covers approximate nearest neighbor algorithms (HNSW, IVF, DiskANN), embedding space geometry, query-document asymmetry (bi-encoder architectures), distance metrics (cosine, dot product, L2), and the fundamental trade-off between index build time, memory usage, and retrieval latency. Dense retrieval excels at semantic matching but struggles with exact keyword lookups and rare terms.