Dense Retrieval

Dense retrieval prioritizes semantic similarity rather than exact lexical overlap. Even if a query and a document use different words, they can still match if they are semantically close. This is especially important in modern QA, RAG, and semantic search systems. However, strong performance requires high-quality embeddings, the right index structure, and task-aligned training.