Asymmetric Search

Embedding approach where documents/images are represented by multiple vectors (one per token/patch) rather than a single vector, enabling fine-grained semantic matching.

An information retrieval approach using dense vector representations (embeddings) to encode queries and documents. Unlike sparse methods like BM25, dense retrieval captures semantic meaning in continuous vector spaces, enabling neural search and forming the foundation of modern RAG systems.

A search approach that understands the meaning and intent of queries rather than just matching keywords. Using vector embeddings and similarity measures, semantic search finds conceptually relevant results even when exact terms don't match, enabling natural language queries and cross-lingual retrieval.

Comparison of dense vector retrieval (neural embeddings) and sparse retrieval (keyword-based) approaches including strengths, weaknesses, and when to use hybrid methods.

Advanced multi-vector retrieval model creating token-level embeddings with late interaction mechanism, featuring denoised supervision and improved memory efficiency over original ColBERT.

Agentic Search and Memory Retrieval technique by Supermemory using parallel reader agents and search agents that achieved ~99% accuracy on LongMemEval benchmark.