The Faiss Library
By Matthijs Douze, Alexandr Guzhva, Chengqi Deng, Jeff Johnson, Gergely Szilvasy, Pierre-Emmanuel Mazaré, Maria Lomeli, Lucas Hosseini, Herv'e J'egou. IEEE Transactions on Big Data, 2024
768 citations recorded by the source corpus.
Vector databases typically manage large collections of embedding vectors. As AI applications are growing rapidly, the number of embeddings that need to be stored and indexed is increasing. The Faiss library is dedicated to vector similarity search, a core functionality of vector databases. Faiss is a toolkit of indexing methods and related primitives used to search, cluster, compress and transform vectors. This paper describes the trade-offs in vector search and the design principles of Faiss in terms of structure, approach to optimization and interfacing. We benchmark key features of the library and discuss a few selected use cases to highlight its broad applicability.
Subfields: Classics
View Paper: https://www.semanticscholar.org/paper/d4999c6f04c357414f66066f0fa06be14faa8efc
Citations in the corpus
- Evaluating Vector-Based Search and Question Answering Approaches for an Information System (2026, unknown): “Since Mikolov [5] showed that semantic proximity can be learned from co-occurrence and subsequently expressed through metrics, and Reimers [6] expanded this approach to sentences and whole documents, it seems of interest to apply this approach to document retrieval tasks.”
- On the Challenges and Opportunities of Learned Sparse Retrieval for Code (2026, unknown): “For ℎ𝑛𝑠𝑤 , we use faiss [6] and parameters M = 32, efConstruction = 40 and vary efSearch .”