An encoder that independently generates embeddings for queries and documents for scalable retrieval.
More about Bi-Encoder
Bi-Encoders independently encode queries and documents into vector embeddings, allowing for efficient similarity computation in vector databases. While less accurate than cross-encoders, bi-encoders are computationally efficient, making them ideal for large-scale retrieval tasks like dense retrieval.
Bi-encoders are a core component in systems like retrieval augmentation pipelines and semantic search, enabling fast and scalable retrieval.