Definition
A family of open-source embedding models developed by the Beijing Academy of Artificial Intelligence (BAAI) designed to map text into high-dimensional vector spaces; it is a top-tier choice for RAG pipelines, offering a trade-off between model size (Small, Base, Large) and retrieval accuracy.
BGE is the model that generates the vectors (encoder), not the database that stores them or the LLM that reads them.
"A high-precision library cataloging system that assigns every book a unique multi-dimensional coordinate based on its specific subject matter."
- MTEB (Massive Text Embedding Benchmark)(Evaluation Framework)
- Dense Retrieval(Core Mechanism)
- BGE Re-ranker(Complementary Component)
- Cosine Similarity(Mathematical Prerequisite)
Conceptual Overview
A family of open-source embedding models developed by the Beijing Academy of Artificial Intelligence (BAAI) designed to map text into high-dimensional vector spaces; it is a top-tier choice for RAG pipelines, offering a trade-off between model size (Small, Base, Large) and retrieval accuracy.
Disambiguation
BGE is the model that generates the vectors (encoder), not the database that stores them or the LLM that reads them.
Visual Analog
A high-precision library cataloging system that assigns every book a unique multi-dimensional coordinate based on its specific subject matter.