Karpukhin et al.

A local embedding model, a second smaller model that converts text into vectors, is essential for retrieval-augmented generation (RAG) setups, as it retrieves relevant document chunks for the LLM. Acc…