Jasper and Stella: Distilling State-of-the-Art Embedding Models

Jasper and Stella are state-of-the-art embedding models developed to address challenges in dense retrieval for applications like FAQ systems and Retrieval-Augmented Generation (RAG). The Jasper model, built upon the Stella embedding model, achieved the No. 3 position on the Massive Text Embedding Benchmark (MTEB) leaderboard as of December 24, 2024, with an average score of 71.54 across 56 datasets