> it handles queries and documents differently to embed them in the same vector space I assume this might be behind the state of the art, then. As of a year ag

> it handles queries and documents differently to embed them in the same vector space I assume this might be behind the state of the art, then. As of a year ago OpenAI managed to get an embeddings model that can do both without any special flag or dual model to treat queries and documents differently.