I was wondering if anyone is aware of any current/ongoing research on similarity search for vector dbs? In the RAG models I have been working on, most of the accuracy of the result hinges on the similarity search returning the right document, and less on the language model. Thanks!