When Large Language Models Meet Vector Databases: A Survey

When Large Language Models Meet Vector Databases: A Survey

6 Feb 2024 | Zhi Jing*, Yongye Su*, Yikun Han*, Bo Yuan*, Haiyun Xu*, Chunjiang Liu*, Kehai Chen*, Min Zhang*
This survey explores the potential synergy between Large Language Models (LLMs) and Vector Databases (VecDBs), addressing challenges such as hallucinations, outdated knowledge, high commercial costs, and memory issues. VecDBs offer efficient storage, retrieval, and management of high-dimensional vector representations, enhancing LLM functionalities. The survey outlines the foundational principles of both LLMs and VecDBs, critically analyzes their integration, and discusses future developments. Key contributions include the use of VecDBs as external knowledge bases, semantic caches, and reliable memories for LLMs, which address issues like domain-specific knowledge, computational costs, and forgetting. The survey also highlights the importance of multimodal data handling and retrieval optimizations in RAG systems, emphasizing the need for hybrid search algorithms and multi-modal data fusion. Finally, it discusses challenges such as data preprocessing, multi-tenancy, cost-effective storage, and knowledge conflict resolution, providing directions for future research.This survey explores the potential synergy between Large Language Models (LLMs) and Vector Databases (VecDBs), addressing challenges such as hallucinations, outdated knowledge, high commercial costs, and memory issues. VecDBs offer efficient storage, retrieval, and management of high-dimensional vector representations, enhancing LLM functionalities. The survey outlines the foundational principles of both LLMs and VecDBs, critically analyzes their integration, and discusses future developments. Key contributions include the use of VecDBs as external knowledge bases, semantic caches, and reliable memories for LLMs, which address issues like domain-specific knowledge, computational costs, and forgetting. The survey also highlights the importance of multimodal data handling and retrieval optimizations in RAG systems, emphasizing the need for hybrid search algorithms and multi-modal data fusion. Finally, it discusses challenges such as data preprocessing, multi-tenancy, cost-effective storage, and knowledge conflict resolution, providing directions for future research.
Reach us at info@study.space
[slides and audio] When Large Language Models Meet Vector Databases%3A A Survey