Publikation
LLMs and Databases: A Synergistic Approach to Data Utilization
Fatma Ozcan; Yeounoh Chung; Yannis Chronis; Lyubomir Ganev; Yawen Wang; Carsten Binnig; Johannes Wehrstein; Gaurav Tarlok Kakkar; Sami Abu-el-haija
In: IEEE Data Engineering Bulletin, Vol. 49, No. 1, Pages 32-44, IEEE, 2025.
Zusammenfassung
Large language models (LLMs) are not merely changing computing; they are igniting a revo-
lution across industries, from healthcare to finance. Their prowess in tackling complex problems,
demonstrated by breakthroughs in chatbots, translation, and code generation, is undeniable. A
notable breakthrough in data management, driven by LLMs, is the advancement of natural language
to SQL translation. This technology has fueled substantial progress, making database interactions
more accessible and enabling the deployment of numerous real-world applications. Yet, even these
powerful models falter with latency-sensitive regression problems, a critical need in database perfor-
mance optimization. Inspired by the foundational principles of LLMs, we are developing pre-trained
cardinality estimation and foundation database models, bridging this gap and unlocking the next
generation of database optimization.
