Poster
in
Workshop: Table Representation Learning Workshop (TRL)

Towards Agentic Schema Refinement

Agapi Rissaki · Ilias Fountalis · Nikolaos Vasiloglou · Wolfgang Gatterbauer

Keywords: LLMs data exploration multi-agent simulation schema refinement data semantics

Project Page [ OpenReview]

Abstract

We introduce a multi-agent framework using LLMs to build a semantic layer over complex databases. By refining the database schema into smaller, more interpretable views, our approach simplifies data querying and exploration. We employ agentic programming, so that these views are discovered and refined through autonomous agent collaboration.Our approach paves the way for LLM-powered exploration of unwieldy databases.

Chat is not available.