LLM + Vector Data @ ICDE 2026

Large Language Models (LLMs) have transformed natural language processing with versatile generative capabilities that streamline AI workflows–from data cleaning and augmentation to function calling and inference. Yet current LLM pipelines still suffer from hallucinations, limited complex reasoning, and outdated or domain-specific knowledge gaps. This workshop explores two complementary solutions. First, vector databases enable retrieval-augmented generation by grounding outputs in semantically indexed external data, reducing factual errors. Second, AI agent frameworks orchestrate multi-step task decomposition, specialized tool invocation, and uncertainty monitoring to prevent incoherent responses. We aim to explore unified prototypes integrating these components and call for evaluating their performance on benchmark query-processing and data-analysis tasks.

For information about the 1st edition of the LLM+Vector Data workshop (ICDE 2025), please visit the previous website.