Introduction

Natural Language to SQL (NL2SQL) has advanced rapidly with Large Language Models (LLMs), achieving strong performance on benchmarks such as Spider and BIRD. However, these gains have not translated into reliable deployment in real-world enterprise and high-stakes application domains, including finance, healthcare, logistics, and enterprise operations.

A key challenge is a persistent gap between benchmark assumptions and production reality. Existing benchmarks typically assume small, clean, and static schemas; fully specified user intent; limited join complexity; few-database settings; and accuracy-centric evaluation. These simplifications support rapid progress on benchmark metrics but fail to reflect real-world conditions.

In practice, enterprise and domain-specific systems are significantly more complex. They span thousands of evolving, poorly documented tables across distributed cloud platforms, with inconsistent business definitions, strict access controls, and heterogeneous data types. This makes both structure and semantics difficult to navigate, requiring user queries to be translated from high-level intent into grounded schemas, disambiguated metrics, and cost-aware execution plans across multiple systems. Despite strong benchmark performance, real-world deployment often fails due to incorrect joins, hallucinated tables or columns, weak grounding in domain metadata, limited scalability, and a lack of cost and system awareness that can lead to unsafe or overly expensive queries.

Meanwhile, NL2SQL is evolving from single-step query generation into agentic analytics systems that must retrieve and reason over schemas, plan multi-step workflows, interact through clarification, execute and validate SQL, and repair failures while operating under cost, latency, and governance constraints.

This workshop frames NL2SQL as a real-world AI systems challenge at the intersection of foundation models, databases, cloud infrastructure, and agentic reasoning. It positions NL2SQL as a core component of enterprise and high-stakes AI infrastructure and aims to shape a research agenda focused on trustworthy, scalable, cross-aware, and production-ready NL2SQL systems, bridging the gap between benchmark success and real-world impact.

Topics

Enterprise NL2SQL and Foundation Models

Large-schema reasoning and retrieval
Metadata-aware query generation
Semantic-layer & business metric reasoning

Agentic and Conversational Data Systems

Ambiguous, interactive, & querying
Agentic AI for data intelligence & query generation
Human-in-the-loop structured data reasoning

Cross-Systems and Cloud-Scale Analytics

Cross-table and cross-database reasoning
Execution-aware & cost-efficient query generation
Scalable architectures for NL-to-data interfaces

Safety, Reliability, and Governance

Reliability and failure analysis in NL2SQL systems
Cost-aware SQL generation
Security, privacy, & governance for NL data access
Deployment-driven evaluation metrics
Runtime optimization and monitoring

Advanced Reasoning and Applications

Geospatial & complex structured data reasoning
High-stakes application domains (e.g., finance, healthcare, enterprise BI)
Robustness under ambiguity, distribution shift, and noisy schemas

Important Dates

See Important Dates for all paper and abstract submission deadlines.

Author and Submission Guidelines

Please follow Author and Submission Guidelines to submit your work.

Committee

Program Chairs

Lindi Liao, George Mason University, USA (Chair)
Callie Liao, Stanford University, USA (Co-Chair)

Program Committee

Zhiqian Chen, Mississippi State University, USA
Kaiqun Fu, Texas Christian University, USA
Weiwei Gong, Oracle, USA
Katie He, Oracle, USA
Maryam Heidari, George Mason University, USA
Sungpack Hong, Oracle, USA
Fanchun Jin, Google, USA
Ge Jin, Purdue University, USA
Ahmad Mousavi, American University, USA
Sujith Ravi, Oracle, USA
Viktor Sanca, Oracle, USA
Chen Shen, Google, USA
Sanjeev Singh, Meta, USA
Gregory Joseph Stein, George Mason University, USA
Yingfan Xu, Oklahoma State University
Alex Wang, Yale University, USA
Zhenwei Zhang, University of Maryland, USA

External Industrial Reviewers

TBD

If you are interested in serving on the workshop program committee or paper reviewing, please apply online.