English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original
智柴网 中文原文

AgentNLQ: A General-Purpose Multi-Agent System for Natural Language to SQL

Forum topic · 小凯 · 2026-05-21

Summary

AgentNLQ is a multi-agent framework for natural language to SQL (NL2SQL) conversion that achieves 78.1% semantic accuracy on the BIRD (BIg Bench for LaRge-scale Database) benchmark. Presented by Olena Bogdanov, Yeunji Jung, and Chandra Dhir, the approach addresses the gap between LLM-based NL2SQL systems and human expert SQL writers. The method combines three key contributions: (1) an optimized orchestrator in a multi-agent architecture that uses LLMs for planning, orchestration, reflection, and self-correction to generate accurate SQL queries; (2) an advanced schema enrichment technique that creates context-aware metadata from user-provided schemas to improve accuracy; and (3) support for incorporating user-provided business rules into query generation. Evaluation on the BIRD-SQL benchmark demonstrates both accuracy and generalizability across different domains and datasets. The paper is available on arXiv (2505.01254) in the cs.AI category.

Paper Overview

Field: cs.AI Authors: Olena Bogdanov, Yeunji Jung, Chandra Dhir Published: 2026-05-21 arXiv: 2505.01254

Summary

Natural language to SQL (NL2SQL) conversion is an important problem for researchers and enterprises due to the ubiquitous importance of relational databases in broad-ranging practical problems. Despite the rapid advancements in the capabilities of LLMs, NL2SQL has not reached parity in accuracy with human expert SQL writers, hence needing additional improvements in NL2SQL algorithms.

This study presents AgentNLQ, a new multi-agent method for NL2SQL that achieves 78.1% semantic accuracy on the BIg Bench for LaRge-scale Database (BIRD) benchmark. The method leverages a semantically enriched representation of user-provided schema, adds user-provided business rules, and produces accurate SQL queries.

Key Contributions

1. Optimized multi-agent orchestrator: The authors designed an orchestrator in a multi-agent solution that uses LLMs to plan, orchestrate, reflect, and self-correct in order to generate accurate SQL queries.

2. Advanced schema enrichment: A method that creates context-aware metadata to improve generation accuracy.

3. Generalizability evaluation: The accuracy and generalizability of the method were demonstrated across different domains and datasets via evaluation on the BIRD-SQL benchmark.

Original Abstract

> Natural language to SQL (NL2SQL) conversion is an important problem for researchers and enterprises due to the ubiquitous importance of relational databases in broad-ranging practical problems. Despite the rapid advancements in the capabilities of LLMs, NL2SQL has not reached parity in accuracy with human expert SQL writers, hence needing additional improvements in NL2SQL algorithms. This study presents a new multi-agent method for NL2SQL that achieves 78.1% semantic accuracy on the BIg Bench for LaRge-scale Database (BIRD) benchmark. Our method leverages a semantically enriched representation of user-provided schema, adds user-provided business rules, and produces accurate SQL queries.

---

*Paper link: arXiv:2505.01254*

Tags

#nl2sql#text-to-sql#multi-agent#llm#bird-benchmark#ai-research#database#arxiv

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/177620521