AgentNLQ: A General-Purpose Agent for Natural Language to SQL
The article discusses a new multi-agent method for converting natural language to SQL, known as AgentNLQ. This method achieves a semantic accuracy of 78.1% on the BIRD benchmark, addressing the gap between machine and human SQL writing capabilities. Key innovations include an optimized orchestrator and an advanced schema enrichment method to enhance accuracy across various domains.
- ▪AgentNLQ is a general-purpose agent designed for natural language to SQL conversion.
- ▪The method achieves 78.1% semantic accuracy on the BIg Bench for LaRge-scale Database benchmark.
- ▪It incorporates user-provided schema and business rules to generate accurate SQL queries.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.19010 (cs) [Submitted on 18 May 2026] Title:AgentNLQ: A General-Purpose Agent for Natural Language to SQL Authors:Olena Bogdanov, Yeunji Jung, Chandra Dhir, Pareekshitreddy Gaddam, Saurabh Jain, Lakshmi Tumati, Vijay Parthasarathy, Anup Shirgaonkar View a PDF of the paper titled AgentNLQ: A General-Purpose Agent for Natural Language to SQL, by Olena Bogdanov and 7 other authors View PDF HTML (experimental) 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.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
