Same Data, Different Schemas: Robustness of LLM-based Text-to-SQL Conference: Nineth International Workshop on Exploiting Artificial Intelligence Techniques for Data Management ; May 31-June 05, 2026; Bengaluru, India Nineth International Workshop on Exploiting Artificial Intelligence Techniques for Data Management (aiDM ’26), May 31-June 05, 2026, Bengaluru, India DOI: 10.1145/3814940.3815329 ISBN: 979-8-4007-2719-1/2026/05
Nitin Kanchinadam, Aditya Menachery, Amol Deshpande · 2026
Read Same Data, Different Schemas: Robustness of LLM-based Text-to-SQL Conference: Nineth International Workshop on Exploiting Artificial Intelligence Techniques for Data Management ; May 31-June 05, 2026; Bengaluru, India Nineth International Workshop on Exploiting Artificial Intelligence Techniques for Data Management (aiDM ’26), May 31-June 05, 2026, Bengaluru, India DOI: 10.1145/3814940.3815329 ISBN: 979-8-4007-2719-1/2026/05 by Nitin Kanchinadam, Aditya Menachery, Amol Deshpande (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-25838.txt.
Description
Kanchinadam, Menachery and Deshpande test the robustness of language model text-to-SQL when the same data sits under varied schemas. Their framework generates schemas from one entity relationship model, varying shredding choices while questions and data stay fixed. Four models answer the same questions across variants in two domains. Schema shifts model behavior, with very varied answers across equivalent schemas. Adding the original specification improves performance but does not fully remove the inconsistencies.
Catalog record
| CREATOR | Nitin Kanchinadam, Aditya Menachery, Amol Deshpande |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 56792 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.25838