DSMR-SQL: Enhancing Text-to-SQL with Dual-Strategy SQL Generation and Multi-role SQL Selection

Yiming Huang; Jiyu Guo; Jichuan Zeng; Cuiyun Gao; Peiyi Han; Chuanyi Liu

doi:10.1007/978-981-95-2725-0_26

DSMR-SQL: Enhancing Text-to-SQL with Dual-Strategy SQL Generation and Multi-role SQL Selection

Yiming Huang
, Jiyu Guo
, Jichuan Zeng
, Cuiyun Gao^*
, Peiyi Han
, Chuanyi Liu

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Recent advancements in Large Language Models (LLMs) have markedly improved SQL generation. Nevertheless, existing approaches typically rely on single-model designs, limiting their capacity to effectively handle complex user queries. In addition, current methods often face difficulties in selecting the optimal SQL from multiple candidates. To mitigate these limitations, this study presents DSMR-SQL, a two-stage framework consisting of: (1) Dual-Strategy SQL Generation: DSMR-SQL aims to produce a broader spectrum of SQL queries by using multiple models with two strategies: Supervised Fine-Tuning and In-Context Learning; (2) Multi-Role SQL Selection: DSMR-SQL seeks to identify the SQL most aligning with user intent by introducing a collaborative framework involving three roles (i.e., Proposer, Critic, Summarizer). Extensive experiments on various datasets substantiate the efficacy of DSMR-SQL in enhancing SQL generation.

Original language	English
Title of host publication	Chinese Computational Linguistics - 24th China National Conference, CCL 2025, Proceedings
Editors	Maosong Sun, Peiyong Duan, Zhiyuan Liu, Ruifeng Xu, Weiwei Sun, Yubo Chen, Zhiliang Tian, Zhenghao Liu
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	435-468
Number of pages	34
ISBN (Print)	9789819527243
DOIs	https://doi.org/10.1007/978-981-95-2725-0_26
State	Published - 2026
Externally published	Yes
Event	24th China National Conference on Chinese Computational Linguistics, CCL 2025 - Jinan, China Duration: 11 Aug 2025 → 14 Aug 2025

Publication series

Name	Lecture Notes in Computer Science
Volume	16052 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	24th China National Conference on Chinese Computational Linguistics, CCL 2025
Country/Territory	China
City	Jinan
Period	11/08/25 → 14/08/25

Keywords

Dual-Strategy SQL Generation
Large Language Model
Multi-Role SQL Selection
Text to SQL

Access to Document

10.1007/978-981-95-2725-0_26

Cite this

Huang, Y., Guo, J., Zeng, J., Gao, C., Han, P., & Liu, C. (2026). DSMR-SQL: Enhancing Text-to-SQL with Dual-Strategy SQL Generation and Multi-role SQL Selection. In M. Sun, P. Duan, Z. Liu, R. Xu, W. Sun, Y. Chen, Z. Tian, & Z. Liu (Eds.), Chinese Computational Linguistics - 24th China National Conference, CCL 2025, Proceedings (pp. 435-468). (Lecture Notes in Computer Science; Vol. 16052 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-981-95-2725-0_26

Huang, Yiming ; Guo, Jiyu ; Zeng, Jichuan et al. / DSMR-SQL : Enhancing Text-to-SQL with Dual-Strategy SQL Generation and Multi-role SQL Selection. Chinese Computational Linguistics - 24th China National Conference, CCL 2025, Proceedings. editor / Maosong Sun ; Peiyong Duan ; Zhiyuan Liu ; Ruifeng Xu ; Weiwei Sun ; Yubo Chen ; Zhiliang Tian ; Zhenghao Liu. Springer Science and Business Media Deutschland GmbH, 2026. pp. 435-468 (Lecture Notes in Computer Science).

@inproceedings{0649e688794e47558d0705c6de57d3c0,

title = "DSMR-SQL: Enhancing Text-to-SQL with Dual-Strategy SQL Generation and Multi-role SQL Selection",

abstract = "Recent advancements in Large Language Models (LLMs) have markedly improved SQL generation. Nevertheless, existing approaches typically rely on single-model designs, limiting their capacity to effectively handle complex user queries. In addition, current methods often face difficulties in selecting the optimal SQL from multiple candidates. To mitigate these limitations, this study presents DSMR-SQL, a two-stage framework consisting of: (1) Dual-Strategy SQL Generation: DSMR-SQL aims to produce a broader spectrum of SQL queries by using multiple models with two strategies: Supervised Fine-Tuning and In-Context Learning; (2) Multi-Role SQL Selection: DSMR-SQL seeks to identify the SQL most aligning with user intent by introducing a collaborative framework involving three roles (i.e., Proposer, Critic, Summarizer). Extensive experiments on various datasets substantiate the efficacy of DSMR-SQL in enhancing SQL generation.",

keywords = "Dual-Strategy SQL Generation, Large Language Model, Multi-Role SQL Selection, Text to SQL",

author = "Yiming Huang and Jiyu Guo and Jichuan Zeng and Cuiyun Gao and Peiyi Han and Chuanyi Liu",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.; 24th China National Conference on Chinese Computational Linguistics, CCL 2025 ; Conference date: 11-08-2025 Through 14-08-2025",

year = "2026",

doi = "10.1007/978-981-95-2725-0\_26",

language = "英语",

isbn = "9789819527243",

series = "Lecture Notes in Computer Science",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "435--468",

editor = "Maosong Sun and Peiyong Duan and Zhiyuan Liu and Ruifeng Xu and Weiwei Sun and Yubo Chen and Zhiliang Tian and Zhenghao Liu",

booktitle = "Chinese Computational Linguistics - 24th China National Conference, CCL 2025, Proceedings",

address = "德国",

}

Huang, Y, Guo, J, Zeng, J, Gao, C, Han, P & Liu, C 2026, DSMR-SQL: Enhancing Text-to-SQL with Dual-Strategy SQL Generation and Multi-role SQL Selection. in M Sun, P Duan, Z Liu, R Xu, W Sun, Y Chen, Z Tian & Z Liu (eds), Chinese Computational Linguistics - 24th China National Conference, CCL 2025, Proceedings. Lecture Notes in Computer Science, vol. 16052 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 435-468, 24th China National Conference on Chinese Computational Linguistics, CCL 2025, Jinan, China, 11/08/25. https://doi.org/10.1007/978-981-95-2725-0_26

DSMR-SQL: Enhancing Text-to-SQL with Dual-Strategy SQL Generation and Multi-role SQL Selection. / Huang, Yiming; Guo, Jiyu; Zeng, Jichuan et al.
Chinese Computational Linguistics - 24th China National Conference, CCL 2025, Proceedings. ed. / Maosong Sun; Peiyong Duan; Zhiyuan Liu; Ruifeng Xu; Weiwei Sun; Yubo Chen; Zhiliang Tian; Zhenghao Liu. Springer Science and Business Media Deutschland GmbH, 2026. p. 435-468 (Lecture Notes in Computer Science; Vol. 16052 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - DSMR-SQL

T2 - 24th China National Conference on Chinese Computational Linguistics, CCL 2025

AU - Huang, Yiming

AU - Guo, Jiyu

AU - Zeng, Jichuan

AU - Gao, Cuiyun

AU - Han, Peiyi

AU - Liu, Chuanyi

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.

PY - 2026

Y1 - 2026

N2 - Recent advancements in Large Language Models (LLMs) have markedly improved SQL generation. Nevertheless, existing approaches typically rely on single-model designs, limiting their capacity to effectively handle complex user queries. In addition, current methods often face difficulties in selecting the optimal SQL from multiple candidates. To mitigate these limitations, this study presents DSMR-SQL, a two-stage framework consisting of: (1) Dual-Strategy SQL Generation: DSMR-SQL aims to produce a broader spectrum of SQL queries by using multiple models with two strategies: Supervised Fine-Tuning and In-Context Learning; (2) Multi-Role SQL Selection: DSMR-SQL seeks to identify the SQL most aligning with user intent by introducing a collaborative framework involving three roles (i.e., Proposer, Critic, Summarizer). Extensive experiments on various datasets substantiate the efficacy of DSMR-SQL in enhancing SQL generation.

AB - Recent advancements in Large Language Models (LLMs) have markedly improved SQL generation. Nevertheless, existing approaches typically rely on single-model designs, limiting their capacity to effectively handle complex user queries. In addition, current methods often face difficulties in selecting the optimal SQL from multiple candidates. To mitigate these limitations, this study presents DSMR-SQL, a two-stage framework consisting of: (1) Dual-Strategy SQL Generation: DSMR-SQL aims to produce a broader spectrum of SQL queries by using multiple models with two strategies: Supervised Fine-Tuning and In-Context Learning; (2) Multi-Role SQL Selection: DSMR-SQL seeks to identify the SQL most aligning with user intent by introducing a collaborative framework involving three roles (i.e., Proposer, Critic, Summarizer). Extensive experiments on various datasets substantiate the efficacy of DSMR-SQL in enhancing SQL generation.

KW - Dual-Strategy SQL Generation

KW - Large Language Model

KW - Multi-Role SQL Selection

KW - Text to SQL

UR - https://www.scopus.com/pages/publications/105022008571

U2 - 10.1007/978-981-95-2725-0_26

DO - 10.1007/978-981-95-2725-0_26

M3 - 会议稿件

AN - SCOPUS:105022008571

SN - 9789819527243

T3 - Lecture Notes in Computer Science

SP - 435

EP - 468

BT - Chinese Computational Linguistics - 24th China National Conference, CCL 2025, Proceedings

A2 - Sun, Maosong

A2 - Duan, Peiyong

A2 - Liu, Zhiyuan

A2 - Xu, Ruifeng

A2 - Sun, Weiwei

A2 - Chen, Yubo

A2 - Tian, Zhiliang

A2 - Liu, Zhenghao

PB - Springer Science and Business Media Deutschland GmbH

Y2 - 11 August 2025 through 14 August 2025

ER -

Huang Y, Guo J, Zeng J, Gao C, Han P, Liu C. DSMR-SQL: Enhancing Text-to-SQL with Dual-Strategy SQL Generation and Multi-role SQL Selection. In Sun M, Duan P, Liu Z, Xu R, Sun W, Chen Y, Tian Z, Liu Z, editors, Chinese Computational Linguistics - 24th China National Conference, CCL 2025, Proceedings. Springer Science and Business Media Deutschland GmbH. 2026. p. 435-468. (Lecture Notes in Computer Science). doi: 10.1007/978-981-95-2725-0_26