Elamur Research⌕
☰
home / papers / 2026 / Rose-SQL: Role-State Evolution Guided Structured Reasoning for Multi-Turn Text-to-SQL

Rose-SQL: Role-State Evolution Guided Structured Reasoning for Multi-Turn Text-to-SQL

Paper · 2026 · Le Zhou, Feng Yao, Fengcai Qiao et al.

Summary

Rose-SQL addresses multi-turn (conversational) Text-to-SQL, where a model must turn the current question plus dialogue history into SQL. Prior work either prompts large proprietary LLMs, which is fragile, or fine-tunes small models, which is expensive. The authors test whether small open reasoning models can handle the task through in-context learning alone. Their training-free framework introduces the Role-State, which tags schema elements with their function in the query and encodes the query's overall shape as a binary ten-slot vector. Gain Dependency Analysis scores how much each earlier turn helps the current one, using perplexity reduction weighted by recency. Evolutionary Trajectory Searching then retrieves similar dialogue sequences from training data and keeps only those whose Role-State transitions are structurally isomorphic. Finally, Augmented Hierarchical Reasoning makes the model plan the reasoning and the Role-State before writing SQL. On SParC and CoSQL with Qwen3 models, Rose-SQL beats ICL baselines at 4B and outperforms the fine-tuned Track-SQL at 8B and 14B. This suggests that structural guidance can replace task-specific training.

Key points

  • The Role-State vector flags which of ten structures appear in the target SQL. These are the selected columns, joins, filtering conditions, ordering, and grouping. They also include the three set operations (union, except, and intersect) and the "in" and "nin" (not-in) membership checks. Comparing these vectors across turns enables isomorphism checks.
  • The Gain Dependency Score is the dot product of two vectors. The first holds normalized perplexity drops, PPL(Q_i) vs. PPL(Q_i | Q_h). The second holds distance weights h/i. Before scoring, questions are rewritten with M-schema.
  • On SParC (Dev), Rose-SQL improves Qwen3 by 14.77–20.62% across the 4B–14B sizes without fine-tuning. It also shows gains on other reasoning backbones.
  • Earlier ICL methods transfer poorly to small models. ACT-SQL on Qwen3-8B loses 18.41% QM-EX and 19.71% IM-EX on SParC.
  • The authors list three limitations:
    • matching trajectories are scarce for rare dialogue patterns;
    • the lightweight Role-State has limited expressiveness;
    • inference costs more than direct prompting.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🏛️Chinese Academy of SciencesBeijing · China246 works · 24,952 citations🔬Institute of AutomationBeijing · China87 works · 14,883 citations🎓National University of Defense TechnologyChangsha · China20 works · 1,339 citations🎓Guangdong University of TechnologyGuangzhou · China5 works · 112 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksComputer Vision and Pattern Recognition · 3,486 works

🧭 Topics 3Research subjects and languages the work is about.Reasoning · 1,685 worksSemantic parsing & meaning representation · 193 worksCode & programming languages · 168 works

Cite

Cite as Le Zhou, Feng Yao, Fengcai Qiao, Bo Xu, Fangyuan Wang and Boyan Xu. 2026. Rose-SQL: Role-State Evolution Guided Structured Reasoning for Multi-Turn Text-to-SQL. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.2107

@inproceedings{zhou2026rose,
  title = {{Rose-SQL: Role-State Evolution Guided Structured Reasoning for Multi-Turn Text-to-SQL}},
  author = {Le Zhou, Feng Yao, Fengcai Qiao, Bo Xu, Fangyuan Wang and Boyan Xu},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.2107},
  url = {https://doi.org/10.18653/v1/2026.acl-long.2107}
}

References

30 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Yitao Cai and Xiaojun Wan. 2020. Igsql: Database schema interaction graph based neural model for context-dependent text-to-sql generation. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6903–6912. find
  2. Bingfeng Chen, Shaobin Shi, Yongqi Luo, Boyan Xu, Ruichu Cai, and Zhifeng Hao. 2025. Track-SQL: Enhancing generative language models with dual-extractive modules for schema and context tracking in multi-turn text-to-SQL. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of… find
  3. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Peiyi Wang, Qihao Zhu, Runxin Xu, Ruoyu Zhang, Shirong Ma, Xiao Bi, Xiaokang Zhang, Xingkai Yu, Yu Wu, Z. F. Wu, Zhibin Gou, Zhihong Shao, Zhuoshu Li, Ziyi Gao, Aixin Liu, and 175 others. 2025. Deepseek-r1 incentivizes reasoning in llms through rein… find
  4. Feng He, Zijun Chen, Xinnian Liang, Tingting Ma, Yunqi Qiu, Shuangzhi Wu, and Junchi Yan. 2025. Protoreasoning: Prototypes as the foundation for generalizable reasoning in llms. Preprint, arXiv:2506.15211. — [arXiv:2506.15211](https://arxiv.org/abs/2506.15211) find
  5. Tiansheng Huang, Sihao Hu, Fatih Ilhan, Selim Furkan Tekin, Zachary Yahn, Yichang Xu, and Ling Liu. 2025. Safety tax: Safety alignment makes your large reasoning models less reasonable. Preprint, arXiv:2503.00555. — [arXiv:2503.00555](https://arxiv.org/abs/2503.00555) find
  6. Binyuan Hui, Ruiying Geng, Qiyu Ren, Binhua Li, Yongbin Li, Jian Sun, Fei Huang, Luo Si, Pengfei Zhu, and Xiaodan Zhu. 2021. Dynamic hybrid relation network for cross-domain context-dependent semantic parsing. Preprint, arXiv:2101.01686. — [arXiv:2101.01686](https://arxiv.org/abs/2101.01686) find
  7. Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, and 1 others. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276.
  8. Aaron Jaech, Adam Kalai, Adam Lerer, Adam Richardson, Ahmed El-Kishky, Aiden Low, Alec Helyar, Aleksander Madry, Alex Beutel, Alex Carney, and 1 others. 2024. Openai o1 system card. arXiv preprint arXiv:2412.16720.
  9. Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. 2023. Efficient memory management for large language model serving with pagedattention. In Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems Principles.
  10. Jinyang Li, Binyuan Hui, Ge Qu, Jiaxi Yang, Binhua Li, Bowen Li, Bailin Wang, Bowen Qin, Ruiying Geng, Nan Huo, and 1 others. 2024. Can llm already serve as a database interface? a big bench for large-scale database grounded text-to-sqls. Advances in Neural Information Processing Systems, 36.
  11. Aixin Liu, Bei Feng, Bing Xue, Bingxuan Wang, Bochao Wu, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, and 1 others. 2024. Deepseek-v3 technical report. arXiv preprint arXiv:2412.19437. find
  12. Yifu Liu, Yin Zhu, Yingqi Gao, Zhiling Luo, Xiaoxia Li, Xiaorong Shi, Yuntao Hong, Jinyang Gao, Yu Li, Bolin Ding, and Jingren Zhou. 2025. Xiyan-sql: A novel multi-generator framework for text-to-sql. find
  13. Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, and 1 others. 2023. Self-refine: Iterative refinement with self-feedback. Advances in Neural Information Processing Systems, 36:46534–46594. find
  14. OpenAI. 2025. gpt-oss-120b & gpt-oss-20b model card. Preprint, arXiv:2508.10925.
  15. Aske Plaat, Annie Wong, Suzan Verberne, Joost Broekens, Niki van Stein, and Thomas Back. 2024. Reasoning with large language models, a survey. arXiv preprint arXiv:2407.11511.
  16. Jiexing Qi, Jingyao Tang, Ziwei He, Xiangpeng Wan, Yu Cheng, Chenghu Zhou, Xinbing Wang, Quanshi Zhang, and Zhouhan Lin. 2022. Rasat: Integrating relational structures into pretrained seq2seq model for text-to-sql. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Proces… find
  17. Yinggang Sun, Ziming Guo, Haining Yu, Chuanyi Liu, Xiang Li, Bingxuan Wang, Xiangzhan Yu, and Tiancheng Zhao. 2024. Qda-sql: Questions enhanced dialogue augmentation for multi-turn text-to-sql. Preprint, arXiv:2406.10593. — [arXiv:2406.10593](https://arxiv.org/abs/2406.10593) find
  18. Zhenheng Tang, Xiang Liu, Qian Wang, Peijie Dong, Bingsheng He, Xiaowen Chu, and Bo Li. 2025. The lottery llm hypothesis, rethinking what abilities should llm compression preserve? Preprint, arXiv:2502.17535. — [arXiv:2502.17535](https://arxiv.org/abs/2502.17535) find
  19. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, and 1 others. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824– 24837.
  20. Fengli Xu, Qianyue Hao, Zefang Zong, Jingwei Wang, Yunke Zhang, Jingyi Wang, Xiaochong Lan, Jiahui Gong, Tianjian Ouyang, Fanjin Meng, Chenyang Shao, Yuwei Yan, Qinglong Yang, Yiwen Song, Sijian Ren, Xinyuan Hu, Yu Li, Jie Feng, Chen Gao, and Yong Li. 2025. Towards large reasoning models: A survey o…
  21. Xiaojun Xu, Chang Liu, and Dawn Song. 2017. Sqlnet: Generating structured queries from natural language without reinforcement learning. arXiv preprint arXiv:1711.04436.
  22. Tao Yu, Rui Zhang, Heyang Er, Suyi Li, Eric Xue, Bo Pang, Xi Victoria Lin, Yi Chern Tan, Tianze Shi, Zihan Li, and 1 others. 2019a. Cosql: A conversational text-to-sql challenge towards cross-domain natural language interfaces to databases. In Proceedings of the 2019 Conference on Empirical Methods…
  23. Tao Yu, Rui Zhang, Kai Yang, Michihiro Yasunaga, Dongxu Wang, Zifan Li, James Ma, Irene Li, Qingning Yao, Shanelle Roman, and 1 others. 2018. Spider: A large-scale human-labeled dataset for complex and cross-domain semantic parsing and text-to-sql task. In Proceedings of the 2018 Conference on Empir…
  24. Tao Yu, Rui Zhang, Michihiro Yasunaga, Yi Chern Tan, Xi Victoria Lin, Suyi Li, Heyang Er, Irene Li, Bo Pang, Tao Chen, and 1 others. 2019b. Sparc: Cross-domain semantic parsing in context. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4511– 4523.
  25. Hanchong Zhang, Ruisheng Cao, Lu Chen, Hongshen Xu, and Kai Yu. 2023. Act-sql: In-context learning for text-to-sql with automatically-generated chain-of-thought. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 3501–3532. find
  26. Hanchong Zhang, Ruisheng Cao, Hongshen Xu, Lu Chen, and Kai Yu. 2024. Coe-sql: In-context learning for multi-turn text-to-sql with chain-of-editions. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Vol… find
  27. Yanzhao Zheng, Haibin Wang, Baohua Dong, Xingjun Wang, and Changshan Li. 2022. Hie-sql: History information enhanced network for context-dependent text-to-sql semantic parsing. In Findings of the Association for Computational Linguistics: ACL 2022, pages 2997–3007. find
  28. Victor Zhong, Caiming Xiong, and Richard Socher. 2017. Seq2sql: Generating structured queries from natural language using reinforcement learning. CoRR, abs/1709.00103.
  29. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, and 1 others. 2025. Qwen3 technical report. arXiv preprint arXiv:2505.09388.
  30. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. 2023. Tree of thoughts: Deliberate problem solving with large language models. Advances in neural information processing systems, 36:11809–11822. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhou-2026-rose-sql-role-state-evolution/