Improving Demonstration Diversity by Human-Free Fusing for Text-to-SQL

Dingzirui Wang, Longxu Dou, Xuanliang Zhang, Qingfu Zhu, Wanxiang Che


Abstract
In-context learning with large language models (LLMs) is the current mainstream method for text-to-SQL. Previous studies have explored selecting relevant demonstrations from a human-labeled demonstration pool, but these methods lack diversity and incur high labeling costs. In this work, we address measuring and enhancing the diversity of the text-to-SQL demonstration pool. First, we introduce a diversity metric and present that the diversity of the existing labeling data can be further enhanced. Motivated by these findings, we propose Fused that iteratively fuses demonstrations to create a diverse demonstration pool based on human labeling or even from scratch with LLMs, reducing labeling costs. Fused achieves an average improvement of 2.1% based on existing labeling and 5.5% from scratch on several mainstream datasets, demonstrating its effectiveness.
Anthology ID:
2024.findings-emnlp.65
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2024
Month:
November
Year:
2024
Address:
Miami, Florida, USA
Editors:
Yaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
1193–1207
Language:
URL:
https://aclanthology.org/2024.findings-emnlp.65
DOI:
10.18653/v1/2024.findings-emnlp.65
Bibkey:
Cite (ACL):
Dingzirui Wang, Longxu Dou, Xuanliang Zhang, Qingfu Zhu, and Wanxiang Che. 2024. Improving Demonstration Diversity by Human-Free Fusing for Text-to-SQL. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 1193–1207, Miami, Florida, USA. Association for Computational Linguistics.
Cite (Informal):
Improving Demonstration Diversity by Human-Free Fusing for Text-to-SQL (Wang et al., Findings 2024)
Copy Citation:
PDF:
https://preview.aclanthology.org/landing_page/2024.findings-emnlp.65.pdf