AGRO-SQL: Agentic Group-Relative Optimization with High-Fidelity Data Synthesis
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866908737237680128 |
|---|---|
| author | Yang, Cehua Xiao, Dongyu Lin, Junming Song, Yuyang Yan, Hanxu Guo, Shawn Zhang, Wei Yang, Jian Tang, Mingjie Dai, Bryan |
| author_facet | Yang, Cehua Xiao, Dongyu Lin, Junming Song, Yuyang Yan, Hanxu Guo, Shawn Zhang, Wei Yang, Jian Tang, Mingjie Dai, Bryan |
| contents | The advancement of Text-to-SQL systems is currently hindered by the scarcity of high-quality training data and the limited reasoning capabilities of models in complex scenarios. In this paper, we propose a holistic framework that addresses these issues through a dual-centric approach. From a Data-Centric perspective, we construct an iterative data factory that synthesizes RL-ready data characterized by high correctness and precise semantic-logic alignment, ensured by strict verification. From a Model-Centric perspective, we introduce a novel Agentic Reinforcement Learning framework. This framework employs a Diversity-Aware Cold Start stage to initialize a robust policy, followed by Group Relative Policy Optimization (GRPO) to refine the agent's reasoning via environmental feedback. Extensive experiments on BIRD and Spider benchmarks demonstrate that our synergistic approach achieves state-of-the-art performance among single-model methods. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_23366 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | AGRO-SQL: Agentic Group-Relative Optimization with High-Fidelity Data Synthesis Yang, Cehua Xiao, Dongyu Lin, Junming Song, Yuyang Yan, Hanxu Guo, Shawn Zhang, Wei Yang, Jian Tang, Mingjie Dai, Bryan Databases Artificial Intelligence The advancement of Text-to-SQL systems is currently hindered by the scarcity of high-quality training data and the limited reasoning capabilities of models in complex scenarios. In this paper, we propose a holistic framework that addresses these issues through a dual-centric approach. From a Data-Centric perspective, we construct an iterative data factory that synthesizes RL-ready data characterized by high correctness and precise semantic-logic alignment, ensured by strict verification. From a Model-Centric perspective, we introduce a novel Agentic Reinforcement Learning framework. This framework employs a Diversity-Aware Cold Start stage to initialize a robust policy, followed by Group Relative Policy Optimization (GRPO) to refine the agent's reasoning via environmental feedback. Extensive experiments on BIRD and Spider benchmarks demonstrate that our synergistic approach achieves state-of-the-art performance among single-model methods. |
| title | AGRO-SQL: Agentic Group-Relative Optimization with High-Fidelity Data Synthesis |
| topic | Databases Artificial Intelligence |
| url | https://arxiv.org/abs/2512.23366 |