arXiv AI September 12, 2026 By Ai Jian, Xiaoyun Zhang, Eryu Guo, Wanrou Du, Jingqing Ruan, Jiangbo Pei, Weipeng Zhang, Ke Zeng, Xunliang Cai

TRUST-SQL: Tool-Integrated Multi-Turn Reinforcement Learning for Text-to-SQL over Unknown Schemas

Read the original on arXiv AI →

TRUST‑SQL tackles Text‑to‑SQL parsing in environments where the full database schema is unknown, a common scenario in enterprise settings with many noisy tables. The method models the problem as a Partially Observable Markov Decision Process and uses a four‑phase protocol to verify only relevant metadata, guided by a Dual‑Track GRPO strategy that separates exploration from execution rewards. Experiments on five benchmarks show significant gains, with the 4B and 8B variants outperforming base models by 30.6% and 16.6% respectively, while matching or exceeding baselines that rely on pre‑loaded schemas.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

arXiv Computation and Language

Aug 24

Human-Level Text-to-SQL via Reinforcement Learning on Verified Data, Without Pipeline Engineering

arXiv:2603.20004v4 Announce Type: replace-cross Abstract: Translating natural language questions to SQL queries (Text-to-SQL) is a long-standing problem in database research. Recent efforts have focu...

By Yuxuan Zhu, Tengjun Jin, Yoojin Choi, Daniel Kang

llms reinforcement-learning fine-tuning benchmarks data-engineering

arXiv AI

Sep 17

DualSQL: Text-to-SQL with Multi-Agent Reinforcement Learning

DualSQL is a Text-to-SQL system that uses two agents sharing a single model backbone, enabling joint optimization via multi-agent reinforcement learning. The approach incorporates three database access tools for multi-step reasoning, rollout guardrails to stabilize training, and a new SQL correctness metric called robust execution match (REX). Trained on only 3,755 examples, DualSQL-4B reaches 68.0% execution accuracy on the BIRD dev set, while DualSQL-8B achieves 71.1%, surpassing prior state‑of‑the‑art single‑model solutions with 32B parameters.

By Shijie Chen, Yu Gan, Yeounoh Chung, Jiani Zhang, Quannan Li, Sravan Babu Bodapati, Cody J. Greer, Yu Su, Fatma Ozcan

agents reinforcement-learning benchmarks safety data-engineering

Hugging Face Trending Papers

Sep 16

DualSQL: Text-to-SQL with Multi-Agent Reinforcement Learning

State-of-the-art Text-to-SQL systems are typically multi-agent pipelines centered around two fundamental tasks: schema linking and SQL generation. However, existing work trains separate models for eac...

agents reinforcement-learning benchmarks safety data-engineering

arXiv AI

Jun 2

APEX-SQL: Talking to the data via Agentic Exploration for Text-to-SQL

arXiv:2602. 16720v2 Announce Type: replace-cross Abstract: Text-to-SQL systems powered by Large Language Models have excelled on academic benchmarks but struggle in complex enterprise environments.

By Bowen Cao, Weibin Liao, Yushi Sun, Dong Fang, Haitao Li, Wai Lam

llms agents efficiency benchmarks data-engineering

arXiv Machine Learning

Jun 15

Reward-SQL: Boosting Text-to-SQL via Stepwise Execution-Aware Reasoning and Process-Supervised Rewards

arXiv:2505. 04671v3 Announce Type: replace-cross Abstract: Recent advances in large language models (LLMs) trained with reinforcement learning (RL) have improved Text-to-SQL performance.

By Yuxin Zhang, Meihao Fan, Ju Fan, Mingyang Yi, Yuyu Luo, Guoliang Li, Bin Wu, Wenchao Zhou

llms reinforcement-learning safety data-engineering

arXiv AI

Sep 7

A Cost-Aware Agentic Architecture for NL-to-SQL over Nested Enterprise Schemas, with a New Benchmark

The paper introduces the DevRev NL2SQL benchmark, featuring 900 execution‑verified queries that test natural‑language‑to‑SQL systems on nested, graph‑like enterprise schemas, and proposes the Semantic Depth Score (SDS) as a rubric for analytical reasoning depth. It also presents a cost‑aware, single‑generation agentic architecture that includes schema selection, metadata retrieval, and error‑repair components tailored to these complex schemas. On the DevRev benchmark, the system achieves 91.7% answer correctness, outperforming the next‑best baseline by 54.6 percentage points, and remains competitive on the Spider 2.0 Snowflake dataset.

By Yoga Sri Varshan Varadharajan, Ajay Yadav, Ritesh Goru, Prateek Chaudhury, Constantine Caramanis, Prateek Jain, Divyateja Pasupuleti, Sunil Kumar Pandey

agents benchmarks data-engineering

One email a morning, machine-written

Related stories

Human-Level Text-to-SQL via Reinforcement Learning on Verified Data, Without Pipeline Engineering

DualSQL: Text-to-SQL with Multi-Agent Reinforcement Learning

DualSQL: Text-to-SQL with Multi-Agent Reinforcement Learning

APEX-SQL: Talking to the data via Agentic Exploration for Text-to-SQL

Reward-SQL: Boosting Text-to-SQL via Stepwise Execution-Aware Reasoning and Process-Supervised Rewards

A Cost-Aware Agentic Architecture for NL-to-SQL over Nested Enterprise Schemas, with a New Benchmark