arXiv cs.CL· Dong-Jae Koh, Huisu Kim, SeongHwan Yoon, Lasse M. Jantsch, Chun-Hee Lee, Seonghyeon Lee, Young-Kyoon Suh·· 13 小时前AI 评分31
LLM 能否完成数据库规范化？DNBENCH 基准与多智能体框架 MARS 发布
Can LLMs Normalize Databases? A Benchmark and Multi-Agent Framework for Schema Normalization
AI 导读
研究团队提出 Database Normalization Benchmark（DNBENCH），含 3,275 个样本，用三轴协议从 1NF 到 BCNF 评估 LLM 驱动的数据库规范化，覆盖 Single、Complex、Real World 三个难度层级，揭示依赖推断、模式分解与跨表约束重建的反复失败。
正文 · 原文
Abstract:Large Language Models (LLMs) are increasingly used to generate structured outputs, but their reliability remains unclear when those outputs must satisfy database-level constraints. We study this issue through database normalization, involving reasoning about functional dependencies, lossless join decompositions, and inter-table constraints. We introduce a Database Normalization Benchmark (DNBENCH), comprising 3,275 samples for evaluating LLM-driven database normalization from 1NF to BCNF. DNBENCH uses a three-axis protocol to measure semantic equivalence, structural accuracy, and logical validity. Across Single, Complex, and Real World levels, DNBENCH uncovers recurring failures in dependency inference, schema decomposition, and inter-table constraint reconstruction. We further propose Multi-Agent Reasoning for Schemas (MARS), which separates evidence extraction, violation diagnosis, and decomposition planning from schema generation and verification. MARS improves the DNB-SCORE by 82.0% over the single-prompt baseline. All artifacts will be released upon acceptance.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2609.11141 [cs.CL] |
| (or arXiv:2609.11141v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2609.11141 arXiv-issued DOI via DataCite |
Submission history
From: Dong Jae Koh [view email]
[v1]
Thu, 10 Sep 2026 06:34:35 UTC (11,388 KB)
[v2]
Wed, 21 Oct 2026 11:41:32 UTC (13,261 KB)
来源：arXiv cs.CL · arxiv.org