ISO接地NFR规范能否提升LLM代码生成质量？
原标题：Does ISO-Grounded NFR Specification Improve LLM Code Generation? A Comparison of Rich and Structured Interventions against a Natural-Language Baseline
AI 摘要
该研究比较了在LLM代码生成中，基于ISO/IEC 25010:2023的丰富自然语言和结构化JSON两种NFR规范方式，与RobuNFR风格的单行基线。研究发现，ISO接地增强能改善静态质量代理指标并降低对提示措辞的敏感性，但不能可靠提升功能正确性；当ISO内容固定时，散文和JSON格式差异不大。
正文节选
Does ISO-Grounded NFR Specification Improve LLM Code Generation? A Comparison of Rich and Structured Interventions against a Natural-Language BaselineConference: 20th Brazilian Symposium on Software Components, Architectures, and Reuse; September 8–11, 2026; São Paulo, SP, Brazil Abstract. In LLM-based code generation, Non-Functional Requirements (NFRs) are often specified as terse one-line phrases. We ask whether grounding those specifications in ISO/IEC 25010:2023, either as rich natural-langu