Think Before You Code: Dual Reasoning for the NLSafety-Utility Trade-Off in LLM Code Generation
结构化安全审计:在LLM生成代码的正确性与内容安全之间平衡
专题命中 测试时计算 :reasoning(title,summary_cn);chain-of-thought(abstract)
AI总结 本文提出NLSafety-Utility Duality Score和Dual Reasoning,通过结构化安全审计和任务导向的代码审查,提升LLM生成代码的安全性和正确性,实验显示其在多个模型上显著提升SUDS评分。
Comments 13 pages. Accepted at the 41st IEEE/ACM International Conference on Automated Software Engineering (ASE 2026). Artifact: https://doi.org/10.5281/zenodo.19245736