OPT-Kraken：LLM优化建模幻觉的事后修复框架
原标题：Fix What's Wrong, Keep What's Right: Hallucination Repair for LLM-Based Optimization Modeling
AI 摘要
研究者提出 OPT-Kraken，一个无需训练的 LLM 优化建模幻觉修复框架，针对自然语言运筹学问题生成的数学模型与求解器代码之间的不一致进行事后修复。该框架提取问题需求与实例数据，通过问题-模型保真度、模型-代码一致性和数据接地检查定位错误，并按影响范围选择修复粒度，用契约约束允许的编辑。在已有优化建模基准上，它对多种自动建模框架和微调模型均带来正向收益，能修复零样本产物中的幻觉并保留原本正确的内容。
正文节选
Fix What’s Wrong, Keep What’s Right: Hallucination Repair for LLM-Based Optimization Modeling Abstract Large language models (LLMs) are increasingly being explored to automate operations research modeling, reduce modeling time and lower the expertise barrier to optimization based decision support. However, this process often introduces inconsistencies between generated artifacts and the original problem, which we refer to as optimization modeling hallucinations. Despite progress in general-purpo