面向运筹学的不确定性感知模拟推理框架
原标题：Uncertainty-Aware Simulation-Based Inference for Operations Research with Large Language Models
AI 摘要
该研究提出了一种面向运筹学（OR）数学建模的不确定性感知推理框架，无需训练即可提升大语言模型（LLM）的建模可靠性。该方法通过短视前瞻模拟评估中间候选步骤，量化下游预测不确定性，并利用重要性重采样动态选择更可能生成连贯数学公式的候选。在NL4OPT、MAMO和IndustryOR等多个OR基准上的实验表明，该框架优于标准和低温基线，为可靠的OR公式生成提供了高效、免训练的范式。
正文节选
Computer Science > Machine Learning Title:Uncertainty-Aware Simulation-Based Inference for Operations Research with Large Language Models View PDF HTML (experimental) Abstract:Deploying large language models (LLMs) for operations research (OR) tasks remains challenging because correctness depends on a coherent modeling process, not merely a correct final answer. Standard autoregressive generation operates on a myopic policy, which sometimes fails to anticipate whether a partial formu