OR for AI That Does OR: Routing LLMs up the Escalator inside the OSCAR Framework

📄 arXiv: 2610.00912v1 📥 PDF

作者: Jinzhi Bu, Haixin Tang, Huanan Zhang

分类: cs.AI, math.OC

发布日期: 2026-10-01

💡 一句话要点

提出OSCAR框架以优化大语言模型的决策过程

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 优化建模 大语言模型 决策过程 成本效益 OSCAR框架 模拟器 编码器 审查器

📋 核心要点

  1. 现有方法在将业务描述转化为优化模型时，可能导致约束或目标的错误表示，从而影响求解结果的准确性。
  2. OSCAR框架通过离线模拟器验证候选模型，并在可行性之外继续搜索改进方案，优化了决策过程。
  3. 在五个基准问题上，OSCAR实现了95%至100%的准确率，且相较于传统方法显著降低了成本，提升了效率。

📝 摘要（中文）

大型语言模型能够将业务描述转化为优化模型，但生成的可执行代码可能会错误地表示约束或目标，导致求解器返回错误问题的最优解。为此，本文提出了OSCAR（优化建模模拟器、编码器和审查器），该框架通过离线模拟器对候选模型进行验证，并在可行性之外继续搜索改进方案。我们在简化的已知先验设置下，给出了成本排序升级的最优条件，并在五个基准问题上实现了95%至100%的准确率，显示出OSCAR在成本和性能上的优势。

🔬 方法详解

问题定义：本文旨在解决大型语言模型在优化建模中可能产生的约束和目标错误表示的问题，现有方法在多次使用优化建模时，可能导致求解器返回错误问题的最优解。

核心思路：OSCAR框架通过引入离线模拟器来验证候选模型的有效性，并在可行性之外继续寻找改进方案，从而提高决策的准确性和效率。

技术框架：OSCAR的整体架构包括三个主要模块：模拟器、编码器和审查器。模拟器用于验证决策示例，编码器负责生成优化模型，而审查器则确保生成的模型符合预设的操作规则。

关键创新：OSCAR的核心创新在于其能够在不依赖先验知识的情况下，通过成本排序的方式优化模型调用顺序，显著提升了决策效率和准确性。

关键设计：OSCAR设计了简单的操作规则和可调设置，以适应不同的预算和保密需求，同时建议企业维护标记的决策示例，以便在模型解释与实际规则冲突时进行调整。

🖼️ 关键图片

fig_0
fig_1

📊 实验亮点

在五个基准问题上，OSCAR实现了95%至100%的准确率，使用两个小型开源权重的大语言模型，且每次尝试的准确率分别为29%和48%。与Codex和Claude Code相比，OSCAR的平均令牌成本分别降低了3.1倍和5.8倍，显示出显著的成本效益。

🎯 应用场景

OSCAR框架具有广泛的应用潜力，尤其适用于需要频繁进行优化建模的企业。通过提供准确且低成本的优化模型，OSCAR能够帮助企业在决策过程中提高效率，降低成本，并适应不断变化的市场需求和技术进步。

📄 摘要（原文）

Large language models can translate business descriptions into optimization models, but executable code may misrepresent constraints or objectives. A solver can then return an optimal solution to the wrong problem. Even when the solution satisfies the intended operating rules, a better plan may exist. For organizations that repeatedly use optimization modeling, an LLM-based framework should produce accurate formulations at low cost and, ideally, run locally. We study how to verify improvements and allocate attempts across LLMs that differ in price and capability. We develop OSCAR (Optimization modeling by Simulator, Coder, And Reviewer), which uses an offline Simulator certified against labeled decision examples to compare candidates and continues searching beyond feasibility. We model the search for the next certified improvement as sequential decisions under unobserved difficulty: which LLMs to call and when to stop. In a simplified known-prior setting, we give conditions under which cost-ordered escalation is optimal. For general menus, we derive a prior-free competitive guarantee. On five benchmark problems, OSCAR achieves 95% to 100% accuracy at the reported settings using two small open-weight LLMs, each deployable locally on a single GPU. Their single-attempt accuracies average 29% and 48%. In five runs per problem, Codex and Claude Code incur average token costs 3.1 and 5.8 times OSCAR's, respectively. OSCAR supports open-weight models locally or in the cloud, depending on budget and confidentiality requirements. Firms should maintain labeled decision examples of feasible and infeasible decisions to clarify plain-language operating rules. OSCAR follows these labels when an LLM's interpretation conflicts with them. As LLM capabilities and prices change, OSCAR's simple operating rules and adjustable settings help firms adapt their model choices and benefit from these advances.