When is a large language model a good calibration agent? Physics-labeled prompting for budgeted engineering simulator calibration

查看原文 / View Original 编辑 / Edit

Computer Science > Machine Learning arXiv:2610.22379 (cs) [Submitted on 1 Oct 2026] Title: When is a large language model a good calibration agent? Physics-labeled prompting for budgeted engineering simulator calibration Authors: Ruo-Qian Wang.

计算机科学 > 机器学习 arXiv:2610.22379 (cs) [2026年10月1日提交] 标题：大型语言模型何时能成为优秀的校准代理？用于预算受限工程模拟器校准的物理标签提示法作者：Ruo-Qian Wang。

Abstract: Calibrating expensive engineering simulators requires tuning model parameters to match reference data, a process heavily constrained by computational costs. This study investigates whether Large Language Models (LLMs), conditioned on physical diagnostic descriptions of simulation errors, can propose informative parameter updates more efficiently than conventional derivative-free optimizers.

摘要：校准昂贵的工程模拟器需要调整模型参数以匹配参考数据，这一过程受到计算成本的严重制约。本研究探讨了以模拟误差的物理诊断描述为条件的大型语言模型（LLM），是否能比传统的无导数优化器更有效地提出信息量大的参数更新建议。

We test this on a cylinder-wake Large-Eddy Simulation and a damped-oscillator inverse problem. Physics-labeled prompts are evaluated against four text ablations (removed, anonymized, padded, misdirected) and standard optimizers (Nelder-Mead, Bayesian optimization, CMA-ES, and random search).

我们在圆柱绕流大涡模拟和阻尼振子反问题上进行了测试。物理标签提示法与四种文本消融（移除、匿名化、填充、误导）以及标准优化器（Nelder-Mead、贝叶斯优化、CMA-ES 和随机搜索）进行了对比评估。

Results demonstrate a strong early semantic advantage: the first physics-informed LLM proposal improves the initial design in 23 of 25 trials. After five evaluations, the median calibration error is 21.9 times lower than physics-free controls. However, this advantage is transient - disappearing by fifteen evaluations - and inherent LLM decoder variability yields a 9.8-fold performance spread.

结果显示出强大的早期语义优势：在25次试验中，首次物理信息LLM建议有23次改善了初始设计。在五次评估后，中位校准误差比无物理信息的对照组低21.9倍。然而，这种优势是短暂的——在十五次评估后消失——且LLM解码器固有的变异性导致了9.8倍的性能差异。

To address this, we introduce a hybrid workflow that uses the LLM as a warm-start, passing accumulated evaluations to a conventional optimizer. This two-stage numerical continuation policy reaches target accuracy in just 9 simulator runs, outperforming the LLM alone (12 runs), Nelder-Mead (17 runs), and Bayesian optimization (>60 runs).

为了解决这个问题，我们引入了一种混合工作流，将LLM用作热启动，并将累积的评估结果传递给传统优化器。这种两阶段数值延拓策略仅需9次模拟运行即可达到目标精度，优于单独使用LLM（12次运行）、Nelder-Mead（17次运行）和贝叶斯优化（>60次运行）。

Supported by sensitivity analysis validating the test cases, this hybrid physics-labeled LLM scheme proves highly effective as a calibration agent under strict evaluation budgets.

在验证测试案例的敏感性分析支持下，这种混合物理标签LLM方案被证明在严格的评估预算下作为校准代理非常有效。