ControlBench

ControlBench是由伊利诺伊大学厄巴纳-香槟分校等机构创建的数据集，包含147个本科控制工程问题，旨在评估大型语言模型如GPT-4、Claude 3 Opus和Gemini 1.0 Ultra在解决控制工程问题上的能力。该数据集涵盖了控制设计的广度、深度和复杂性，包括系统动力学、PID/loopshaping设计及稳定性/鲁棒性分析等。数据集的创建过程经过精心设计，确保每个问题都经过人工验证，以LaTeX格式呈现，并提供详细的解决方案。ControlBench的应用领域主要集中在评估和提升人工智能在控制工程教育与研究中的集成，特别是在解决本科级别的控制问题方面。

ControlBench is a dataset developed by the University of Illinois Urbana-Champaign and other institutions. It includes 147 undergraduate control engineering problems, and is designed to assess the problem-solving capabilities of large language models (LLMs) such as GPT-4, Claude 3 Opus and Gemini 1.0 Ultra in the field of control engineering. This dataset covers the breadth, depth and complexity of control design, encompassing system dynamics, PID/loopshaping design, stability and robustness analysis, and other relevant topics. The creation of the dataset followed rigorous standards: every problem has been manually verified, all content is formatted in LaTeX, and detailed solutions are provided for each problem. The primary applications of ControlBench focus on evaluating and advancing the integration of artificial intelligence into control engineering education and research, particularly for addressing undergraduate-level control engineering problems.