MultiOOP
收藏资源简介:
MultiOOP是一个多语言面向对象编程基准数据集,涵盖了六种流行语言(Python、PHP、C++、C#、Java、JavaScript),每种语言有267个任务。该数据集旨在评估大型语言模型(LLMs)在面向对象编程任务中的生成代码能力,并解决现有基准数据集在编程语言、任务粒度和测试用例数量方面的不平衡问题。数据集创建过程包括将现有的单语言面向对象编程基准转换为多语言基准,并设计了一个自动测试用例生成框架,以确保评估结果的可靠性。MultiOOP数据集的应用领域包括代码生成、大型语言模型、面向对象编程和多语言编程。
MultiOOP is a multilingual object-oriented programming benchmark dataset that covers six popular programming languages: Python, PHP, C++, C#, Java and JavaScript, with 267 tasks for each language. This dataset aims to evaluate the code generation capabilities of Large Language Models (LLMs) on object-oriented programming tasks, and addresses the imbalance issues of existing benchmark datasets in terms of programming languages, task granularity and the number of test cases. The dataset creation process includes converting existing single-language object-oriented programming benchmarks into multilingual benchmarks, and designing an automated test case generation framework to ensure the reliability of evaluation results. The application scenarios of the MultiOOP dataset include code generation, large language models, object-oriented programming and multilingual programming.




