DevBench
收藏资源简介:
DevBench数据集由上海人工智能实验室开发,旨在评估大型语言模型在软件开发中的应用能力。数据集包含22个高质量的仓库,覆盖Python、C/C++、Java和JavaScript等语言,涉及机器学习、网络服务等多个应用领域。每个仓库都经过精心设计,以确保评估的全面性和准确性,包括软件设计文档、环境配置文件、实现代码以及验收和单元测试用例。DevBench不仅测试模型的代码生成能力,还评估其在复杂软件开发任务中的表现,如理解复杂的仓库结构、管理编译过程和掌握高级编程概念。
DevBench is developed by Shanghai AI Laboratory, aiming to evaluate the capabilities of large language models (LLMs) in software development. The dataset consists of 22 high-quality code repositories, covering programming languages including Python, C/C++, Java, JavaScript and others, spanning multiple application domains such as machine learning and web services. Each repository is meticulously designed to ensure the comprehensiveness and accuracy of the evaluation, including software design documents, environment configuration files, implementation code, as well as acceptance and unit test cases. DevBench not only tests the code generation capability of models, but also evaluates their performance in complex software development tasks, such as comprehending intricate repository structures, managing compilation processes, and mastering advanced programming concepts.




