ScenEval
收藏资源简介:
ScenEval是由牛津布鲁克斯大学工程、计算与数学学院创建的一个大规模Java编程任务数据集,包含12864个任务,涵盖多种编程场景和来源。数据集通过手动和自动方式从教科书、在线学习网站和Stack Overflow论坛中提取,每个任务都附有JSON格式的场景信息作为元数据。ScenEval旨在通过场景化测试评估大型语言模型在代码生成方面的能力,特别关注于解决代码生成的复杂性和性能问题。
ScenEval is a large-scale Java programming task dataset created by the School of Engineering, Computing and Mathematics at Oxford Brookes University. It contains 12,864 tasks covering diverse programming scenarios and sources. The dataset is extracted from textbooks, online learning platforms, and the Stack Overflow forum through both manual and automated methods, with each task accompanied by scenario information in JSON format as metadata. ScenEval aims to evaluate the code generation capabilities of large language models (LLMs) via scenario-based testing, with a particular focus on addressing the complexity and performance issues in code generation.




