STUDENTEVAL
收藏资源简介:
STUDENTEVAL是一个由东北大学、韦尔斯利学院和奥伯林学院的学生创建的数据集,包含1749个初学者编写的Python编程问题提示,用于评估大型代码语言模型。该数据集通过学生与Code LLM的交互式工作创建,旨在解决现有基准测试中专家提示与初学者需求不匹配的问题。STUDENTEVAL通过提供多样化的提示,帮助评估模型在处理不同技能水平提示时的表现,特别关注于教育领域中代码生成模型的应用和教学方法的改进。
STUDENTEVAL is a dataset developed by students from Northeastern University, Wellesley College, and Oberlin College. It contains 1,749 Python programming problem prompts authored by novice programmers, and is designed for evaluating large code language models. This dataset is built through interactive collaboration between students and Code LLMs, aiming to address the mismatch between expert-written prompts and the actual needs of beginner learners in existing benchmark datasets. By providing diverse prompts, STUDENTEVAL facilitates the evaluation of model performance when handling prompts of varying skill levels, with a particular focus on the application of code generation models in educational contexts and the improvement of teaching methodologies.

- 1StudentEval: A Benchmark of Student-Written Prompts for Large Language Models of Code东北大学、韦尔斯利学院、奥伯林学院 · 2023年



