VEX19/VEX
收藏资源简介:
VEX v1.0 Stable 是一个用于自动短答案评分(ASAG)的真实世界数据集,收集自一门实时大学级数据库系统课程。数据集包含来自173名学生的30,682条学生回答,涉及239个问题标识符。每个实例包括自由文本的学生答案、序数成绩和相关元数据。数据集包含人工标注(黄金)和模型生成(白银)的标签,旨在支持超越点分数预测的评估,包括决策级评分行为。数据集支持标签级评估(如分数预测)和决策级评估(如通过/失败、排名、成绩映射),并允许结合人工标注和模型生成的标签进行训练。数据集主要使用德语(约95%),包含少量英语回答(约5%)。
VEX v1.0 Stable is a real-world dataset for automated short-answer grading (ASAG), collected from a live university-level database systems course. The dataset contains 30,682 student responses from 173 students across 239 question identifiers. Each instance consists of a free-text student answer, an ordinal grade, and associated metadata. The dataset includes both human-annotated (gold) and model-generated (silver) labels and is designed to support evaluation beyond pointwise score prediction, including decision-level grading behavior. The dataset supports label-level evaluation (e.g., score prediction) and decision-level evaluation (e.g., pass/fail, ranking, grade mapping), and allows training with a combination of human-annotated and model-generated labels. The dataset is predominantly in German (~95%), with a small proportion of English responses (~5%).




