CharlyR/varbench-metric-evaluation

该数据集包含了代码和相应的指令、结果描述等信息，旨在评估代码修补和生成任务的性能。数据集中的每个示例都包含了原始代码、修补后的代码、预测结果以及多种评估指标。数据集分为raw和treated两种配置，其中treated配置额外提供了人工评分。训练集包含了大量的代码示例，用于训练和评估相关模型。

The dataset consists of code and corresponding instructions, result descriptions, etc., aiming to evaluate the performance of code repair and generation tasks. Each example in the dataset contains original code, patched code, prediction results, and various evaluation metrics. The dataset is divided into two configurations: raw and treated, with the treated configuration additionally providing human scores. The training set includes a large number of code examples for training and evaluating related models.