ReproduceBench
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/AI9Stars/AutoReproduce
下载链接
链接失效反馈官方服务：
资源简介：
该数据集提供了一个带有验证实施的标准基准，旨在评估研究论文中描述实验的复现能力。此外，它还包括了新颖的评价指标，用以评估实验复现的准确性和执行忠实度。该数据集的任务是评估自动实验复现的能力。
This dataset provides a standard benchmark with validated implementations, aimed at evaluating the reproducibility of experiments detailed in research papers. Furthermore, it incorporates novel evaluation metrics to assess the accuracy of experimental reproduction and the fidelity of execution. The core task of this dataset is to evaluate the capability of automated experiment reproduction.
提供机构：
AutoReproduce Project
