cmu-lti/interactive-swe

Interactive SWE-bench是一个软件工程问题解决的增强型数据集，包含500个经过人工验证的GitHub问题与拉取请求对。每个条目包括原始详细问题和简化版本的问题陈述，以及解决问题的关键补丁和环境配置信息。数据集适用于训练和评估自动化问题解决模型，支持详细的 leaderboard 对比研究。

Interactive SWE-bench is an enhanced dataset for software engineering issue resolution, containing 500 human-validated GitHub Issue-Pull Request pairs. Each entry includes both the original detailed issue and a simplified problem statement, along with the key patch and environment setup information for resolving the issue. The dataset is suitable for training and evaluating automated issue resolution models, supporting detailed leaderboard comparison studies.