VmaxRL/bugpilot-bugintro-lm-modify-gpt55-repaired-34-cleaned-1k-20260526T082957Z

该数据集是一个软件工程测试与修复数据集，包含1000个训练样本，总大小约155MB。数据集中每个样本涉及代码仓库、测试框架、补丁生成和验证等信息，具体字段包括agent_instruction_md（代理指令）、base_commit（基础提交）、docker_image（Docker镜像）、fail_to_pass（失败到通过的测试列表）、gold_patch（黄金补丁）、problem_statement_md（问题陈述）、requirements_md（需求说明）等。元数据部分（metadata）进一步记录了测试计数、数据集ID、源仓库等详细信息，表明数据集用于支持代码自动修复、测试验证或软件工程研究任务。This dataset is a software engineering testing and repair dataset containing 1000 training samples with a total size of approximately 155MB. Each sample in the dataset includes information related to code repositories, testing frameworks, patch generation, and verification, with specific fields such as agent_instruction_md, base_commit, docker_image, fail_to_pass (list of tests transitioning from fail to pass), gold_patch, problem_statement_md, requirements_md, and more. The metadata section further records details like test counts, dataset IDs, source repositories, etc., indicating that the dataset is designed to support tasks such as automated code repair, test validation, or software engineering research.