VmaxRL/bugpilot-bugintro-lm-modify-gpt55-1k

BugPilot BugIntro LM_modify GPT-5.5 1k数据集是一个关于BugPilot LM_modify任务的语料库，来源于Demiurge LM_modify恢复工件。数据集包含1000行数据，列包括实例ID、补丁、从失败到通过的序列、从通过到通过的序列、图像名称、代码仓库和问题陈述。问题陈述使用azure/gpt-5.5-1重新生成。此外，数据集还提供了补丁统计信息作为辅助元数据，包括行级和仓库级的统计数据，如添加的行数、删除的行数、总更改行数、更改的文件数等。

The BugPilot BugIntro LM_modify GPT-5.5 1k dataset is a corpus for the BugPilot LM_modify task, exported from the Demiurge LM_modify recovery artifacts. It contains 1000 rows with columns including instance_id, patch, FAIL_TO_PASS, PASS_TO_PASS, image_name, repo, and problem_statement. Problem statements were regenerated for all rows using azure/gpt-5.5-1. Additionally, the dataset includes patch statistics as sidecar metadata, such as row-level and repo-level aggregates, including lines of code added, lines removed, total changed lines, files changed, etc.