nvidia/Nemotron-Cascade-SFT-SWE

Nemotron-Cascade-SFT-SWE数据集是用于SWE代码修复任务的RL训练数据，由SWE-Bench-Train、SWE-reBench、SWE-Smith、R2E-Gym/R2E-Gym-Subset和SWE-Fixer-Train组成。根据难度选择SFT和RL阶段的训练数据，并排除评估数据集中存在的仓库实例以避免数据污染。数据集的提示遵循agentless mini框架，包含三个任务：错误代码定位、代码修复和测试用例生成。响应是通过DeepSeek-R1-0528生成的。使用该数据集进行SFT和RL训练后，8B和14B模型的pass@1解决率分别达到37.2和43.1（无TTS）。数据集的统计信息包括各任务的样本数量，结构示例包括类别、来源、消息、生成器、补丁和思考模式等字段。

The Nemotron-Cascade-SFT-SWE dataset is the RL training data for SWE code repairing task, consisting of SWE-Bench-Train, SWE-reBench, SWE-Smith, R2E-Gym/R2E-Gym-Subset and SWE-Fixer-Train. We select the training data for SFT and RL stages based on its difficulty. Also, to avoid data contamination, we exclude all instances originating from repositories present in the evaluation dataset. We create the prompts following the agentless mini framework, consisting of three tasks buggy code localization, code reparing and test case generation. The response is generated using DeepSeek-R1-0528. Using the data for SFT and RL, we reach pass@1 resolve rate (without TTS) 37.2 and 43.1 for 8B and 14B models. The statistics for our training data include the number of samples for each task. An example of datum includes category, source, messages, generator, patch, and thinking mode.