ByteDance-Seed/Multi-SWE-RL

Multi-SWE-RL数据集是一个由Multi-SWE-RL社区构建的开源数据集，旨在为复杂的软件工程任务提供高质量的强化学习（RL）数据。该数据集包含新收集的未注释数据以及从Multi-SWE-bench中丢弃的实例，可用于训练和评估解决现实世界编码挑战的自主代理。

The Multi-SWE-RL dataset is an open-source dataset built by the Multi-SWE-RL community, aiming to provide high-quality reinforcement learning (RL) data for complex software engineering tasks. This dataset includes newly collected unannotated data and discarded instances from Multi-SWE-bench, which can be used for training and evaluating autonomous agents that solve real-world coding challenges.