philippbecker/troll_data
收藏Hugging Face2025-11-01 更新2025-11-15 收录
下载链接：
https://hf-mirror.com/datasets/philippbecker/troll_data
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含在TROLL项目中使用的几个数据集，这些数据集用于评估大型语言模型的强化学习。具体包括：GSM8k数据集、DAPO数据集（分为DAPO训练集和DAPO评估集），以及EURUS数据集。DAPO和EURUS数据集在准备过程中进行了特定的修改或筛选，以适应项目需求。这些数据集主要与数学推理相关。
The dataset includes several datasets used in the TROLL project, which are used to evaluate reinforcement learning for large language models. Specifically, it includes the GSM8k dataset, the DAPO dataset (divided into DAPO Train and DAPO Eval), and the EURUS dataset. The DAPO and EURUS datasets have been specifically modified or filtered in preparation for the projects needs. These datasets are primarily related to mathematical reasoning.
提供机构：
philippbecker
