zen-E/GSM8k-Aug-NL
收藏Hugging Face2025-04-08 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/zen-E/GSM8k-Aug-NL
下载链接
链接失效反馈官方服务：
资源简介：
该数据集用于训练CODI模型，包含两个子数据集GSM8k-Aug和GSM8k-Aug-NL。GSM8k-Aug通过去除自然语言交错的隐式CoT训练，仅保留结构化数学表达式。GSM8k-Aug-NL保留了自然语言解释，用于评估方法的泛化能力和对更冗长CoT的压缩有效性。
This dataset is used to train the CODI model and includes two subsets, GSM8k-Aug and GSM8k-Aug-NL. GSM8k-Aug is trained with implicit CoT by removing interleaved natural language, leaving only structured mathematical expressions. GSM8k-Aug-NL preserves natural language explanations to evaluate the generalizability and effectiveness of the approach in compressing more verbose CoTs.
提供机构：
zen-E
