UWNSL/MATH_training_split_distill_small_teacher

该数据集是一个数学问题集，包含问题（problem）和解决方案（solution）字段。它由MATH数据集中7.5k的训练分割部分构成，使用Qwen/Qwen2.5-3B-Instruct小型教师模型生成答案。数据集的目的是研究不同学生模型从大型教师模型与小型教师模型进行知识蒸馏的效果，特别是小型模型在学习复杂推理链方面的限制。数据集遵循统一问答许可协议和知识共享署名-非商业性使用4.0国际许可协议。

This dataset is a collection of math problems, including problem and solution fields. It is constructed from the 7.5k training split of the MATH dataset, using the Qwen/Qwen2.5-3B-Instruct small teacher model to generate answers. The purpose of the dataset is to study the effect of knowledge distillation from large teacher models to small teacher models on different student models, especially the limitations of small models in learning complex chains of reasoning. The dataset follows the Tongyi Qianwen License Agreement and the CC BY-NC 4.0 license.