lime-nlp/orz_math_difficulty

该数据集是对Open Reasoner Zero数据集中的问题进行难度评分的标注数据集。它使用了Qwen 2.5-MATH-7B模型来评估问题的难度，并为每个问题计算了一个难度分数。这个难度分数可以作为构建自适应课程的信号。Open Reasoner Zero是一个包含57,000个推理密集型问题的数据集，用于训练和评估基于强化学习的大型语言模型方法。

This dataset is an annotated dataset that scores the difficulty of problems in the Open Reasoner Zero dataset. It uses the Qwen 2.5-MATH-7B model to evaluate the difficulty of problems and calculates a difficulty score for each problem. This difficulty score serves as a signal for constructing adaptive curricula. Open Reasoner Zero is a dataset consisting of 57,000 reasoning-intensive problems used to train and evaluate reinforcement learning-based methods for large language models.