MINT

该数据集是一个评估大型语言模型在多轮交互中使用工具和自然语言反馈解决问题的能力的基准。它包含了专注于推理、编程和决策制定的多样化评估数据集，这些数据集被精心挑选并整合成一个紧凑的子集，以便进行高效的评估。该任务的目的是评估大型语言模型的多轮交互能力。

This dataset is a benchmark for evaluating the ability of large language models (LLMs) to solve problems through tool usage and natural language feedback during multi-turn interactions. It includes diverse evaluation datasets focused on reasoning, programming, and decision-making. These datasets have been carefully selected and consolidated into a compact subset to enable efficient assessment. The objective of this task is to evaluate the multi-turn interaction capabilities of large language models.