INSTRUCTEVAL

该数据集是一个专为指令微调大型语言模型设计的综合评估套件，它不仅评估模型的问题解决能力，还考察其写作能力以及对人类价值观的契合度。该数据集包含多种任务，旨在评估模型的写作能力以及与人类价值观的一致性，强调了指令微调数据在模型性能中的重要性。这项任务的目的是对指令微调后的大型语言模型进行评估。

This dataset is a comprehensive evaluation suite specifically developed for assessing instruction-tuned large language models (LLMs). It covers multiple dimensions of model performance, including problem-solving ability, writing proficiency, and alignment with human values. The dataset comprises a variety of tasks, which aim to evaluate models' writing capabilities and their consistency with human values, while highlighting the critical importance of instruction-tuning data in model performance. The core goal of this evaluation task is to assess instruction-tuned large language models.