Response Optimization Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/henryyantq/OptimaLLM
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了一套日常生活中常遇到的五个问题，旨在评估经过改进的GPT-3.5模型在与基准模型以及GPT-4模型的对比中，在回答这些问题时的表现。这些问题既包括事实性问题，也包括推理性问题，评估标准基于人类专家对回答的准确性、简洁性和完整性进行判断。此次评估的规模为五个问题，任务是对语言模型输出的响应进行评价和优化。
This dataset contains five daily-life questions commonly encountered in real life, intended to evaluate the performance of the enhanced GPT-3.5 model when answering these queries, compared against both the baseline model and GPT-4. These questions cover both factual questions and reasoning-based problems. The evaluation criteria are based on human experts' assessments of the accuracy, conciseness, and completeness of the model's responses. With a total of five questions for this evaluation, the core task is to assess and optimize the responses generated by language models.
提供机构：
OpenAI (for GPT-3.5-Turbo model)
