Synthetic Conversation Dataset

该数据集通过模拟两个具有不同人格特征的大型语言模型之间的对话来构建，反映了它们生活中不断变化的事实。这使得我们能够在合成对话环境中严格控制和评估推理能力。该数据集包含了3组对话，共检查了1,008个问题。其任务是对包含演变状态和推理的对话中的语言模型进行评估。

This dataset is constructed by simulating dialogues between two large language models (LLMs) with distinct personality traits, which mirrors the dynamically evolving factual details within their simulated life contexts. This setup enables rigorous control over and evaluation of reasoning capabilities within a synthetic conversational environment. The dataset includes 3 groups of conversations, with a total of 1,008 examined questions. Its core task is to evaluate large language models in conversational scenarios that involve evolving states and reasoning processes.