HiddenProfile

该数据集是一个用于评估多代理大型语言模型系统的基准，它采用了来自社会心理学文献中的隐藏剖面范式。该基准包含了从先前的人类受试者研究中改编的任务，并评估了不同大型语言模型的表现。数据集规模涵盖了九个任务，横跨多样化的场景。具体任务旨在评估在分布式知识下，多代理系统中的集体推理能力。

This dataset is a benchmark for evaluating multi-agent large language model systems, which adopts the hidden profile paradigm from social psychology literature. This benchmark includes tasks adapted from prior human-subject research, and evaluates the performance of various large language models. It covers nine tasks across diverse scenarios. The specific tasks are designed to assess the collective reasoning capabilities of multi-agent systems under distributed knowledge.