thoughtworks/gemma_psychometrics_personas_responses
收藏资源简介:
该数据集是一个多配置集合,专注于心理评估和人格测试任务,用于分析AI模型在模拟不同人格特征下的行为表现。它包括多个子数据集,如AdvBench(对抗性基准)、Emo Bench(情绪基准)、HEXACO(六大人格特质模型)、SJT(社会判断任务)和TruthfulQA(真实性问答)。每个子数据集包含详细的个人资料(如年龄、性别、教育背景、心理特征)、问题与回答、模型响应以及评估指标(如正确性、规范化答案)。数据集旨在支持研究AI模型在人格模拟、情绪理解、社会判断和真实性评估方面的性能,适用于NLP、心理学和AI伦理研究。
This dataset is a multi-configuration collection focused on psychological assessment and personality testing tasks, designed to analyze AI model behavior under simulated personality traits. It includes multiple sub-datasets such as AdvBench (adversarial benchmark), Emo Bench (emotional benchmark), HEXACO (six-factor personality model), SJT (social judgment task), and TruthfulQA (truthfulness question answering). Each sub-dataset contains detailed persona profiles (e.g., age, gender, education, psychological traits), questions and answers, model responses, and evaluation metrics (e.g., correctness, normalized answers). The dataset aims to support research on AI model performance in personality simulation, emotional understanding, social judgment, and truthfulness assessment, applicable to NLP, psychology, and AI ethics studies.




