Choice Trials Dataset

该数据集包含了来自人工智能模型（Llama 2和Llama 3）在60,000个独立选择试验中的回应，在这些试验中，大型语言模型从五个刺激集合中挑选出偏好的赌博选项。此外，该数据集还按大型语言模型和赌博数据集分类，汇总了多个随机化种子的回应，这允许评估在不同呈现格式下的传递性违规情况。该数据集的规模为60,000个独立选择试验，其任务是评估人工智能生成选择中的偏好传递性。

This dataset comprises responses from AI models (Llama 2 and Llama 3) across 60,000 independent choice trials, where large language models select preferred gambling options from five stimulus sets. Furthermore, the dataset aggregates responses across multiple randomization seeds, categorized by both the large language model and the gambling dataset, which facilitates evaluation of transitivity violations under different presentation formats. Consisting of 60,000 independent choice trials, this dataset is designed to assess preference transitivity in AI-generated choices.