PersonaX
收藏资源简介:
PersonaX是一个多模态数据集集合,旨在全面分析公共特征。它由两部分组成:CelebPersona,包含来自CelebA数据集的9444位公众人物,以及AthlePersona,覆盖7个主要体育联盟的4181名职业运动员。每个数据集都包括由三个高性能大型语言模型推断的行为特征评估,以及面部图像和结构化的传记信息。PersonaX提供了一个研究LLM推断行为特征与视觉和传记属性相结合的独特基础,促进了多模态特征分析和因果推理的新途径。
PersonaX is a multimodal dataset collection designed for comprehensive analysis of public characteristics. It comprises two core components: CelebPersona, which encompasses 9,444 public figures sourced from the CelebA dataset, and AthlePersona, which covers 4,181 professional athletes across seven major sports leagues. Each dataset includes behavioral trait assessments inferred by three high-performance large language models, alongside facial images and structured biographical information. PersonaX provides a unique foundation for investigating the integration of LLM-inferred behavioral traits with visual and biographical attributes, opening up novel avenues for multimodal feature analysis and causal reasoning.

- 1通过Zayed大学人工智能学院,卡内基梅隆大学,加利福尼亚大学圣地亚哥分校,澳大利亚国立大学 · 2025年



