PersonaCUABench: Benchmarking Computer-Use Agents Serving Diverse Personas

Oct 16, 2026ยท
Haoming Li
,
Defu Cao
,
Xu Hu
,
Yingtao Luo
Yiyang "Diana" Wang
Yiyang "Diana" Wang
,
Xianghe Xu
,
Jiashu He
,
Wanting Jiang
,
Yifan Zhang
,
Yunhan Zhao
,
Siyang Wu
,
Chen Yang
,
Haixin Wang
,
Zheyuan Deng
,
Taoran Li
,
Jingwen Zhang
,
Zhejian Peng
,
Shushu Liang
,
Yue Zhao
,
Bingzhe Li
,
Feng Chen
,
Yifan Wang
,
Jintao Huang
,
Yijun Wang
,
Yuexing Hao
,
Xiaomin Li
ยท 1 min read
Type
Publication
arXiv preprint โ€” under submission

PersonaCUABench evaluates computer-use agents through the people they serve. It holds tasks fixed while varying user personas, letting agents interact with requests that may omit requirements, add constraints, or call for approval before consequential actions.

The benchmark distinguishes task completion from meeting a user’s expressed needs, examining both agent performance and user satisfaction.