GUI agents built on large language and vision-language models still struggle on unseen applications and complex multi-step tasks, as completing real GUI task…
机构:西安交通大学
来源:arXiv 2609.39547 | AI4Papers 论文推荐平台