Benchmarking General Mobile Assistants in Challenging Real-World Scenarios

arXiv:2608.27477v1 Announce Type: new Abstract: Graphical user interfaces have emerged as an important environment for evaluating autonomous AI agents on multimodal interactive tasks.