@misc{indiciae567abd069140,
  title = {Do LLMs Need to See Everything? A Benchmark and Study of Failures in LLM-driven Smartphone Automation using Screentext vs. Screenshots},
  author = {Shiquan Zhang and Tianyi Zhang and Le Fang and Simon D'Alfonso and Hong Jia and Vassilis Kostakos},
  year = {2026},
  url = {https://arxiv.org/abs/2604.17817},
  note = {Source identifier: 2604.17817}
}