LLM-Powered GUI Agents in Phone Automation: Surveying Progress and Prospects
A maintained paper list and taxonomy accompanying a TMLR survey on using LLMs to automate phone interfaces, covering prompting, training-based methods, datasets, and benchmarks.
This repository maintains the paper list behind a survey, accepted to TMLR, on large-language-model-powered GUI agents for phone automation. It organizes the field into a taxonomy of representative work and tracks milestones across four areas: prompt engineering, which adapts pre-trained LLMs through crafted prompts without changing model parameters; training-based methods, which fine-tune or reinforcement-learn on GUI-specific data to improve mobile-interface understanding; datasets; and benchmarks.
The list has been updated repeatedly since the survey's first release, with the maintainers stating they intend to keep updating it and inviting contributions by pull request, issue, or email. The repository itself is a reference companion to the paper rather than a runnable tool, aimed at researchers who want a citation-ready, categorized entry point into the phone-GUI-agent literature.