Researchers have introduced SwipeGen, a novel tool designed to synthesize human-like swipe interactions for GUI agents, addressing a key limitation in their real-world applicability. Existing agents often fail at complex tasks due to rigid swipe execution, lacking the fine-grained control over trajectory, speed, and timing that humans naturally employ. To tackle this, SwipeGen generates diverse and realistic swipe data, accompanied by SwipeBench, a new benchmark for evaluating swipe interaction quality. Experiments demonstrate that agents utilizing SwipeGen can achieve up to a 2.46x improvement in swipe execution success rates. AI
IMPACT Enhances the usability and effectiveness of GUI agents by enabling more human-like interaction capabilities.
RANK_REASON The cluster describes a new research paper introducing a novel tool and benchmark for improving GUI agents. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
