Show GN: VLM이 유저 시선을 얼마나 예측할 수 있는지 실제 아이트래킹 데이터로 비교

This research demonstrates how Vision Language Models can predict user gaze patterns on UI interfaces, providing crucial insights for UX design and usability testing. By validating VLM predictions against real eye-tracking data using the UEyes dataset, it establishes a scalable, cost-effective alternative to expensive eye-tracking studies. This has significant implications for automated accessibility assessment, user experience optimization, and advancing VLM capabilities in understanding human visual attention.