Moving Beyond Sparse Grounding with Complete Screen Parsing Supervision
- Said Gürbüz
- Sunghwan Hong
- et al.
- 2026
- ICML 2026
Conference paper
Said Gürbüz is a Ph.D. student at ETH Zürich in the Computer Vision and Geometry Group, led by Prof. Marc Pollefeys, and a Pre-Doctoral Researcher at IBM Research Zurich with Dr. Peter W. J. Staar on the Docling team.
His research focuses on efficient vision-language models and screen understanding for computer-use agents. He holds an M.Sc. in Computer Science from EPFL, where he worked on multimodal document understanding and contributed to the Docling and SmolDocling projects at IBM Research, with the latter presented at ICCV 2025.