SemVink: Advancing VLMs' Semantic Understanding of Optical Illusions via Visual Global Thinking

Open in new window