[CVPR 2026 Oral] "INSID3: Training-Free In-Context Segmentation with DINOv3" - View it on GitHub