Paper 'DONUT: A Decoder-Only Model for Trajectory Prediction' accepted to ICCV 2025
Sa2VA-i achieved 3rd place in the ICCV 2025 RVOS Challenge
DITR won the CVPR 2025 ScanNet++ Challenge
Paper 'EoMT' accepted to CVPR 2025
Multiple publications at CVPR, ICCV, WACV, and arXiv on topics including Vision Transformers, 3D segmentation, diffusion model fine-tuning, and efficient semantic segmentation
CVPR 2024 paper 'ALGM' proposed adaptive local-then-global token merging for efficient ViT-based segmentation
CVPR 2023 paper 'Content-aware Token Sharing' introduced a content-aware token sharing mechanism