Decoding the Functional Roles of Register and High-Norm Patch Tokens in Vision Transformers
cs.CV
Submitted: 2026-10-02
Updated: 2026-10-02
Code: https://github.com/facebookresearch/dinov2
Terminology
Sources
- Sparse Autoencoders Find Highly Interpretable Features in Language Models
- Vision Transformers Need Registers
- Archetypal SAE: Adaptive and Stable Dictionary Learning for Concept Extraction in Large Vision Models
- Language Models Can Explain Visual Features via Steering
- Scaling and evaluating sparse autoencoders
- Vision Transformers Don't Need Trained Registers
- DINOv2: Learning Robust Visual Features without Supervision
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models