What DINO saw: ALiBi positional encoding reduces positional bias in Vision Transformers

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Pawlowsky, Moritz, Vamvakeros, Antonis, Weiss, Alexander, Bielefeld, Anja, Cooper, Samuel J., Docherty, Ronan
Format:	Preprint
Published:	2026
Subjects:	Computer Vision and Pattern Recognition Materials Science
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Internet

https://arxiv.org/abs/2603.16840