AI RESEARCH
Finding Distributed Object-Centric Properties in Self-Supervised Transformers
arXiv CS.AI
•
ArXi:2603.26127v1 Announce Type: cross Self-supervised Vision Transformers (ViTs) like DINO show an emergent ability to discover objects, typically observed in [CLS] token attention maps of the final layer. However, these maps often contain spurious activations resulting in poor localization of objects. This is because the [CLS] token, trained on an image-level objective, summarizes the entire image instead of focusing on objects. This aggregation dilutes the object-centric information existing in the local, patch-level interactions.