AI RESEARCH
ASAP: Attention Sink Anchored Pruning
arXiv CS.LG
•
ArXi:2605.22372v1 Announce Type: new Vision Transformers (ViTs) face severe computational bottlenecks due to the quadratic complexity of self-attention at high resolutions. Existing token reduction methods rely on local metrics - such as single-layer attention scores - that are inherently vulnerable to the attention sink phenomenon, where uninformative tokens are paradoxically preserved over salient foreground objects. We propose ASAP (Attention Sink Anchored Pruning), a.