The Structural Origin of Attention Sink: Variance Discrepancy, Super Neurons, and Dimension Disparity

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Siquan, Jiang, Kaiqi, Sun, Jiacheng, Hu, Tianyang
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!