1 paper · 1 filter
Weikang Meng, Yadan Luo, Xin Li +2
Linear attention has emerged as a promising alternative to softmax-based attention, leveraging kernelized feature maps to reduce complexity from quadratic to linear in sequence len…