Mixture of Attention Heads: Selecting Attention Heads Per Token.

Xiaofeng Zhang, Yikang Shen, Zeyu Huang, Jie Zhou, Wenge Rong, Zhang Xiong

Browse the full EMNLP paper archive.