I3D: Transformer Architectures with Input-Dependent Dynamic Depth for Speech Recognition.

Yifan Peng, Jaesong Lee, Shinji Watanabe

Browse the full ICASSP paper archive.