Fovea Transformer: Efficient Long-Context Modeling with Structured Fine-To-Coarse Attention.

Ziwei He, Jian Yuan, Le Zhou, Jingwen Leng, Bo Jiang

Browse the full ICASSP paper archive.