LSG Attention: Extrapolation of Pretrained Transformers to Long Sequences.

Charles Condevaux, Sbastien Harispe

Browse the full PAKDD paper archive.