매일 아침, 어제의 AI를 한 통으로 정리해 보내드립니다메일로 받아보기

METAL LAB

AraSSM: A bidirectional state-space encoder for Arabic masked language modeling

arXiv:2608.082562026-08-11

arXiv:2608.08256v1 Announce Type: new Abstract: Pretrained Transformer encoders such as AraBERT, MARBERT, and CAMeLBERT have become the standard backbone for Arabic natural language understanding, but their self-attention mechanism scales quadratically with sequence length, which limits efficiency on long documents. Mamba, a selective state-space model (SSM), offers linear-time sequence modeling as a competitive alternative to attention, yet no dedicated bidirectional Mamba encoder pretrained sp

저자 · Ahmed Amine Aliane, Hassina Aliane, Nasredine Semmar

arXiv에서 원문 보기