AI Hub
All papers
Architecture

Mamba: Linear-Time Sequence Modeling with Selective State Spaces

Carnegie Mellon·December 1, 2023

Albert Gu, Tri Dao

View on arXiv

TL;DR

A selective state-space model that matches Transformers while scaling linearly with sequence length.

Why it matters

Reignited interest in attention-free architectures for long-context, efficient modeling.

Related terms