Motion Mamba: Efficient and Long Sequence Motion Generation with Hierarchical and Bidirectional Selective SSM
Paper • 2403.07487 • Published • 16
Papers - Mamba - Structured State Space Model
-
-
LocalMamba: Visual State Space Model with Windowed Selective Scan
Paper • 2403.09338 • Published • 8 -
Cobra: Extending Mamba to Multi-Modal Large Language Model for Efficient Inference
Paper • 2403.14520 • Published • 35 -
SiMBA: Simplified Mamba-Based Architecture for Vision and Multivariate Time series
Paper • 2403.15360 • Published • 13 -
Gamba: Marry Gaussian Splatting with Mamba for single view 3D reconstruction
Paper • 2403.18795 • Published • 20 -
Jamba: A Hybrid Transformer-Mamba Language Model
Paper • 2403.19887 • Published • 112 -
MambaMixer: Efficient Selective State Space Models with Dual Token and Channel Selection
Paper • 2403.19888 • Published • 12 -
Locating and Editing Factual Associations in Mamba
Paper • 2404.03646 • Published • 3 -
Sigma: Siamese Mamba Network for Multi-Modal Semantic Segmentation
Paper • 2404.04256 • Published • 6 -
HSIDMamba: Exploring Bidirectional State-Space Models for Hyperspectral Denoising
Paper • 2404.09697 • Published • 1
Caduceus: Bi-Directional Equivariant Long-Range DNA Sequence Modeling
Paper • 2403.03234 • Published • 14Note Gated MLP mechanism. This is depicted in the left-most schematic in Figure 1. An incoming sequence is copied and projected to twice the input dimension. One copy is then passed through a causal convolution, followed by the SiLU/Swish non-linear activation (Ramachandran et al., 2017) and then finally through the selective SSM. The other copy has the SiLU non-linearity applied to it and then gates the SSM output. The gated representation is then projected back to the original dimension D.