// ∿ New Architectures: SSM & Linear Attention
Mamba (Selective State Spaces)
Input-dependent state spaces with linear-time cost and Transformer-like quality.
Mamba 选择性状态空间 2023 Advanced core
IN PLAIN WORDS
S4's fixed parameters cannot selectively remember. Mamba makes them depend on the token and uses a hardware-aware parallel scan. Inference uses constant memory per step with no KV cache.
Used in: Jamba, Falcon Mamba, hybrid models
TIMELINE · prerequisites → this → what builds on it
01Original paper
Mamba: Linear-Time Sequence Modeling with Selective State Spaces
Albert Gu, Tri Dao · 2023
arXiv ↗ ✓ Verified · arXiv API
02Where it sits
Learn the left column first; the right column builds on this one.
Learn first
Then you can learn
Related:Gated Delta Networks
Left: prerequisites · right: builds on this · dashed: related. Full map →
03Learning resources
Official docs, explainers, courses and reference code. All links checked.
04Where the course teaches it
On learning paths:Generative Models & Frontier Research
For AI / data
- Verification: verified, checked 2026-10-21.
- JSON: /api/learn/algorithms.json (id:
mamba)