https://www.youtube.com/watch?v=7jlZlSxZZ1g&t=420s

https://www.youtube.com/watch?v=MWkZMV-6uCw&t=7s

📄 Paper Info

Title: Mamba: Linear-Time Sequence Modeling with Selective State Spaces

Authors: Albert Gu, Tri Dao

Year / Venue: 1 Dec 2023

Link: https://arxiv.org/pdf/2312.00752


🧠 1. One-Sentence Summary

Mamba is a sequence modeling architecture that achieves Transformer-level performance with linear-time scaling by introducing a selective state-space mechanism (S6) that allows the model to dynamically prioritize or forget information based on the input.


🎯 2. Main Contributions


📊 3. Experiments & Results


⚠️ 4. Limitations


💡 5. My Understanding (Personal Notes)