Browse Teach Curate

Input-Dependent Selection in State Space Models

making the state-space parameters a function of the current token is what gave SSMs the content-based filtering they lacked, and it is the gating node from the LSTM section returning in new clothes

jeremy Video

Mamba State Space Models Enhancing LLMs in Natural Language Processing

State Space Models (SSMs) formalize sequence modeling through continuous dynamical systems defined by state equations that evolve over time and output equations mapping states to predictions, fundame…