Interpretability

A Comparative Analysis of Representation Flow in State-Space and Transformer Architectures
We study the feature flow across layers in state-space models compared to Transformers.