Mamba and Transformers through state-space duality

Published on September 24, 2026, the article develops theoretical links between state-space models (SSMs), such as Mamba, and attention variants. Its approach uses structured semiseparable matrix decompositions to relate these architectures.

According to the description, this structure helps engineers understand the relationship between SSMs and attention models. To assess the scope of the result, consult the original article and check how it presents its decompositions and conclusions; the available description gives no further details about specific methods or findings.