TIDES: Selective SSMs with Implicit Time-Awareness for Irregular Series
October 4, 2026
TIDES improves Selective State Space Models like Mamba by allowing them to handle irregular timestamps natively. It moves input dependence from the time discretization step to the diagonal state matrix, reconciling the expressivity of selective models with the continuous nature of S5.
HOW THIS AFFECTS YOU
●
builderThis provides a more robust SSM architecture for time-sensitive sensor or financial data applications.
●
researcherYou can build models that handle irregular time-series data without losing per-token expressivity.