Abstract: Analog compute-in-memory (ACiM) systems show promise for energy-efficient AI inference, but their performance is hindered by variations in conductance, resulting in reduced accuracy. This ...