This paper aims to uncover true synaptic connectivity from neural population activity using a self-supervised learning approach that addresses confounding effects from correlation-driven artifacts.