Abstract: Speech separation using ad-hoc microphone arrays has been explored, but there is still significant room for improvement, especially in complex scenarios with varying channel conditions.