mask quality assessment for referring audio-visual segmentation