Abstract: We propose a novel stabilized semi-supervised training method to solve the chal-lenging problem of covid lesion segmentation in CT scans. We first study the limita-tions of current models and based on our findings we introduce a lightweight SU-Net(Small U-Net) architecture. During training we feed the CT scans in sorted order oflesion occupancy and calculate a reliability score at each epoch to determine the stop-ping criteria. We test the proposed method on the largest publicly available COVID CTdataset called MOSMED dataset. By harnessing around 800 un-labelled COVID CTvolumes comprising 25k CT slices, we improve the segmentation accuracy by around2-4 dice percentage points depending upon the availability of labelled training data. We also compare our method with a recently published COVID lesion segmentation methodcalled Semi-InfNet. The proposed method outperforms Semi-InfNet model and achievesstate-of-the-art covid segmentation result on MOSMED dataset.