This report describes our approach for the EGO4D 2023 Visual Query 2D Localization Challenge. Our method aims to reduce the number of False Positives (FP) that occur because of high similarity between the visual crop and the proposed bounding boxes from the baseline's Region Proposal Network (RPN). Our method uses a transformer to determine similarity in higher dimensions which is used as our prior belief. The results are then combined together with the similarity in lower dimensions from the Siamese Head, acting as our measurement, to generate a posterior which is then used to determine the final similarity of the visual crop with the proposed bounding box. Our code is publicly available $\href{https://github.com/s-m-asjad/EGO4D_VQ2D}{here}$.