Examples of false-positive candidates from the network committee that were cleaned with post-processing scripts. A major fraction of contaminants correspond to crowded fields and/or cutouts with nonzero background emission in g, r, and/or i-band, which are rejected by SExtractor and HSC pixel-level flags (top row) or by the modeling network from S23a (bottom row). We show the average score of the network committee and the HSC ID in the top and bottom, respectively, of each panel, which is ~10″ on a side. We further show the predicted lens center as a cross with a length corresponding to the predicted uncertainty, and the Einstein radius as a solid green circle with predicted uncertainty bounds as dotted white lines. Some of the lower limits on the Einstein radii are near zero and are thus not visible.