Thumb on the Scale: Optimal Loss Weighting in Last Layer Retraining | Read Paper on Bytez