Precision (left) and recall (right) of the merger class as a function of stellar mass for each method. The filled symbols show the metrics for the TNG dataset and the empty symbols for the Horizon-AGN dataset. For most methods and in both simulations, precision remains constant with mass, but then decreases with increasing mass at M_*> 10¹¹ M_⊙. There is a sharp downward trend in recall with decreasing mass for both datasets.