Fig. 2

Bedshift and similarity score results. A Similarity scores for shift, drop, and add perturbations, along with their pairwise combinations. B Summary of similarity score change for different perturbation levels measured. For a given scenario, each perturbation is classified as either “increasing,” “not used,” or “held constant.” To condense the information from panel A, the amount of decrease shown for each Scenario is the average across all three levels in the corresponding Scenario in panel A. C Summary of the sensitivity of each metric to the three different perturbations