scib-metrics get_results() default min_max_scale flipped True->False in 0.5.6; it can change the winner

Benchmarker.get_results() and plot_results_table() defaulted to min_max_scale=True up to scib-metrics 0.5.5. From 0.5.6 (July 2025, changelog entry for PR 215) the default is False, i.e. raw metric means. Tables made with default calls before and after that release are therefore not comparable. On the scIB human pancreas data with 6 embeddings (Unintegrated, ComBat, Harmony, Scanorama, scVI, scANVI), averaged over 10 seeds, raw Total ranked scANVI first and scVI third, while min-max-scaled Total ranked Harmony first and scVI fifth. Scaled scores are also sensitive to which methods are in the pool, since each metric is rescaled over the methods present: removing the Unintegrated baseline reordered the other five in most runs. Always pass min_max_scale explicitly and report it along with the method set.