There is confusion in the Evaluate API between classification and binary_soft_classification terms (the user may be confused which one of these two should they be using in a particular case).
As binary_soft_classification is used as means to evaluate outlier detection analysis, we should rename binary_soft_classification to outlier_detection.
Note: Ultimately, we may want to be able to perform all kinds of classification-related evaluations (binary vs multi-class, soft vs hard) using classification evaluation stanza. But for now simply introducing 1:1 mapping between analyses names and evaluations names should increase API understandability for the users