Data semantics and styling#

These functions deliberately accept seaborn-style arguments, but the wrapper owns model-specific preprocessing before delegating to plotting machinery. Keep data semantics separate from visual style:

• first validate x/y/hue and class/score meaning;

• then choose kind and visual options;

• retain the returned Matplotlib object when later composition/customization is required; and

• save figures at the application boundary rather than burying file writes in reusable data-preparation code.