Describe the workflow you want to enable

CategoricalNB : it needs to add proper treatment for unseen values in validation data (validation data has new values than train data )
to
https://scikit-learn.org/stable/modules/generated/sklearn.naive_bayes.CategoricalNB.html

meantime OrdinalEncoder put all unseen to one category - 0
CategoricalNB assumes that the sample matrix is encoded (for instance with the help of OrdinalEncoder) such that all categories for each feature are represented with numbers
where
is the number of available categories of feature .

using min_categories
also is not solving this problem

Describe your proposed solution

not sure
but unseen values in data row should be used for inference