extract_original_inputs() returns @x_original_inputs — the column names
present in the raw data before any feature engineering.
extract_feature_names() returns @x_names — the names after feature
engineering (i.e. what the model actually sees).