I don't love that they used such different functional forms for each effect. It makes me think they're overfitting the data and putting too much weight on the specific features of the function they chose.

Maybe they explain these choices well in the paper, but it's a red flag to me.