Recent work by Wayland, Coupette, and Rieck (2024) proposes a method to characterize and compare the latent embedding spaces arising from machine learning models. Their method is based on persistent homology and allows variability and sensitivity analysis of various hyperparameter choices for these models. Inspired by this idea but focusing on the case of classification problems, we would like to develop tools for a similar analysis. In this talk, we define a variant of multiparameter persistence landscapes, which can be seen as a generalization of the definition in the recent work by Vipond (2020). For practical applications, we are interested in the landscapes that are defined over a poset that is a product of $\mathbb R$ and a subposet of an inclusion poset. We discuss the properties of this definition, the theoretical challenges, and future directions.