1: \begin{abstract}

2: The atomic scale dynamics of halide perovskites have a direct impact not only on their thermal stability but their optoelectronic properties.

3: Progress in machine learned potentials has only recently enabled modeling the finite temperature behavior of these material using fully atomistic methods with near first-principles accuracy.

4: Here, we systematically analyze the impact of heating and cooling rate, simulation size, model uncertainty, and the role of the underlying exchange-correlation functional on the phase behavior of \ce{CsPbX3} with X=Cl, Br, and I, including both the perovskite and the $\delta$-phases.

5: We show that rates below approximately \SI{30}{\kelvin\per\nano\second} and system sizes of at least a few ten thousand atoms are indicated to achieve convergence with regard to these parameters.

6: By controlling these factors and constructing models that are specific for different exchange-correlation functionals we then show that the semi-local functionals considered in this work (SCAN, vdW-DF-cx, PBEsol, and PBE) systematically underestimate the transition temperatures separating the perovskite phases while overestimating the lattice parameters.

7: Among the considered functionals the vdW-DF-cx functional yields the closest agreement with experiment, followed by SCAN, PBEsol, and PBE.

8: Our work provides guidelines for the systematic analysis of dynamics and phase transitions in inorganic halide perovskites and similar systems.

9: It also serves as a benchmark for the further development of machine-learned potentials as well as exchange-correlation functionals.

10: \end{abstract}

11: