Max-function formulation of the ReLU depth-separation conjecture

Let $n=2^k$, and let $\operatorname{ReLU}_n(k)$ be the set of continuous piecewise linear functions on $\mathbb{R}^n$ computable with $k$ hidden layers. Max-function formulation of the depth-separation conjecture. For $n=2^k$, the function

$$(x_1,\ldots,x_n)\longmapsto\max\{0,x_1,\ldots,x_n\}$$

is not contained in $\operatorname{ReLU}_n(k)$. This statement is presented as equivalent to the preceding depth-separation conjecture, based on a result of Wang. It is open; for example, it is unknown whether two hidden layers suffice for $\max\{0,x_1,x_2,x_3,x_4\}$.