T → 0: the distribution collapses onto the arg-max (a frozen system occupies its ground state). T → ∞: it flattens toward uniform (thermal agitation wins). This is the Boltzmann–Gibbs distribution wearing machine-learning clothes.