Bound for sum of squares $r_{2k}(m)$ for $m \geq 1$

Question

I happened to read in (Iwaniec-Kowalkski) Analytic Number Theory book that the Sum of Squares function satisfies the bound $r_{2k}(m) << m^{k-1+\epsilon}$ for $m \geq 1$. But $\epsilon$ is not mentioned explicitly as positive here (though I believe so). Can someone clarify about that? Also, would there be any obvious (or even non-obvious) reason why such an implied constant can't exist for negative $\epsilon$, i.e., if it is so?