This can be understood from solid-state physics.
First of all, for the investigated system with lattice constants $\vec{a}_1,\vec{a}_2$ and $\vec{a}_3$ (length), the reciprocal lattice vectors (1/length) can be calculated as:
\begin{align}
\vec{b}_1 & = 2\pi \dfrac{\vec{a}_2\times\vec{a}_3}{\Omega} \\
\vec{b}_2 & = 2\pi \dfrac{\vec{a}_3\times\vec{a}_1}{\Omega} \\
\vec{b}_3 & = 2\pi \dfrac{\vec{a}_1\times\vec{a}_2}{\Omega} \\
\end{align}
in which $\Omega=\vec{a}_1\cdot[\vec{a_2}\times\vec{a}_3]$ is the volume of the investigated system.
Then, back to the Kohn-Sham equation solvers like QE and VASP, they will do many integrals in reciprocal space with numerical techniques, which means that you should sample the reciprocal space along $\vec{b}_1, \vec{b}_2$ and $\vec{b}_3$. For example, if you adopt a $N_1 \times N_2 \times N_3$ k-mesh, you can calculated the resolution as
\begin{align}
\dfrac{\vec{b}_1}{N_1}, \qquad \dfrac{\vec{b}_2}{N_2}, \qquad \dfrac{\vec{b}_3}{N_3},
\end{align}
respectively. Basically, this just tell you how they are sampling the simulated cell in reciprocal space.
Hope it helps.