1: \begin{abstract}

2: In this paper, we design non-resonant low-regularity numerical integrators for the cubic nonlinear stochastic Schr$\ddot{\rm o}$dinger equations (SNLSE). First, we begin with a mean-square convergence analysis for an explicit low-regularity stochastic integrator for the cubic SNLSE and, then, under the assumption of initial data of size $\mathcal{O}(\varepsilon), \varepsilon \in (0,1]$, following the {regularity compensation oscillation} technique \cite{sch0,ala,bao}, we provide rigorous long-term error estimates in the space $L^2(\Omega, H^\sigma), \sigma >1/2$. Our findings reveal improved long-time behaviour allowing for an error of size

3:  $\mathcal{O}\left(\tau\cdot \varepsilon^{{\rm min}(4,2(q-2))}\right), q>2$, up to time of size $\mathcal{O}(\varepsilon^{-2})$. Numerically findings confirm the superior long-time behaviour of our new schemes compared to classical numerical schemes.

4: \end{abstract}

5: