A Gaussian variable $X_i\sim {\cal N}(0,1)$ is anti-concentrated in the following sense: for any $\epsilon>0$ we have:
$$
\mathbf{P}( |X_i| \leq \epsilon ) = O(\epsilon).
$$
Hence if we consider a polynomial $P(X_1,..., X_n)$ of degree $d$, 
in $n$ independent Gaussian variables, then it is clear that the best anti-concentration bound we can hope is
$$
\mathbf{P}( P(X_1,..., X_n) \leq \epsilon) = O(\epsilon^{1/d}).
$$
This follows immediately when we consider a monomial of the form $X_i^d$ for some Gaussian variable $X_i$, $i\in [n]$.
A seminal result by Carbery and Wright shows that this is tight
up to a factor of $d$:
$$
\mathbf{P}( P(X_1,..., X_n) \leq \epsilon) \leq d \epsilon^{1/d}.
$$

However, what is known when we assume in addition that $P$ is $\textbf{multi-linear}$ in the variables $X_i$.  Is it anti-concentrated more like the degree-1 uni-variate case, or does it behave like the $d$-power monomial case?