In on hiding information from an oracle, the authors (Abadi, Feigenbaum, and Kilian) wrote:
$(\mathsf{NP/poly} \cap \mathsf{co\text-NP}{/poly})$ ... is not known to be equal to $(\mathsf{NP} ∩ \mathsf{co\text-NP}){/poly}$.
They highlighted that in the conference paper, they mistook the two classes. Apparently, the latter is a subset of the former, but we don't know if the containment is strict.
Assuming $X$ and $Y$ are complexity classes, and $F$ is a set of functions specifying the length of the advice strings, are there recent results comparing $(X_{/F} \cap Y_{/F})$ and $(X \cap Y)_{/F}$, resolving issues like the one pointed above?