Implementability constraints and accessibility hypergraphs

If

for an affine Hermitian-matrix-valued map $\mathcal{F}$, the restricted bias is the larger optimum of two semidefinite programs with objectives $\pm \mathrm{tr}(M\Lambda (X_J))$. Convexity alone does not imply that an arbitrary effect class has an exact semidefinite representation.