Consider the following optimization problem: The optimized variable A0,B0,A1,B1 are dimension-2 real symmetric matrices. All of their eigenvalues have absolute value 1 . rho is a given dim-4 matrix. The objective function is max Tr[(A0⊗B0 +A0⊗B1+A1⊗B0 -A1⊗B1)*rho]. Here ⊗ is the Kronecker Product. The background is from the Tsirelson’s bound in quantum information.