 Solution

In a local hidden-variable description, assign predetermined values $X_j,Y_j\in \{+1,-1\}$ to the two possible measurements at photon $j$. The perfect correlations of b requireMultiplying these identities cancels each squared X value and gives $\boxed{Y_1{Y}_2{Y}_3=-1}$. Thus local realism predicts only the YYY outcomes $++-,$ $+-+,$ $-++,$ and $---$. Combining this parity rule with the measured unbiased single-Y marginals forces all four frequencies to be $1/4$: four weights subject to normalization and the three zero-mean conditions are uniquely uniform.

Quantum mechanics instead has $k=3$ in the amplitude formula of b, giving ${\mathcal{A}}_3=(1+abc)/4$. ThereforeQuantum mechanics predicts each positive-product outcome with probability $1/4$ and every negative-product outcome with probability zero. The local and quantum allowed sets are disjoint.

This is the GHZ theorem: in the ideal case the contradiction follows from deterministic product constraints, an all-versus-nothing argument. A usual two-party Bell-state test instead violates a statistical Bell inequality assembled from several settings. Both require appropriate locality and measurement-choice assumptions and careful treatment of experimental imperfections; the distinction is the deterministic form of the ideal three-party argument.