Quantum bounds for compiled XOR games and $d$-outcome CHSH games
Matilde Baroni, Quoc-Huy Vu, Boris Bourdoncle, Eleni Diamanti, Damian Markham, Ivan Šupić
TL;DR
The paper investigates whether Kalai et al.'s quantum-homomorphic-encryption-based compiler preserves the quantum bound for nonlocal games beyond CHSH in a single-prover setting. Using a cryptographic sum-of-squares (SOS) decomposition and a pseudo-expectation map, the authors transfer nonlocal bounds to compiled games while controlling security via a negligible function $\delta_{\mathrm{QHE}}(\kappa)$, enabling rigorous upper bounds on compiled scores. They prove that the quantum bound is preserved for XOR games and for $d$-outcome SATWAP/CHSH-type inequalities, and they establish computational self-testing results for pairs of qubit measurements and for three Pauli observables within compiled protocols. The framework combines SOS techniques, NPA-type reasoning, and cryptographic guarantees to enable device-independent certification on a single quantum device, with implications for practical quantum certification and cryptographic protocols; concurrent work and several open directions are also discussed.
Abstract
Nonlocal games play a crucial role in quantum information theory and have numerous applications in certification and cryptographic protocols. Kalai et al. (STOC 2023) introduced a procedure to compile a nonlocal game into a single-prover interactive proof, using a quantum homomorphic encryption scheme, and showed that their compilation method preserves the classical bound of the game. Natarajan and Zhang (FOCS 2023) then showed that the quantum bound is preserved for the specific case of the CHSH game. Extending the proof techniques of Natarajan and Zhang, we show that the compilation procedure of Kalai et al. preserves the quantum bound for two classes of games: XOR games and d-outcome CHSH games. We also establish that, for any pair of qubit measurements, there exists an XOR game such that its optimal winning probability serves as a self-test for that particular pair of measurements.
