Abstract
To date all self-tests for high dimensional systems are confined to many-qubit states. This is due to two restrictions in the literature: the standard techniques for proving self-testing results apply only to qubits, and there are few suitable non-local games that are known to require high dimensional non-qubit states to achieve their optimal quantum value. In this paper we address these two problems. Specifically, we generalize the usual self-testing framework to qits and then apply it to a generalization of the magic square game, giving a self-test for two maximally entangled pairs of qits. This self-test has two parties, perfect completeness for honest quantum players, and requires only a constant number of measurement settings, regardless of the dimension .