Bell-inequality violations establish that two systems share some quantum entanglement. We give a simple test to certify that two systems share an asymptotically large amount of entanglement, n EPR states. The test is efficient: unlike earlier tests that play many games, in sequence or in parallel, our test requires only one or two CHSH games. One system is directed to play a CHSH game on a random specified qubit i, and the other is told to play games on qubits {i, j}, without knowing which index is i.The test is robust: a success probability within δ of optimal guarantees distance O(n 5/2 √ δ) from n EPR states. However, the test does not tolerate constant δ; it breaks down for δ =Ω(1/ √ n). We give an adversarial strategy that succeeds within δ of the optimum probability using onlyÕ(δ −2 ) EPR states.Accepted in Quantum 2018-08-17, click title to verify arXiv:1610.00771v2 [quant-ph] 30 Aug 2018We address this natural question. We develop a test for n EPR states worth of entanglement, 1 √ 2 n (|00 + |11 ) ⊗n . We will explain this test below, but first let us give some more context.A Bell-inequality violation certifies that there is some entanglement; a nearly optimal violation can certify that the entanglement is of the specific form of an EPR state [MMMO06, MYS12, RUV13]. Wu et al. [WBMS16] have shown that two CHSH games, played in parallel, can be used to test for n = 2 EPR states. One might reasonably suppose that if playing one or two CHSH games can show some entanglement, then playing many CHSH games should suffice to show lots of entanglement. It is not so simple. Different games need not be independent from each other, and complicated dependencies could conceivably allow low-dimensional systems to act like higher-dimensional ones [CRSV17].The first test for asymptotically many EPR states was given in [RUV13]. The motivation for this test was to develop a scheme for delegated quantum computation, allowing one to securely run a quantum circuit across several untrusted devices. The test tolerates polynomially small deviations from the ideal behavior, and in principle can be run with polynomial overhead, but it is severely impractical. The ideal systems need to share much more entanglement, N = n O(1) n EPR states, than is being certified. The test is based on playing N CHSH games sequentially. This is highly constraining: the answer to one game must be given before the question for the next game, and yet the whole sequence of questions and answers must be space-like separated from one system to the other. Finally, the test tolerates only an inverse-polynomial error rate in the systems.McKague [McK16] gave a much-improved test for n EPR states. His test uses only n EPR states; none are lost in the analysis. They are all measured, and the measurements must still be space-like separated from one system to the other, but they can be performed in parallel. The final distance from 1where δ is an upper bound on the error for each ofÕ(n) test settings. Thus to achieve error , one should set δ ∼ 8 /n 4 , so each E...