IBM demonstrates quantum computers that exceed classical verification limits

IBM and its research partners have reached a milestone in quantum computing: their quantum processors can now perform calculations that surpass what classical computers can verify through traditional simulation methods. This development signals real progress in the field, though it introduces a new challenge: if a quantum computer solves a problem that classical machines cannot double-check, how do you trust the answer is correct?

The company and collaborators from the University of Chicago, Qedma, RIKEN, BlueQubit, and Algorithmiq published three research papers this week describing different approaches to building confidence in quantum results. All three remain unpublished preprints awaiting peer review.

Building verification into quantum circuits#

The first research effort involved encoding a 70-qubit circuit using 97 physical qubits on IBM's Heron R3 processors. Of those 97, 70 qubits performed the actual computation while 27 qubits—known as ancillas—served as error-detection checks. The researchers began with a Clifford circuit, which classical computers can simulate efficiently, then added 468 T-gates that perform 45-degree phase rotations. These T-gates made the overall circuit extremely difficult for classical methods to reproduce.

When the ancilla qubits detected errors (triggering what's called a nonzero syndrome), the team discarded those runs—a technique called postselection. The approach generated about 2,000 usable results from over 16 minutes of computation. Critically, the error-detection data allowed researchers to establish a fidelity lower bound (a measure of how accurately the quantum state matched the intended state) with 95% confidence.

"Verification remains one of the biggest challenges in firmly establishing experimental quantum advantage," notes Bill Fefferman of the University of Chicago. The key insight here is that the quantum hardware itself provides evidence of correctness—rather than relying solely on analyzing output samples.

Testing across different quantum architectures#

The second approach took a different tack. Researchers from Qedma simulated a Floquet Ising model (which models how magnetic systems respond to repeating external pulses) using 51 and 74 qubits. In the early cycles, both quantum and classical calculations agreed. But as the simulation deepened, classical methods either stopped converging or produced conflicting answers, while the error-mitigated quantum results remained stable and showed persistent oscillations.