Why establishing trust matters for quantum computing

IBM says three new quantum advantage experiments could help users trust an answer no ordinary computer can check.

Three papers involving IBM and researchers from the University of Chicago, Qedma, Algorithmiq and other organizations describe calculations that reached regimes where leading classical approaches became impractical or produced inconsistent answers. They describe calculations that pushed beyond the practical reach of leading classical methods, along with new ways to check whether the quantum answers can be trusted. The studies also present ways to assess the quantum output in those regimes that can’t be accessed by conventional machines.

“Verification remains one of the biggest challenges in firmly establishing experimental quantum advantage,” Bill Fefferman, an Associate Professor at the University of Chicago, said in a news release.

Quantum computing is a way to process information using the laws that govern matter at extremely small scales. Ordinary computers use bits recorded as zeros or ones. Quantum devices rely on qubits, which can represent combinations of those states and tackle certain problems differently.

Under IBM’s definition, quantum advantage occurs when a quantum processor performs a task beyond what known classical computing methods can achieve alone and the outcome can still be rigorously validated.

That standard creates a practical difficulty, according to IBM researchers. If a conventional machine can easily repeat the work, no advantage has been shown. Once the task exceeds classical reach, however, the familiar means of checking it disappears.

The University of Chicago and IBM collaboration used a structured alternative to random circuit sampling, a benchmark involving patterns that become increasingly difficult for conventional computers to simulate.

The authors encoded a 70-qubit logical computation across 97 physical qubits, the hardware components that carry quantum information. The encoding was designed to detect errors that might otherwise disrupt the operation.

IBM and the university said the processor finished in roughly 15 minutes, while leading classical simulations would require impractical resources. The paper establishes a statistical lower bound on fidelity, a measure of how closely the machine produced the intended quantum state.

Qedma and IBM used as many as 74 qubits to model the behavior of a magnetic system exposed to regular pulses of energy. In the hardest portion, one leading classical method failed to converge, while another remained highly sensitive to where the computation was cut off, according to the article.

The collaborators examined the output with separate error-reduction techniques and repeated select portions on Quantinuum processors, which use a different hardware design. They said the agreement across methods and platforms strengthened the evidence behind the finding.

Algorithmiq used 56 qubits to explore how information spreads through quantum matter. After classical simulations generated conflicting predictions, the company altered processor calibrations and noise patterns and repeated the experiment across several IBM systems. Both results reported consistent estimates.

The Algorithmiq paper presents its quantum estimate as the most credible among the approaches considered, rather than as independently established ground truth.

IBM has reiterated that these are claims of quantum advantage and the next step is for the classical community to respond. The company has placed all three papers on its Quantum Advantage Tracker, where outside groups can test the claims against improved classical algorithms.

“Advances in verification have the potential to unlock practical applications for the next generation of quantum computers,” Soumik Ghosh, a doctoral student at the University of Chicago and co-author, said in the release published about the demonstration.