On July 30, 2026, IBM announced that three research papers demonstrated what it describes as reliable quantum progress, with quantum computers performing computations in regimes where classical simulation methods fail, while incorporating mechanisms to verify the quality of the results. The work addressed the verification of random circuit sampling, the observation of quantum phenomena in Floquet dynamics, and verification of the computation process when no reliable classical answer is available for comparison.
These results come amid an ongoing discussion over whether quantum computers have actually surpassed the best classical methods, rather than merely producing outputs that are difficult to verify. According to IBM, a quantum advantage claim does not end the comparison but instead opens the results to further testing through the Quantum Advantage Tracker, which follows quantum advantage candidates and compares them with available classical methods.
Verification Built Into Random Circuit Sampling
The paper co-authored by researchers from IBM and the University of Chicago focused on the problem of verifying Random Circuit Sampling, one of the leading methods for demonstrating the gap between quantum computing and classical simulation. Cross-Entropy Benchmarking requires calculating ideal output probabilities, a task that can become extremely costly with large circuits. This led previous experiments to rely on smaller or simpler circuits and then infer performance at sizes that cannot be examined directly.
The researchers proposed a structured alternative called doped Clifford sampling, in which classically intractable T gates are added to an efficiently simulable Clifford circuit while preserving a structure that helps detect errors. The circuit was embedded within a spacetime code that uses auxiliary qubits distributed across the qubits and evolves the circuit over time. In an experiment involving 70 logical qubits, selecting executions consistent with consistency checks produced an effective reduction in gate errors of approximately tenfold while maintaining practical execution rates.
This method makes it possible to derive a strict lower bound on the accuracy of the logical computation based on a reference circuit that can be verified classically, along with syndrome and logical-error information. Verification thus becomes part of the computational framework rather than relying solely on an external statistical metric.
Quantum Phenomena Beyond Leading-Edge Simulation
In another effort, Qedma, in collaboration with RIKEN and BlueQubit, studied Floquet dynamics, namely the response of an interacting quantum system to recurring energy pulses. Using circuits of up to 74 qubits, the researchers tracked the system's magnetization over time and observed persistent oscillations that did not appear in two advanced classical simulation methods using one of the world's largest supercomputers at RIKEN.
In the experiment's most challenging regimes, the two classical methods diverged and did not provide a consistent answer, while quantum computing continued to resolve the dynamics. The experiments used Qedma's QESEM software on IBM Quantum systems, along with error-mitigation methods, and portions of the experiment were repeated on the Quantinuum platform. IBM considered the agreement between independent estimators and results from the other platform to strengthen confidence that the observed behavior reflects the underlying physics.
Verifying the Process Rather Than the Answer
Algorithmiq developed an algorithm to estimate the Loschmidt echo operator, a quantity that tracks how information spreads in inhomogeneous quantum systems. The experiments involved 56 qubits and reached regimes in which at least three leading classical-simulation groups produced divergent predictions, which also differed from the quantum results.
Rather than relying on an unavailable classical answer, the researchers applied the same error-mitigation method across the equivalent of five quantum computers with different noise characteristics and found consistent results. They also showed that rigorous error mitigation can produce unbiased estimates and quantitative error margins when an accurate model of device noise is available.
Other candidates currently listed in the Quantum Advantage Tracker include work from Q-CTRL, BlueQubit, and the Birla Institute of Technology and Science, Pilani. IBM emphasizes that comparisons between quantum and classical computing will continue as these results undergo further testing.