Quantum Noise Mapped in Record-Breaking 92 Qubits

Researchers at IBM Quantum and several universities have developed a new method for mapping noise in up to 92 qubits using gate set Pauli noise learning.

This approach, which involves learning a set of noise channels that are mathematically equivalent to the true but unidentifiable noise channels present in the hardware, has been shown to be effective in characterizing and mitigating noise across a complete gate set.

The team's method uses a self-consistent characterization of learnable parameters to eliminate inconsistencies arising from unlearnable aspects of quantum noise, even as the number of qubits increases.

This research addresses a long-standing issue with identifying the complete noise model within quantum systems and provides experimental evidence quantifying the magnitude of gauge-induced inconsistencies in practical error mitigation protocols.