Quantum computing for much of its history has had at least one problem that has been challenging to solve — it's simply not always accurate.

In new breakthrough research published today by IBM, conducted in partnership with a group at the University of California Berkeley, the group is claiming a significant breakthrough with the most accurate quantum circuits yet, on systems using quantum processors with over 100 qubits.

A qubit is the unit of measurement for a quantum system designed to give some measure of the computing power. IBM's most powerful production quantum system today uses a 127-qubit processor codenamed Eagle. IBM has laid out a roadmap to reach 433 qubits at some point in the future.

"With applications that we're looking at — whether it's ground-state simulations, kernel methods, quantum machine learning or different molecular interactions — you're trying to look at getting the accurate value out of the quantum state," Katie Pizzolato, director, IBM quantum strategy and applications research, told SDxCentral. "Getting the accurate value at scale is difficult because we have a lot of inherent noise in the system."

Sarah Sheldon, senior manager, quantum theory and capabilities at IBM Quantum, explained to SDxCentral that quantum noise is really anything that affects operations that are being performed on quantum computers.

That noise can be anything that causes information loss and can be due to decoherence, material defects and other disruptions in the environment. What IBM's research has done is figure out how to do error mitigation for the noise, which improves the overall accuracy of the system at scale.

Pizzolato commented that the idea of error correction is common in classical silicon computers and helps make them useful. IBM has long held the view that error correction was also possible with quantum computing, though full error correction likely requires a very large number of qubits.

"What we're seeing now is that we can already get good results out of a quantum computer without having to go to an error-corrected framework with a huge qubit overhead," Pizzolato explained. "But we still have to deal with the noise in our systems somehow and that's where these error-mitigation methods come from; where instead of using a huge qubit overhead, we are able to extract good results by combining the outputs of many noisy experiments and essentially doing some classical post-processing."

As an analogy, Pizzolato said the error-mitigation techniques can be thought of as being similar in basic concept to how noise-canceling headphones work. With noise-canceling headphones, additional signals are dropped on top of the raw signal to cancel out the effect of noise.

"In our case, we're learning what noise exists on our device, and we're implementing some additional operations, so that when we look at many instances of measuring these circuits, on average, we get a result that has mitigated noise value," she said.

With more accurate results possible at large scale, Pizzolato said the path to quantum advantage will become clearer.

The concept of quantum advantage is about clearly identifying the right places where quantum computing will offer better outcomes than classical computers. Pizzolato added that IBM's view is that thanks to the new error-mitigation methods, organizations will be able to evaluate accurate values at scale.

"Quantum is not going to be good for everything," Pizzolato said. "We have to find new circuits that exist in places that can challenge classical methods and then we have to find really cool things to do with them."