$2 / month

Comment and reply across TML.

Start or manage membership

Most pulsed

Latest from other desks

tml.the modern leap
Tech / Future Tech

Quantum X Labs Cuts Decoding Failures on IBM Processors

The AI-powered error-correction program now benchmarks across IBM Fez, Kingston and Pittsburgh: up to 20% fewer logical decoding failures in preliminary tests, by the company's release.

TML Tech DeskNov 11, 9:27 PM ET · 1 min read

Edited and fact-checked by Stephen Sprague

Technicians install an IBM Quantum System One; Quantum X Labs’ error-correction advance is built on IBM quantum hardware. Photo: IBM Research / CC BY 2.0, via Wikimedia Commons.
Summary

Quantum X Labs reports up to 20% fewer logical decoding failures in preliminary quantum error-correction benchmarks run across three IBM processors.

Tap a passage. React to the part that matters.

Quantum X Labs says its AI decoder produced up to 20 percent fewer logical failures in preliminary tests on IBM quantum processors.

The company expanded its error-correction benchmarking to experimental data from IBM's Fez, Kingston and Pittsburgh processors, calibrating noise models per machine before training its decoder. The program builds on earlier work using datasets generated by Google's quantum hardware.

Error correction is the field's bottleneck: physical qubits fail often enough that useful machines need many of them combined into logical qubits, and a decoder has to spot and fix errors as they occur, because a logical qubit that isn't corrected promptly stops carrying information. Quantum X Labs' approach is hardware-agnostic on paper: characterize each processor's noise, calibrate the model, then train the decoder with GPU acceleration and measure logical decoding failures against the uncalibrated baseline. The workflow it describes runs measurement, calibration, training and evaluation as one pipeline across the three IBM datasets. Its earlier research used surface-code datasets associated with Google's quantum program, so the IBM expansion is the cross-vendor test of whether the calibration method travels.

The caveats are in the company's own wording: the 20 percent figure is preliminary, from benchmarks on experimental data, and no peer-reviewed paper accompanies the release. The announcement came through Gix Internet, the Tel Aviv-listed company that holds about 13 percent of Quantum X Labs (2,818,585 shares), which marks it as an investor-facing milestone as much as a scientific one. What would settle the claim is the standard the field applies to every decoder: published logical error rates per round, across more codes and more machines, reproduced by someone else. Until then the honest summary is narrow: a decoder calibrated to each machine's noise failed less often in the company's own preliminary tests, on three of IBM's processors.

Sources

The conversation

A little more perspective.

Reading is free. Verified members can comment and reply for $2 a month.

The conversation.

Loading comments…

Keep your curiosity going.

Some links here are affiliate links. If you buy through one, we might make a commission — it never changes what you pay. As an Amazon Associate I earn from qualifying purchases.