Skip to content
RESEARCH · ERROR CORRECTION

Q-CTRL moves high-fidelity magic states into a surface code

The reported storage result keeps about one run in twenty.

Published Updated Source report Sources checked
Reading tools for text size, copying, printing and your reading list
My reading list
Silver IBM Quantum System Two enclosure behind glass at Hyde Park Labs in Chicago.
IBM Quantum System Two in Chicago. Contextual photograph, not the identified experimental processor.Illustrative photoIBM

Q-CTRL researchers report moving a carefully prepared “magic state” into a surface error-correcting code on IBM’s Nighthawk hardware. These states supply operations needed for universal fault-tolerant computing. The November 17 preprint was highlighted by the company on November 18.[1][2]

Before transfer, the team measured a logical error probability of 0.002%, retaining 25% of experimental runs. After transfer by lattice surgery and ten rounds, fidelity was about 99.1%, but only 4.6% of runs remained after filtering. The preparation and storage numbers describe different stages.[1]

The price of keeping the best runs

Imagine repeating the ten-round experiment at the published retention rate. A retained outcome would require roughly 22 attempts on average (1 ÷ 0.046). This is arithmetic from the reported yield. Fidelity describes how closely the retained states resemble the intended state. It gives no count of individually certified correct runs.[1]

Which runs reached the fidelity estimate?

Runs retained after ten rounds
4.6% of attempts
Fidelity among retained runs
99.1% ± 0.2 percentage points

The first bar measures how often a run passed selection. The second measures how closely the selected states matched the target after lattice-surgery transfer and ten stabilizer rounds. Each bar uses its own denominator.

Read the preprint’s storage result

These tests used offline decoding. A computer that consumes a state during a calculation also has to decode the measurements and act on them in time. Without a timed end-to-end demonstration, the paper cannot tell a user how many usable states per second would reach an algorithm.[1]

Keep your reading trail

Evidence & context

2 linked sources

Preprint and company report

How to read the evidence and dates

An announcement records what an organization stated. An experiment reports an observation. A simulation or resource estimate depends on its assumptions. QubitWire’s analysis connects these sources; the labels are not quality scores.

Published is the article’s first QubitWire publication. Updated marks a substantive content change. Sources checked records a review of the underlying evidence. An older paper can support a new explanation, and another press release about the same work is not independent replication.

For material corrections, check the article’s context note and source trail. A routine formatting or link change does not create a new scientific result.

Sources

  1. Research preprint
  2. Q-CTRL technical report
READ NEXT

Continue exploring

All Research stories
  1. More on error correction

    A shared metal layer could link errors across a quantum chip

    Why checking individual qubits is not the same as checking their relationships.

    Published · Source report · Sources checked · About 2 min read
  2. More on error correction

    AI Can Filter Bad Quantum Runs. The Rejected Runs Still Count.

    A new preprint uses neural networks to flag likely decoding failures. Large error reductions apply to retained simulated results, with a cost in discarded attempts.

    Published · Source report · Sources checked · About 2 min read