Randomized benchmarking (RB)
The average error per gate, found by running random sequences of gates of growing length and measuring how fast the results decay.
When results can be compared
Variants (standard, interleaved, simultaneous, direct) give different numbers. Compare only the same variant on the same kind of gate.
Where it comes from
Academic research (Knill and colleagues; Magesan, Gambetta and Emerson).
- Type
- Benchmark, component level
- Unit
- error per gate
- Better is
- Lower
- Applies to
- gate-based machines
Results
| Machine | Value | Set-up | Published | Measured by | Grade | Record |
|---|---|---|---|---|---|---|
| Google Willow | 3.5 × 10⁻⁴ | single qubit mean simultaneous | 9 Dec 2024 | Maker | B | r-000007 |