QArena · Error correction
Submit
QEC decoder accuracy on fixed circuit-level memory experiments
A decoder decides whether error correction works on real circuits. Fixed shots with hidden labels let decoders be compared on identical circuit-level data.
Instance files
Each JSON file defines one instance of this problem. Instances differ by code and noise model. The verifier reads the same files, and the verifier bundle contains all of them together with an accepted example.
| File | Instance | Parameters | Size |
|---|---|---|---|
| surface9-si1000-v1-dev.json | Surface d = 9 (development) | debug instance, 20,000 shots with public labels; X and Z memory, 25 rounds, 2,000 detectors, 1 observable | 28.9 KB |
| bb72-bravyi-v1-dev.json | BB [[72,12,6]] (development) | debug instance, 20,000 shots with public labels; X and Z memory, 6 cycles, 504 detectors, 12 observables | 58.1 KB |
| bb144-bravyi-v1-dev.json | BB [[144,12,12]] (development) | debug instance, 20,000 shots with public labels; X and Z memory, 12 cycles, 1,872 detectors, 12 observables | 58.5 KB |
| color7-uniform-v1-dev.json | Color d = 7 superdense (development) | debug instance, 20,000 shots with public labels; X and Z memory, 21 rounds, 756 detectors, 1 observable | 28.7 KB |
| honeycomb8-em3-v1-dev.json | Honeycomb 8 × 12 (development) | debug instance, 20,000 shots with public labels; H and V memory, 24 rounds (72 sub-rounds), 1,184 detectors, 1 observable | 29.0 KB |
| surface9-si1000-v1.json | Surface d = 9, SI1000 p = 0.003 | 200,000 shots with private labels; X and Z memory, 25 rounds, 2,000 detectors, 1 observable | 2.4 KB |
| bb72-bravyi-v1.json | BB [[72,12,6]], p = 0.003 | 200,000 shots with private labels; X and Z memory, 6 cycles, 504 detectors, 12 observables | 5.5 KB |
| bb144-bravyi-v1.json | BB [[144,12,12]], p = 0.004 | 200,000 shots with private labels; X and Z memory, 12 cycles, 1,872 detectors, 12 observables | 6.0 KB |
| color7-uniform-v1.json | Color d = 7 superdense, p = 0.002 | 200,000 shots with private labels; X and Z memory, 21 rounds, 756 detectors, 1 observable | 2.2 KB |
| honeycomb8-em3-v1.json | Honeycomb 8 × 12, EM3 p = 0.01 | 200,000 shots with private labels; H and V memory, 24 rounds (72 sub-rounds), 1,184 detectors, 1 observable | 2.5 KB |
Detector data, circuits and error models
Each experiment has its frozen Stim circuit, the full detector error model (DEM), a decomposed DEM for matching decoders (surface and honeycomb), and the detection events in Stim's b8 format (ceil(D/8) bytes per shot, bit i = detector i, little-endian bit order). Development instances also have their labels. Ranked labels are never published. Every file's SHA-256 is in the instance's JSON file. Unranked hardware reference: Google Quantum AI's below-threshold surface-code data (google_105Q_surface_code_d3_d5_d7.zip, 5.7 GB, public labels; not mirrored here and not a ranked instance).
| Instance | Files |
|---|---|
| Surface d = 9 (development) surface9-si1000-v1-dev | X circuit 0.04 MB · X full DEM 0.35 MB · X decomposed DEM 0.40 MB · X detectors 2.50 MB · X labels 0.01 MB · Z circuit 0.04 MB · Z full DEM 0.35 MB · Z decomposed DEM 0.40 MB · Z detectors 2.50 MB · Z labels 0.01 MB |
| BB [[72,12,6]] (development) bb72-bravyi-v1-dev | X circuit 0.07 MB · X full DEM 1.17 MB · X detectors 0.63 MB · X labels 0.02 MB · Z circuit 0.07 MB · Z full DEM 1.29 MB · Z detectors 0.63 MB · Z labels 0.02 MB |
| BB [[144,12,12]] (development) bb144-bravyi-v1-dev | X circuit 0.28 MB · X full DEM 4.78 MB · X detectors 2.34 MB · X labels 0.02 MB · Z circuit 0.28 MB · Z full DEM 5.09 MB · Z detectors 2.34 MB · Z labels 0.02 MB |
| Color d = 7 superdense (development) color7-uniform-v1-dev | X circuit 0.02 MB · X full DEM 0.30 MB · X detectors 0.95 MB · X labels 0.01 MB · Z circuit 0.02 MB · Z full DEM 0.25 MB · Z detectors 0.95 MB · Z labels 0.01 MB |
| Honeycomb 8 × 12 (development) honeycomb8-em3-v1-dev | H circuit 0.36 MB · H full DEM 0.36 MB · H decomposed DEM 0.37 MB · H detectors 1.48 MB · H labels 0.01 MB · V circuit 0.36 MB · V full DEM 0.36 MB · V decomposed DEM 0.37 MB · V detectors 1.48 MB · V labels 0.01 MB |
| Surface d = 9, SI1000 p = 0.003 surface9-si1000-v1 | X circuit 0.04 MB · X full DEM 0.35 MB · X decomposed DEM 0.40 MB · X detectors 25.00 MB · Z circuit 0.04 MB · Z full DEM 0.35 MB · Z decomposed DEM 0.40 MB · Z detectors 25.00 MB |
| BB [[72,12,6]], p = 0.003 bb72-bravyi-v1 | X circuit 0.07 MB · X full DEM 1.17 MB · X detectors 6.30 MB · Z circuit 0.07 MB · Z full DEM 1.29 MB · Z detectors 6.30 MB |
| BB [[144,12,12]], p = 0.004 bb144-bravyi-v1 | X circuit 0.28 MB · X full DEM 4.78 MB · X detectors 23.40 MB · Z circuit 0.28 MB · Z full DEM 5.09 MB · Z detectors 23.40 MB |
| Color d = 7 superdense, p = 0.002 color7-uniform-v1 | X circuit 0.02 MB · X full DEM 0.30 MB · X detectors 9.50 MB · Z circuit 0.02 MB · Z full DEM 0.25 MB · Z detectors 9.50 MB |
| Honeycomb 8 × 12, EM3 p = 0.01 honeycomb8-em3-v1 | H circuit 0.36 MB · H full DEM 0.36 MB · H decomposed DEM 0.37 MB · H detectors 14.80 MB · V circuit 0.36 MB · V full DEM 0.36 MB · V decomposed DEM 0.37 MB · V detectors 14.80 MB |