Decoder benchmarks
Inspect logical failure rates, decoding time and the recorded comparison conditions.
Run and compare decoder cases
rsinter runs cases, merges outputs, and produces comparison plots. Smoke runs check wiring; checked full artifacts support only the recorded cases and environments.
These benchmark harnesses require a source checkout and reference Python packages. Follow the surface comparison setup or BB comparison setup, then run from the repository root. A smoke run is a small check that the workflow works; it does not establish a decoder ranking.
make surface-decoder-compare-smoke
make bb-circuit-bposd-compare-smokeA decoder predicts a logical flip from detector events and an error model. Actual logical outcomes are held out for scoring.
| Campaign | What it checks | Evidence |
|---|---|---|
| Surface decoder | End-to-end local wiring | Smoke and checked comparison |
| BB-circuit BP-OSD | Environment and dataset readiness | Readiness and checked comparison |
Surface-code and BP-OSD comparisons
The checked plots show only their named cases. Full reproduction details are available inside each result.
Loading decoder evidence.
Smoke and readiness checks
Open local smoke and readiness evidence. BP-OSD parity regression checks are also listed in the evidence guide.