Does it let a wrong result through?
wrong runs accepted.
Good runs rejected: 0 of 8.
31 real OpenFOAM runs · 95% interval 0–14%
Does it catch
the wrong
problem?
20 of 20
wrong problems rejected.
Right problems rejected: 0 of 7.
27 recorded OpenFOAM runs · 95% interval 84–100%
Can I see why?
31 of 31
verdicts re-derived from their evidence, byte for byte.
95% interval 89–100% · BENCHMARKS §7 · 11 Oct 2026
icoFoam_cavity::cfl_blowup
icoFoam · ΔT 0.05 s · 0.5 s
Reason 1 of 3 · cfl.max · critical
Courant number exceeds stability limit: max Co=9.66 at t=0.15 exceeds limit 1.0; still 8.92 over the final fifth of the run
Read
icoFoam_cavity__cfl_blowup.log.gz (t = 0.15)
VER-5.3 · grade C
reject
Not checked · 8
- Solution fields were not read
- Discretisation error was not quantified
- Accuracy against a reference was not assessed
- Field values physically plausible
- Flow regime matches turbulence modeling
- Near-wall y+ compatible with wall treatment
- Net boundary mass flux ~ 0
- No specification given
Open the full report report.html · 10,506 bytes
Open the accepted run's report report.html · 12,855 bytes
Every number, with its source20
Certified here means measured by fault injection on real solver runs, with the error rate published. Without a specification it decides whether a run is valid; with one, whether it is the problem that was asked.
| Number | What it counts | Interval | Source |
|---|---|---|---|
| Valid? · 31 real runs | |||
| 31 of 31 | real runs judged correctly. | 95% CI 89–100% | BENCHMARKS §1 · 11 Oct 2026 |
| 0 of 23 | wrong results accepted. | 95% CI 0–14% | BENCHMARKS §1 · 11 Oct 2026 |
| 0 of 8 | good results rejected. | 95% CI 0–32% | BENCHMARKS §1 · 11 Oct 2026 |
| 13 of 23 | wrong results the exit code alone accepts. | 95% CI 37–74% | RESULTS.json exit codes · 11 Oct 2026 |
| 3 of 3 | held-out runs judged correctly. | 95% CI 44–100% | BENCHMARKS §1 · 11 Oct 2026 |
| 23 of 23 | faults named among the top two causes. | 95% CI 86–100% | BENCHMARKS §2 · 11 Oct 2026 |
| 8 of 23 | broken setups caught before the solver started. | 95% CI 19–55% | BENCHMARKS §3 · 11 Oct 2026 |
| 0 of 8 | good setups flagged before the solver started. | 95% CI 0–32% | BENCHMARKS §3 · 11 Oct 2026 |
| 0.032 s | median to verify a run. | BENCHMARKS §6 · 11 Oct 2026 | |
| 31 of 31 | evidence packs re-derived byte for byte. | 95% CI 89–100% | BENCHMARKS §7 · 11 Oct 2026 |
| 31 of 31 | verdicts identical on a second run. | 95% CI 89–100% | BENCHMARKS §7 · 11 Oct 2026 |
| The problem that was asked? · 20 + 7 runs | |||
| 8 of 20 | wrong-problem runs caught without the task. | 95% CI 22–61% | BENCHMARKS §3b · 3 Oct 2026 |
| 20 of 20 | wrong-problem runs caught against the task. | 95% CI 84–100% | BENCHMARKS §3b · 3 Oct 2026 |
| 20 of 20 | caught against the drafted specification, unedited. | 95% CI 84–100% | BENCHMARKS §3b · 3 Oct 2026 |
| 0 of 7 | false alarms on the untouched tutorials. | 95% CI 0–35% | BENCHMARKS §3b · 3 Oct 2026 |
| SilentBench · 28 dev tasks | |||
| 0 of 28 | copied tutorials accepted. | 95% CI 0–12% | BENCHMARKS §3c · 3 Oct 2026 |
| 24 of 28 | accepted with only the named values set. | 95% CI 69–94% | BENCHMARKS §3c · 3 Oct 2026 |
| 28 of 28 | reference answers accepted. | 95% CI 88–100% | BENCHMARKS §3c · 3 Oct 2026 |
| Repeatable? · 36 scenarios, 69 artefacts | |||
| 36 of 36 | certification scenarios pass on a real solver. | 95% CI 90–100% | certification/RESULTS.json · 12 Oct 2026 · 12 Oct 2026 |
| 69 of 69 | artefacts byte-identical across platforms. | 95% CI 95–100% | RM-34 · 11 Oct 2026 · courant 0.4.0 · 11 Oct 2026 |
Modelled, not measured
One number on this site is a model, not a measurement: the expected saving EXPECTED.
The model