How we measure our own noise

No cycle is open. Nothing on this page is a current measurement.

LanguageRungNoise (arm A)DepthGap
C#1 G1 measured24.4%[8.6–41.5]36n/a c1
Dart1 G1 measured11.9%[6.6–15.9]38n/a c1
Elixir1 G1 measured45.6%[6.0–68.0]44n/a c1
Erlang1 G1 measured11.9%[6.9–16.8]32n/a c1

Rung 1 machine judge · 2 + one internal senior rater · 3 + external third-party rater (κ ≥ 0.80). No language is at rung 3 yet.

What these numbers do not tell you

Reproduce our draw. Seed, sampler version and the full repo list — keeps and discards. GET /api/public/langtrain/holdout/1

Everything on this page is served by /api/public/langtrain/programme and is readable without an account.