omni-benchOMNI

Parity

Did the port change the transcripts? Each report diffs two scored runs whose identity tuples were held equal where it matters — the verdict is pass only when every per-language Δ stays inside its tolerance.

No published core 0.6 Parity Reports.