sema-evalsIndependent evaluations

A2A Drift — 20260723T072703363Z-order-20260714

Exploratory model pilot. Not preregistered, not confirmatory evidence. Worker is model-driven; requester, transport, registries, drift injection, and middleware remain deterministic. Ground-truth driftDetected is middleware-only (ADR 0015).

Condition Trials Drift trials Detected Silent executions Correct halts False halts Task successes
baseline 40 20 0/20 20/20 0/20 2/40 18/40
advertised-voluntary 40 20 20/20 0/20 18/20 0/40 38/40
advertised-enforced 40 20 20/20 0/20 20/20 0/40 37/40

Drift-scoped columns use drift-injected trials as the denominator; the rest use all trials in the condition. Every count is recomputed from trials.public.jsonl at build time.