sema-evalsIndependent evaluations

A2A Drift — 20260722T081949541Z-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 8 4 0/4 4/4 0/4 0/8 4/8
advertised-voluntary 8 4 4/4 0/4 4/4 0/8 8/8
advertised-enforced 8 4 4/4 0/4 4/4 0/8 8/8

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.