sema-evalsIndependent evaluations

x402 Contract Drift — 20260723T092012753Z-order-20260716

Exploratory paper-only model pilot, not confirmatory evidence and not facilitator/on-chain conformance evidence. Only the payer decision is model-driven. Seller, registry drift, verification, enforcement, in-process transport, simulated payload, and settlement remain deterministic. No wallet, signing, network, facilitator, or production write is available.

Condition Trials Drift trials Detected Silent payments Correct refusals False refusals Task successes Provider failures Malformed outputs
baseline 60 30 0/30 17/30 13/30 6/60 35/60 0/60 2/60
advertised-voluntary 60 30 30/30 0/30 24/30 6/60 48/60 0/60 1/60
advertised-enforced 60 30 30/30 0/30 30/30 7/60 50/60 0/60 3/60

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.