Adaptive multi‑verifier systems are usually compared by endpoint quality‑cost gaps, even when the verifier catalog, availability, accounting, online filtration, or scorer change with the policy. We model verifier routing as a contract‑conditioned identification problem. The contract records request support, verifier catalog, realized availability, resource accounting, online filtration, and post‑trace scoring; a matched route differs only in the policy coordinate.
ROUTEAUDIT adds three measurable objects to this contract. A contract lattice averages coordinate increments over every admissible bridge order and reports the resulting attribution together with its path sensitivity. A policy‑independent response tape identifies paired sequential contrasts when adaptive policies reveal different observations. For incomplete matching, request‑level bounds use whichever potential outcome remains observed, yielding a sharp finite‑population interval.
The protocol commits paid observations and ledger events before the oracle joins and returns an attribution certificate for each comparison. On two held‑out raw‑tail caches, matched static SF+SA equals the cascade, assigning apparent gains of 0.1797 and 0.1250 to the verifier‑set edge. On 1,319 held‑out task requests, the learned model and RLVR report qualities of 0.9522 and 0.9553 versus 0.9484 for matched static; the RLVR‑static paired difference is +0.0068 with a request‑paired interval $[0.0015,0.0122]$ and a training‑seed‑by‑request hierarchical interval $[0.0006,0.0131]$.
Controlled attribution recovery yields a route mean absolute error of 0.0011 and an endpoint reconstruction error of 0.0004. Factorial, bridge‑order, and stochastic‑provider studies evaluate the certificate interface; RLVR supplies a learned‑policy stress test under the same identification contract.
Review