Establish broader policy gold beyond state-based return predicates

Establish policy gold for informed-confirmation semantics, unsupported verbal claims, disclosure obligations, and external settlement, which are not determined by the captured state clauses used for the declared return-subgoal assessment.

Background

The historical retail assessment evaluates a narrowly defined return subgoal using captured state predicates, including the requested order status, returned item set, and refund-method identifier. The paper explicitly notes that these predicates do not establish several broader policy obligations, including whether informed confirmation was properly obtained, whether unsupported verbal claims were made, whether disclosure requirements were met, or whether external settlement occurred. The corresponding broader policy gold remains unknown, leaving unresolved the construction of authoritative labels for these obligations.

References

Informed confirmation semantics, unsupported verbal claims, disclosure, and external settlement are not established by these state clauses. Broader policy gold remains unknown.

— AgentEval: Evidence-Bounded Assessment of Stateful Agents  (2610.21177 - Gao, 15 Oct 2026) in Appendix: Historical Reference Construction and Assessment, paragraph “Unresolved obligations”