Search papers, labs, and topics across Lattice.
This paper introduces SARC-DQ, a metadata-aware pre-action gating mechanism designed to address the issue of metadata-borne defects in agentic AI systems, which can lead to costly erroneous actions. The authors demonstrate that competent agents frequently fail to detect these defects, acting on stale or superseded information without any data-quality flags, resulting in a 60% error rate in a priced replenishment benchmark. By implementing a downstream-only remediation approach, the proposed method fully recovers losses associated with detectable defects while revealing the critical distinction between evidence integrity and model capability.
Metadata defects can silently cost agentic AI systems 60% in erroneous actions, but a new gating mechanism can fully recover these losses.
Agentic systems act, so a defect in the evidence they retrieve becomes a wrong action with a currency cost. The most dangerous enterprise defects are metadata-borne: a stale price or a superseded record, perfectly well-formed in the payload and betrayed only by freshness, lineage, or provenance. Such a defect never enters the agent's context, and an agent cannot doubt data it cannot see. On a priced replenishment benchmark, a competent agent silently converts an injected metadata-borne defect into a costly action about 60% of the time, with zero data-quality flags and behavioral doubt markers at chance (AUC<= 0.50). Across four model tiers spanning roughly 15x in inference price, the rate stays flat: capability does not buy skepticism. A metadata-aware pre-action gate with downstream-only remediation recovers the loss fully on the signals its predicates cover and not at all on those they miss. A model-free oracle derived from the task's decision geometry tracks the measured rates with MAE 0.015 (Pearson r = 0.876, interval coverage 15/16 cells), giving the flat ladder an analytical form. Evidence integrity is a systems axis distinct from model capability; mitigation depends on enforcement placement and predicate coverage. Code, frozen results, and a deterministic analysis pipeline: https://github.com/besanson/dqSarc
