Rédiger un PREreview

Agentic artificial intelligence (AI) — multi-agent systems that combine large languagemodels with external tools and autonomous planning — are rapidly transitioning fromresearch laboratories into high-stakes domains. Our earlier “Basic” paper introduced afive-axis framework and proposed preliminary metrics such as goal drift and harm reductionbut did not provide an algorithmic instantiation or empirical evidence. This “Advanced”sequel fills that gap. First, we revisit recent benchmarks and industrial deploymentsto show that technical metrics still dominate evaluations: a systematic review of 84papers from 2023–2025 found that 83% report capability metrics while only 30% consider human-centred or economic axes. Second, we formalise an Adaptive Multi-Dimensional Monitoring (AMDM) algorithm that normalises heterogeneous metrics, applies per-axisexponentially weighted moving-average thresholds and performs joint anomaly detectionvia the Mahalanobis distance. Third, we conduct simulations and real-world experiments.AMDM cuts anomaly-detection latency from 12.3 s to 5.6 s on simulated goal drift and reducesfalse-positive rates from 4.5% to 0.9% compared with static thresholds. We present a comparison table and ROC/PR curves, and we reanalyse case studies to surface missingmetrics. Code, data and a reproducibility checklist accompany this paper to facilitate replication.