Tag
This paper proposes BREAD, a scalable baseline-referenced explanation method for diagnosing anomalies in AI-based statistical process monitoring, with theoretical guarantees showing higher faithfulness than LIME under mean-shift anomalies.