Reliably detecting model failures in deployment without labels

Rahul Krishnan (University of Toronto & Vector Institute) · Viet Nguyen (Johns Hopkins University) · Changjian Shui (Vector Institute) · Vijay Giri (University of Pennsylvania) · Siddharth Arya (University of Toronto) · Amol Verma (University of Toronto) · Fahad Razak (University of Toronto)

alert mechanismbenchmark datasetsdisagreement-based algorithmdynamic environments empirical results false positive rateshigh-stakes machine learninginternal medicine datasetmodel retrainingnon-deteriorating shiftspdd monitoringpost-deployment deteriorationpredictive models sample complexity boundstrue positive rates

The distribution of data changes over time; models operating in dynamic environments need retraining. But knowing when to retrain, without access to labels, is an open challenge since some, but not all shifts degrade model performance. This paper formalizes and addresses the problem of post-deployment deterioration (PDD) monitoring. We propose D3M, a practical and efficient monitoring algorithm based on the disagreement of predictive models, achieving low false positive rates under non-deteriorating shifts and provides sample complexity bounds for high true positive rates under deteriorating shifts. Empirical results on both standard benchmark and a real-world large-scale internal medicine dataset demonstrate the effectiveness of the framework and highlight its viability as an alert mechanism for high-stakes machine learning pipelines.