🏢 J.P. Morgan AI Research

This paper introduces a novel, label-free method for detecting harmful distribution shifts in machine learning models deployed in production environments, leveraging a proxy error derived from an erro…