A Backdoor-based Explainable AI Benchmark for Improved Fidelity in Evaluating Attribution Methods

A backdoor-based explainable AI benchmark is proposed for improved fidelity in evaluating attribution methods.