MEA: A Reward-Driven Multi-Agent System for Faithful Model Explanations

Recent years have seen the employment of a plethora of machine learning (ML) models in high-stakes domains, but they remain largely opaque to the practitioners who act on their predictions.