Representational Difference Explanations
This addresses the gap in interpretable model comparison tools for machine learning researchers and practitioners, though it appears incremental as it builds on existing XAI methods.
The authors tackled the problem of comparing learned representations in machine learning models by proposing Representational Differences Explanations (RDX), a method that discovers and visualizes differences, and demonstrated its effectiveness on ImageNet and iNaturalist datasets where existing XAI techniques fail.
We propose a method for discovering and visualizing the differences between two learned representations, enabling more direct and interpretable model comparisons. We validate our method, which we call Representational Differences Explanations (RDX), by using it to compare models with known conceptual differences and demonstrate that it recovers meaningful distinctions where existing explainable AI (XAI) techniques fail. Applied to state-of-the-art models on challenging subsets of the ImageNet and iNaturalist datasets, RDX reveals both insightful representational differences and subtle patterns in the data. Although comparison is a cornerstone of scientific analysis, current tools in machine learning, namely post hoc XAI methods, struggle to support model comparison effectively. Our work addresses this gap by introducing an effective and explainable tool for contrasting model representations.