XAI for Transformers: Better Explanations through Conservative Propagation
This addresses the need for more transparent AI in applications using Transformers, though it is incremental as it extends an existing method.
The paper tackled the problem of unreliable gradient-based explanations for Transformers by identifying Attention Heads and LayerNorm as causes and proposing a stable propagation method, achieving state-of-the-art explanation performance across models and datasets.
Transformers have become an important workhorse of machine learning, with numerous applications. This necessitates the development of reliable methods for increasing their transparency. Multiple interpretability methods, often based on gradient information, have been proposed. We show that the gradient in a Transformer reflects the function only locally, and thus fails to reliably identify the contribution of input features to the prediction. We identify Attention Heads and LayerNorm as main reasons for such unreliable explanations and propose a more stable way for propagation through these layers. Our proposal, which can be seen as a proper extension of the well-established LRP method to Transformers, is shown both theoretically and empirically to overcome the deficiency of a simple gradient-based approach, and achieves state-of-the-art explanation performance on a broad range of Transformer models and datasets.