There is More to Attention: Statistical Filtering Enhances Explanations in Vision Transformers
This work addresses the challenge of generating faithful explanations for Vision Transformers, which is crucial for interpretability in AI applications, though it is incremental as it adapts an existing filtering technique to ViTs.
The authors tackled the problem of noisy attention maps in Vision Transformers (ViTs) for explainable AI by proposing a method that combines attention with statistical filtering, resulting in sharper and more interpretable explanations that outperform or match state-of-the-art methods across multiple datasets.
Explainable AI (XAI) has become increasingly important with the rise of large transformer models, yet many explanation methods designed for CNNs transfer poorly to Vision Transformers (ViTs). Existing ViT explanations often rely on attention weights, which tend to yield noisy maps as they capture token-to-token interactions within each layer.While attribution methods incorporating MLP blocks have been proposed, we argue that attention remains a valuable and interpretable signal when properly filtered. We propose a method that combines attention maps with a statistical filtering, initially proposed for CNNs, to remove noisy or uninformative patterns and produce more faithful explanations. We further extend our approach with a class-specific variant that yields discriminative explanations. Evaluation against popular state-of-the-art methods demonstrates that our approach produces sharper and more interpretable maps. In addition to perturbation-based faithfulness metrics, we incorporate human gaze data to assess alignment with human perception, arguing that human interpretability remains essential for XAI. Across multiple datasets, our approach consistently outperforms or is comparable to the SOTA methods while remaining efficient and human plausible.