Explainable AI: Context-Aware Layer-Wise Integrated Gradients for Explaining Transformer Models
This addresses the interpretability challenge for users of Transformer models across domains like NLP and vision, though it appears incremental as it builds on existing attribution methods.
The paper tackles the problem of interpreting Transformer models by proposing a unified hierarchical attribution framework called Context-Aware Layer-wise Integrated Gradients (CA-LIG), which computes layer-wise Integrated Gradients and fuses them with attention gradients to capture context-sensitive attributions across layers. The result shows that CA-LIG provides more faithful attributions, stronger sensitivity to contextual dependencies, and clearer visualizations than existing methods across diverse tasks and architectures.
Transformer models achieve state-of-the-art performance across domains and tasks, yet their deeply layered representations make their predictions difficult to interpret. Existing explainability methods rely on final-layer attributions, capture either local token-level attributions or global attention patterns without unification, and lack context-awareness of inter-token dependencies and structural components. They also fail to capture how relevance evolves across layers and how structural components shape decision-making. To address these limitations, we proposed the \textbf{Context-Aware Layer-wise Integrated Gradients (CA-LIG) Framework}, a unified hierarchical attribution framework that computes layer-wise Integrated Gradients within each Transformer block and fuses these token-level attributions with class-specific attention gradients. This integration yields signed, context-sensitive attribution maps that capture supportive and opposing evidence while tracing the hierarchical flow of relevance through the Transformer layers. We evaluate the CA-LIG Framework across diverse tasks, domains, and transformer model families, including sentiment analysis and long and multi-class document classification with BERT, hate speech detection in a low-resource language setting with XLM-R and AfroLM, and image classification with Masked Autoencoder vision Transformer model. Across all tasks and architectures, CA-LIG provides more faithful attributions, shows stronger sensitivity to contextual dependencies, and produces clearer, more semantically coherent visualizations than established explainability methods. These results indicate that CA-LIG provides a more comprehensive, context-aware, and reliable explanation of Transformer decision-making, advancing both the practical interpretability and conceptual understanding of deep neural models.