AttentionViz: A Global View of Transformer Attention

TL;DR

AttentionViz introduces a novel visualization method for transformer models that reveals global attention patterns by embedding query and key vectors, enhancing interpretability across language and vision tasks.

Contribution

We propose a new visualization technique based on joint query-key embeddings, enabling global analysis of attention mechanisms in transformers.

Findings

Facilitates understanding of transformer attention patterns

Provides insights into query-key interactions in language and vision models

Improves model interpretability through interactive visualization

Abstract

Transformer models are revolutionizing machine learning, but their inner workings remain mysterious. In this work, we present a new visualization technique designed to help researchers understand the self-attention mechanism in transformers that allows these models to learn rich, contextual relationships between elements of a sequence. The main idea behind our method is to visualize a joint embedding of the query and key vectors used by transformer models to compute attention. Unlike previous attention visualization techniques, our approach enables the analysis of global patterns across multiple input sequences. We create an interactive visualization tool, AttentionViz (demo: http://attentionviz.com), based on these joint query-key embeddings, and use it to study attention mechanisms in both language and vision transformers. We demonstrate the utility of our approach in improving model understanding and offering new insights about query-key interactions through several application scenarios and expert feedback.