Traveling Words: A Geometric Interpretation of Transformers
This provides an intuitive understanding of transformers for researchers and practitioners in NLP, though it is incremental as it builds on prior observations without new performance gains.
The paper tackles the challenge of understanding transformer mechanisms by introducing a geometric interpretation that shows how layer normalization confines features to a hyper-sphere and attention shapes word representations on it, validated by probing a 124M parameter GPT-2 model to reveal attention patterns in early layers and subject-specific heads in deeper layers.
Transformers have significantly advanced the field of natural language processing, but comprehending their internal mechanisms remains a challenge. In this paper, we introduce a novel geometric perspective that elucidates the inner mechanisms of transformer operations. Our primary contribution is illustrating how layer normalization confines the latent features to a hyper-sphere, subsequently enabling attention to mold the semantic representation of words on this surface. This geometric viewpoint seamlessly connects established properties such as iterative refinement and contextual embeddings. We validate our insights by probing a pre-trained 124M parameter GPT-2 model. Our findings reveal clear query-key attention patterns in early layers and build upon prior observations regarding the subject-specific nature of attention heads at deeper layers. Harnessing these geometric insights, we present an intuitive understanding of transformers, depicting them as processes that model the trajectory of word particles along the hyper-sphere.