How to Dissect a Muppet: The Structure of Transformer Embedding Spaces
Published
2022-09-07
Timothee Mickus
,
Denis Paperno
,
Mathieu Constant
Timothee Mickus
ATILF, Université de Lorraine, CNRS
Denis Paperno
Utrecht Universiteit
Mathieu Constant
ATILF, Université de Lorraine, CNRS
Abstract
Pretrained embeddings based on the Transformer architecture have taken the NLP community by storm. We show that they can mathematically be reframed as a sum of vector factors and showcase how to use this reframing to study the impact of each component. We provide evidence that multi-head attentions and feed-forwards are not equally useful in all downstream applications, as well as a quantitative overview of the effects of finetuning on the overall embedding space. This approach allows us to draw connections to a wide range of previous studies, from vector space anisotropy to attention weights.
Article at MIT Press
Presented at EACL 2023