Analyzing Feed-Forward Blocks in Transformers through the Lens of Attention Maps

Transformers are ubiquitous in wide tasks. Interpreting their internals is a pivotal goal. Nevertheless, their particular components, feed-forward (FF) blocks, have typically been less analyzed despite their substantial parameter amounts. We analyze the input contextualization effects of FF blocks b...

Full description

Saved in:
Bibliographic Details
Main Authors Kobayashi, Goro, Kuribayashi, Tatsuki, Yokoi, Sho, Inui, Kentaro
Format Journal Article
LanguageEnglish
Published 01.02.2023
Subjects
Computer Science - Computation and Language
Online AccessGet full text

Cover

More Information
Summary:Transformers are ubiquitous in wide tasks. Interpreting their internals is a pivotal goal. Nevertheless, their particular components, feed-forward (FF) blocks, have typically been less analyzed despite their substantial parameter amounts. We analyze the input contextualization effects of FF blocks by rendering them in the attention maps as a human-friendly visualization scheme. Our experiments with both masked- and causal-language models reveal that FF networks modify the input contextualization to emphasize specific types of linguistic compositions. In addition, FF and its surrounding components tend to cancel out each other's effects, suggesting potential redundancy in the processing of the Transformer layer.
DOI:10.48550/arxiv.2302.00456