GlobEnc: Quantifying Global Token Attribution by Incorporating the Whole Encoder Layer in Transformers

Anonymous

Paper Link: https://openreview.net/forum?id=6d9vG4hBd9s

Paper Type: Long paper (up to eight pages of content + unlimited references and appendices)

Abstract: There has been a growing interest in interpreting the underlying dynamics of Transformers. While self-attention patterns were initially deemed as the primary option, recent studies have shown that integrating other components can yield more accurate explanations. This paper introduces a novel token attribution analysis method that incorporates all the components in the encoder block and aggregates this throughout layers. Through extensive quantitative and qualitative experiments, we demonstrate that our method can produce faithful and meaningful global token attributions. Our experiments reveal that incorporating almost every encoder component results in increasingly more accurate analysis in both local (single layer) and global (the whole model) settings. Our global attribution analysis significantly outperforms previous methods on various tasks regarding correlation with gradient-based saliency scores. Our code is freely available at https://github.com/mohsenfayyaz/GlobEnc.

Presentation Mode: This paper will be presented virtually

Virtual Presentation Timezone: UTC+4

Copyright Consent Signature (type Name Or NA If Not Transferrable): Ali Modarressi

Copyright Consent Name And Address: Iran University of Science & Technology, University St., Hengam St., Resalat Square, Tehran, Iran.