Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

DecompX: Explaining transformers decisions by propagating token decomposition

Modarressi, Ali, Fayyaz, Mohsen, Aghazadeh, Ehsan, Yaghoobzadeh, Yadollah and Pilehvar, Mohammad Taher 2023. DecompX: Explaining transformers decisions by propagating token decomposition. Presented at: 61st Annual Meeting of the Association for Computational Linguistics, Toronto, Canada, 9-14 July 2023. Published in: Rogers, Anna, Boyd-Graber, Jordan and Okazaki, Naoaki eds. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Association for Computational Linguistics, 2649–2664. 10.18653/v1/2023.acl-long.149

[thumbnail of 2023.acl-long.149.pdf] PDF - Published Version
Available under License Creative Commons Attribution.

Download (980kB)

Abstract

An emerging solution for explaining Transformer-based models is to use vector-based analysis on how the representations are formed. However, providing a faithful vector-based explanation for a multi-layer model could be challenging in three aspects: (1) Incorporating all components into the analysis, (2) Aggregating the layer dynamics to determine the information flow and mixture throughout the entire model, and (3) Identifying the connection between the vector-based analysis and the model’s predictions. In this paper, we present DecompX to tackle these challenges. DecompX is based on the construction of decomposed token representations and their successive propagation throughout the model without mixing them in between layers. Additionally, our proposal provides multiple advantages over existing solutions for its inclusion of all encoder components (especially nonlinear feed-forward networks) and the classification head. The former allows acquiring precise vectors while the latter transforms the decomposition into meaningful prediction-based values, eliminating the need for norm- or summation-based vector aggregation. According to the standard faithfulness evaluations, DecompX consistently outperforms existing gradient-based and vector-based approaches on various datasets. Our code is available at https://github.com/mohsenfayyaz/DecompX.

Item Type: Conference or Workshop Item (Paper)
Status: Published
Schools: Business (Including Economics)
Publisher: Association for Computational Linguistics
Date of First Compliant Deposit: 10 December 2024
Last Modified: 10 Dec 2024 17:00
URI: https://orca.cardiff.ac.uk/id/eprint/174557

Actions (repository staff only)

Edit Item Edit Item

Downloads

Downloads per month over past year

Loading...

View more statistics