The Explainability of Transformers: Current Status and Directions
Computers · 04.04.2024 · 13 (4) · 92
DOI: 10.3390/computers13040092 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computers, online erschienen am 04.04.2024. Verfasst von Paolo Fantozzi und 1 weiteren Autor:innen. Die Arbeit wurde bisher 60-mal zitiert (Crossref) und nennt 169 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: European Union–Next Generation EU within the framework of PRIN 2022 Project “MEDICINE+AI, Law and Ethics for an Augmented and Human-Centered Medicine” (2022YB8….
Abstract (Originalsprache)
An increasing demand for model explainability has accompanied the widespread adoption of transformers in various fields of applications. In this paper, we conduct a survey of the existing literature on the explainability of transformers. We provide a taxonomy of methods based on the combination of transformer components that are leveraged to arrive at the explanation. For each method, we describe its mechanism and survey its applications. We find out that attention-based methods, both alone and in conjunction with activation-based and gradient-based methods, are the most employed ones. A growing attention is also devoted to the deployment of visualization techniques to help the explanation process.
Förderung
- European Union–Next Generation EU within the framework of PRIN 2022 Project “MEDICINE+AI, Law and Ethics for an Augmented and Human-Centered Medicine” (2022YB8…
Zitieren
APA
Fantozzi, P., & Naldi, M. (2024). The Explainability of Transformers: Current Status and Directions. Computers, 13(4), 92. https://doi.org/10.3390/computers13040092
Vancouver
Fantozzi P, Naldi M. The Explainability of Transformers: Current Status and Directions. Computers. 2024;13(4):92. doi:10.3390/computers13040092
BibTeX
@article{Fantozzi2024,
author = {Fantozzi, Paolo and Naldi, Maurizio},
title = {The Explainability of Transformers: Current Status and Directions},
journal = {Computers},
year = {2024},
volume = {13},
number = {4},
pages = {92},
doi = {10.3390/computers13040092}
}
EUVIMED