Please use this identifier to cite or link to this item: doi:10.22028/D291-38099
Title: BERT Probe : A python package for probing attention based robustness evaluation of BERT models
Author(s): Khan, Shahrukh
Shahid, Mahnoor
Singh, Navdeeppal
Language: English
Title: Software Impacts
Volume: 13
Publisher/Platform: Elsevier
Year of Publication: 2022
Free key words: Deep learning
Adversarial machine learning
DDC notations: 004 Computer science, internet
Publikation type: Journal Article
Abstract: Transformer models based on attention-based architectures have been significantly successful in establishing state-of-the-art results in natural language processing (NLP). However, recent work about adversarial robustness of attention-based models show that their robustness is susceptible to adversarial inputs causing spurious outputs thereby raising questions about trustworthiness of such models. In this paper, we present BERT Probe which is a python-based package for evaluating robustness to attention attribution based on character-level and word-level evasion attacks and empirically quantifying potential vulnerabilities for sequence classification tasks. Additionally, BERT Probe also provides two out-of-the-box defenses against character-level attention attribution-based evasion attacks.
DOI of the first publication: 10.1016/j.simpa.2022.100310
URL of the first publication:
Link to this record: urn:nbn:de:bsz:291--ds-380999
ISSN: 2665-9638
Date of registration: 21-Nov-2022
Faculty: MI - Fakultät für Mathematik und Informatik
Department: MI - Informatik
Professorship: MI - Keiner Professur zugeordnet
Collections:SciDok - Der Wissenschaftsserver der Universität des Saarlandes

Files for this record:
File Description SizeFormat 
1-s2.0-S2665963822000471-main.pdf420,84 kBAdobe PDFView/Open

This item is licensed under a Creative Commons License Creative Commons