mlsanityData to 14 Oct 2026

Robustness of Explanation Methods for NLP Models

arXiv 2206.12284 ↗First posted 24 Jun 2022cs.CL cs.AIcited by 3

Not on the map. Why?

Abstract (arXiv)

Explanation methods have emerged as an important tool to highlight the features responsible for the predictions of neural networks. There is mounting evidence that many explanation methods are rather unreliable and susceptible to malicious manipulations. In this paper, we particularly aim to understand the robustness of explanation methods in the context of text modality. We provide initial insights and results towards devising a successful adversarial attack against text explanations. To our knowledge, this is the first attempt to evaluate the adversarial robustness of an explanation method. Our experiments show the explanation method can be largely disturbed for up to 86% of the tested samples with small changes in the input sentence and its semantics.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 18 of its 22 references that are in the map. First posted on arXiv 24 Jun 2022.

Builds on

18 of 22 references.

References · 18

  1. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
  2. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,170
  3. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,415
  4. 2018GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understandingcited by 4,899
  5. 2017Axiomatic Attribution for Deep Networkscited by 3,384
  6. 2018Sanity Checks for Saliency Mapscited by 1,156
  7. 2017SmoothGrad: removing noise by adding noisecited by 1,109
  8. 2017Interpretation of Neural Networks is Fragilecited by 444
  9. 2016European Union regulations on algorithmic decision-making and a "right to explanation"cited by 344
  10. 2017The (Un)reliability of saliency methodscited by 333
  11. 2018On the Robustness of Interpretability Methodscited by 314
  12. 2019Explanations can be manipulated and geometry is to blamecited by 216
6 more
  1. 2016Counter-fitting Word Vectors to Linguistic Constraintscited by 140
  2. 2021Do Feature Attribution Methods Correctly Attribute Features?cited by 98
  3. 2019On the Connection Between Adversarial Robustness and Saliency Map Interpretabilitycited by 80
  4. 2018Interpretable Deep Learning under Firecited by 74
  5. 2020Towards Robust Explanations for Deep Neural Networkscited by 43
  6. 2021A Survey on the Robustness of Feature Importance and Counterfactual Explanationscited by 39

Built on by

Cited by year

Citing papers per year: 2022: 1; 2023: 2.

3 papers.

  1. 2023Why do universal adversarial attacks work on large language models?: Geometry might be the answercited by 14
  2. 2022A.I. Robustness: a Human-Centered Perspective on Technological Challenges and Opportunitiescited by 9
  3. 2023DARE: Towards Robust Text Explanations in Biomedical and Healthcare Applicationscited by 1