Gradient-based Analysis of NLP Models is Manipulable
Open accessThe DOI matches its Crossref record; 18 of the 18 references it cites were found in a registry or an index.
Abstract
Gradient-based analysis methods, such as saliency map visualizations and adversarial input perturbations, have found widespread use in interpreting neural NLP models due to their simplicity, flexibility, and most importantly, their faithfulness.In this paper, however, we demonstrate that the gradients of a model are easily manipulable, and thus bring into question the reliability of gradient-based analyses.In particular, we merge the layers of a target model with a FACADE model that overwhelms the gradients without affecting the predictions.This FACADE model can be trained to have gradients that are misleading and irrelevant to the task, such as focusing only on the stop words in the input.On a variety of NLP tasks (text classification, NLI, and QA), we show that our method can manipulate numerous gradient-based analysis techniques: saliency maps, input reduction, and adversarial perturbations all identify unimportant or targeted tokens as being highly important.The code and a tutorial of this paper is available at http://ucinlp.github.io/facade.Input Prediction P(+)=0.9How !great Gradient How great !(a) Original Model, forig Prediction P(+)=0.5 Gradient How great !Input How !great (b) FACADE Model, g 1024 Input Prediction P(+)=0.9Gradient How great !
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325
“Why Should I Trust You?”: Explaining the Predictions of Any Classifier
Ribeiro, M., Singh, S., & Guestrin, C. (2016) · Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations · doi:10.18653/v1/n16-3020 · Cited by 6,682
Self-citationA large annotated corpus for learning natural language inference
Bowman, S. R., Angeli, G., Potts, C., & Manning, C. D. (2015) · Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d15-1075 · Cited by 3,377
Towards A Rigorous Science of Interpretable Machine Learning
Doshi-Velez, F., & Kim, B. (2017) · arXiv · doi:10.48550/arxiv.1702.08608 · Cited by 3,121
Anchors: High-Precision Model-Agnostic Explanations
Ribeiro, M. T., Singh, S., & Guestrin, C. (2018) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v32i1.11491 · Cited by 2,162 · cited 2×
“Furthermore, non-gradient-based approaches, e.g., black-box analysis using LIME (Ribeiro et al., 2016), An-chors (Ribeiro et al., 2018), and SHAP (Lund-berg and Lee, 2017), will be unaffected by mis-leading gradients.”
BackgroundMethodSelf-citationSmoothGrad: removing noise by adding noise
Smilkov, D., Thorat, N., Kim, B., Viégas, F., & Wattenberg, M. (2017) · arXiv · doi:10.48550/arxiv.1706.03825 · Cited by 1,467
Adversarial Examples for Evaluating Reading Comprehension Systems
Jia, R., & Liang, P. (2017) · Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d17-1215 · Cited by 1,410
Sanity Checks for Saliency Maps
Adebayo, J., Gilmer, J., Muelly, M., Goodfellow, I., Hardt, M., & Kim, B. (2018) · arXiv · doi:10.48550/arxiv.1810.03292 · Cited by 1,294
Annotation Artifacts in Natural Language Inference Data
Gururangan, S., Swayamdipta, S., Levy, O., Schwartz, R., Bowman, S., & Smith, N. A. (2018) · Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers) · doi:10.18653/v1/n18-2017 · Cited by 907
Cited by
Explainable Artificial Intelligence Applications in Cyber Security: State-of-the-Art in Research
Zhang, Z., Hamadi, H. A., Damiani, E., Yeun, C. Y., & Taher, F. (2022) · IEEE Access · doi:10.1109/access.2022.3204051 · Cited by 450
How Reliable are Model Diagnostics?
Aribandi, V., Tay, Y., & Metzler, D. (2021) · Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021 · doi:10.18653/v1/2021.findings-acl.155 · Cited by 15
Rationales for Sequential Predictions
Vafa, K., Deng, Y., Blei, D., & Rush, A. (2021) · Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/2021.emnlp-main.807 · Cited by 14
Metadata
- Authors
- Wang, J., Tuyls, J., Wallace, E., & Singh, S.
- Venue
- Findings of the Association for Computational Linguistics: EMNLP 2020
- Published
- 2020
- Pages
- 247-258
- OpenAlex
- W3099954305
- Cited by
- 43
- References
- 18
- FWCI
- 3.02
- Type
- conference paper
- Language
- en
- Open access
- gold
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Gradient-based Analysis of NLP Models is Manipulable. https://app.cochat.ai/papers/10.18653/v1/2020.findings-emnlp.24
- The paper
- Wang, J., Tuyls, J., Wallace, E., & Singh, S. (2020). Gradient-based Analysis of NLP Models is Manipulable. Findings of the Association for Computational Linguistics: EMNLP 2020. https://doi.org/10.18653/v1/2020.findings-emnlp.24
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.