EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Causal Adversarial Training for Improving Language Understanding and Robustness

Bin Zhu, Yi Xie, Yanxia Feng, Fu Lee Wang, Haoran Xie, Yanghui Rao

Computational Linguistics · 11.11.2026 · 1-37

DOI: 10.1162/coli.a.644 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Computational Linguistics, online erschienen am 11.11.2026. Verfasst von Bin Zhu und 5 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Language models suffer from adversarial vulnerability, where imperceptible perturbations in inputs can mislead model predictions. Existing adversarial training methods enhance model robustness by augmenting training data with worst-case adversarial perturbations for each input. However, these approaches overlook the intrinsic causal relationships within data. We construct causal graphs to model the behavior of encoder-based language models under adversarial inputs, suggesting that adversarial vulnerability is often associated with the model’s overfitting to spurious correlations. Within the adversarial training framework, we formulate expressions for Total Effect (TE) and Natural Direct Effect (NDE), and propose a causal regularization objective based on TE-NDE alignment. Experiments on five benchmark datasets show that this causal adversarial training framework improves adversarial robustness across different settings. We further evaluate the model behavior in a controlled sentiment analysis setting, where the quantitative results provide additional evidence for the effectiveness of our method.

Zitieren

APA

Zhu, B., Xie, Y., Feng, Y., Wang, F. L., Xie, H., & Rao, Y. (2026). Causal Adversarial Training for Improving Language Understanding and Robustness. Computational Linguistics, 1-37. https://doi.org/10.1162/coli.a.644

Vancouver

Zhu B, Xie Y, Feng Y, Wang FL, Xie H, Rao Y. Causal Adversarial Training for Improving Language Understanding and Robustness. Computational Linguistics. 2026:1-37. doi:10.1162/coli.a.644

BibTeX
@article{Zhu2026,
  author = {Zhu, Bin and Xie, Yi and Feng, Yanxia and Wang, Fu Lee and Xie, Haoran and Rao, Yanghui},
  title = {Causal Adversarial Training for Improving Language Understanding and Robustness},
  journal = {Computational Linguistics},
  year = {2026},
  pages = {1--37},
  doi = {10.1162/coli.a.644}
}
Kostenlos starten →