Papers
Topics
Authors
Recent
Search
2000 character limit reached

A Survey on Out-of-Distribution Evaluation of Neural NLP Models

Published 27 Jun 2023 in cs.CL | (2306.15261v1)

Abstract: Adversarial robustness, domain generalization and dataset biases are three active lines of research contributing to out-of-distribution (OOD) evaluation on neural NLP models. However, a comprehensive, integrated discussion of the three research lines is still lacking in the literature. In this survey, we 1) compare the three lines of research under a unifying definition; 2) summarize the data-generating processes and evaluation protocols for each line of research; and 3) emphasize the challenges and opportunities for future work.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (53)
  1. Generating natural language adversarial examples. In EMNLP, pages 2890–2896, 2018.
  2. Types of out-of-distribution texts and how to detect them. In EMNLP, pages 10687–10701, 2021.
  3. Adversarial removal of demographic attributes revisited. In EMNLP-IJCNLP, pages 6330–6335, 2019.
  4. Universal adversarial attacks on text classifiers. In ICASSP, pages 7345–7349, 2019.
  5. Nuanced metrics for measuring unintended bias with real data for text classification. In World Wide Web Conference, page 491–500, 2019.
  6. Adversarial filters of dataset biases. In ICML, pages 1078–1088, 2020.
  7. Balanced adversarial training: Balancing tradeoffs between fickleness and obstinacy in NLP models. In EMNLP, 2022.
  8. Seq2sick: Evaluating the robustness of sequence-to-sequence models with adversarial examples. In AAAI, pages 3601–3608, 2020.
  9. Hotflip: White-box adversarial examples for text classification. In ACL, pages 31–36, 2018.
  10. Detecting word sense disambiguation biases in machine translation for model-agnostic adversarial attacks. In EMNLP, pages 7635–7653, 2020.
  11. Black-box generation of adversarial text sequences to evade deep learning classifiers. In 2018 IEEE Security and Privacy Workshops (SPW), pages 50–56, 2018.
  12. Evaluating models’ local decision boundaries via contrast sets. In EMNLP, pages 1307–1323, 2020.
  13. Are we modeling the task or the annotator? an investigation of annotator bias in natural language understanding datasets. In EMNLP-IJCNLP, pages 1161–1166, 2019.
  14. Generalized but not Robust? comparing the effects of data modification methods on out-of-domain generalization and adversarial robustness. In ACL, pages 2705–2718, 2022.
  15. Adversarial texts with gradient methods. arXiv preprint arXiv:1801.07175, 2018.
  16. Explaining and harnessing adversarial examples. In ICLR, 2015.
  17. Annotation artifacts in natural language inference data. In NAACL-HLT, pages 107–112, 2018.
  18. Don’t stop pretraining: Adapt language models to domains and tasks. In ACL, pages 8342–8360, July 2020.
  19. Pretrained transformers improve out-of-distribution robustness. In ACL, pages 2744–2751, 2020.
  20. Adversarial examples are not bugs, they are features. In Advances in NIPS, 2019.
  21. Adversarial example generation with syntactically controlled paraphrase networks. In NAACL-HLT, pages 1875–1885, 2018.
  22. Adversarial examples for evaluating reading comprehension systems. In EMNLP, pages 2021–2031, 2017.
  23. Is bert really robust? a strong baseline for natural language attack on text classification and entailment. In AAAI, pages 8018–8025, 2020.
  24. Learning the difference that makes a difference with counterfactually-augmented data. In ICLR, 2020.
  25. Content selection in deep learning models of summarization. In EMNLP, pages 1818–1828, 2018.
  26. Why machine reading comprehension models learn shortcuts? In ACL-IJCNLP, pages 989–1002, 2021.
  27. Textbugger: Generating adversarial text against real-world applications. In NDSS, 2019.
  28. BERT-ATTACK: Adversarial attack against BERT using BERT. In EMNLP, pages 6193–6202, 2020.
  29. Deep text classification can be fooled. In IJCAI, pages 4208–4215, 7 2018.
  30. Uncovering the connections between adversarial transferability and knowledge transferability. In ICML, pages 6577–6587, 2021.
  31. Character-level white-box adversarial attacks against transformers via attachable subwords substitution. In EMNLP, pages 7664–7676, December 2022.
  32. Right for the wrong reasons: Diagnosing syntactic heuristics in natural language inference. In ACL, pages 3428–3448, 2019.
  33. The effect of natural distribution shift on question answering models. In Hal Daumé III and Aarti Singh, editors, Proceedings of the 37th International Conference on Machine Learning, volume 119 of Proceedings of Machine Learning Research, pages 6905–6916. PMLR, 13–18 Jul 2020.
  34. Universal adversarial perturbations. In CVPR, pages 1765–1773, 2017.
  35. Stress test evaluation for natural language inference. In CoLing, pages 2340–2353, 2018.
  36. Adversarial NLI: A new benchmark for natural language understanding. In ACL, pages 4885–4901, 2020.
  37. The effect of sociocultural variables on sarcasm communication online. In ACM, page 22, 2020.
  38. Mind the style of text! adversarial and backdoor attacks based on text style transfer. In EMNLP, pages 4569–4580, 2021.
  39. Semantically equivalent adversarial rules for debugging NLP models. In ACL, pages 856–865, 2018.
  40. Winogrande: An adversarial Winograd schema challenge at scale. In AAAI, pages 8732–8740, 2020.
  41. Towards debiasing fact verification models. In EMNLP-IJCNLP, pages 3419–3425, 2019.
  42. Adversarial semantic collisions. In EMNLP, pages 4198–4210, 2020.
  43. Universal adversarial attacks with natural triggers for text classification. In NAACL-HLT, pages 3724–3733, 2021.
  44. What makes reading comprehension questions easier? In EMNLP, pages 4208–4219, 2018.
  45. Universal adversarial triggers for attacking and analyzing NLP. In EMNLP-IJCNLP, pages 2153–2162, 2019.
  46. Trick me if you can: Human-in-the-loop generation of adversarial examples for question answering. TACL, 7:387–401, 2019.
  47. Imitation attacks and defenses for black-box machine translation systems. In Bonnie Webber, Trevor Cohn, Yulan He, and Yang Liu, editors, EMNLP, pages 5531–5546, 2020.
  48. EDA: Easy data augmentation techniques for boosting performance on text classification tasks. In EMNLP-IJCNLP, pages 6382–6388, 2019.
  49. Improved ood generalization via adversarial training and pretraing. In Marina Meila and Tong Zhang, editors, ICML, volume 139 of Proceedings of Machine Learning Research, pages 11987–11997. PMLR, 18–24 Jul 2021.
  50. Word-level textual adversarial attacking as combinatorial optimization. In ACL, pages 6066–6080, 2020.
  51. Swag: A large-scale adversarial dataset for grounded commonsense inference. In EMNLP, pages 93–104, 2018.
  52. PAWS: Paraphrase adversaries from word scrambling. In NAACL, pages 1298–1308, 2019.
  53. Generating natural adversarial examples. In ICLR, 2018.
Citations (14)
View on Semantic Scholar

Summary

No one has generated a summary of this paper yet.

Sign Up to Summarize

Paper to Video (Beta)

No one has generated a video about this paper yet.

Sign Up to Generate All Videos Subscribe on YouTube

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Sign Up to Generate

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Generate Now

Continue Learning

We haven't generated follow-up questions for this paper yet.

Generate Now