Evaluating the Faithfulness of Importance Measures in NLP by Recursively Masking Allegedly Important Tokens and Retraining

Evaluating the Faithfulness of Importance Measures in NLP by Recursively Masking Allegedly Important Tokens and Retraining Andreas Madsen author Nicholas Meade author Vaibhav Adlakha author Siva Reddy author 2022-12 text Findings of the Association for Computational Linguistics: EMNLP 2022 Yoav Goldberg editor Zornitsa Kozareva editor Yue Zhang editor Association for Computational Linguistics Abu Dhabi, United Arab Emirates conference publication madsen-etal-2022-evaluating 10.18653/v1/2022.findings-emnlp.125 https://aclanthology.org/2022.findings-emnlp.125/ 2022-12 1731 1751