Robust Infidelity: When Faithfulness Measures on Masked Language Models Are Misleading
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , |
|---|---|
| Format: | Preprint |
| Publié: |
2023
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914818353528832 |
|---|---|
| author | Crothers, Evan Viktor, Herna Japkowicz, Nathalie |
| author_facet | Crothers, Evan Viktor, Herna Japkowicz, Nathalie |
| contents | A common approach to quantifying neural text classifier interpretability is to calculate faithfulness metrics based on iteratively masking salient input tokens and measuring changes in the model prediction. We propose that this property is better described as "sensitivity to iterative masking", and highlight pitfalls in using this measure for comparing text classifier interpretability. We show that iterative masking produces large variation in faithfulness scores between otherwise comparable Transformer encoder text classifiers. We then demonstrate that iteratively masked samples produce embeddings outside the distribution seen during training, resulting in unpredictable behaviour. We further explore task-specific considerations that undermine principled comparison of interpretability using iterative masking, such as an underlying similarity to salience-based adversarial attacks. Our findings give insight into how these behaviours affect neural text classifiers, and provide guidance on how sensitivity to iterative masking should be interpreted. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2308_06795 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Robust Infidelity: When Faithfulness Measures on Masked Language Models Are Misleading Crothers, Evan Viktor, Herna Japkowicz, Nathalie Computation and Language Machine Learning A common approach to quantifying neural text classifier interpretability is to calculate faithfulness metrics based on iteratively masking salient input tokens and measuring changes in the model prediction. We propose that this property is better described as "sensitivity to iterative masking", and highlight pitfalls in using this measure for comparing text classifier interpretability. We show that iterative masking produces large variation in faithfulness scores between otherwise comparable Transformer encoder text classifiers. We then demonstrate that iteratively masked samples produce embeddings outside the distribution seen during training, resulting in unpredictable behaviour. We further explore task-specific considerations that undermine principled comparison of interpretability using iterative masking, such as an underlying similarity to salience-based adversarial attacks. Our findings give insight into how these behaviours affect neural text classifiers, and provide guidance on how sensitivity to iterative masking should be interpreted. |
| title | Robust Infidelity: When Faithfulness Measures on Masked Language Models Are Misleading |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2308.06795 |