Faithfulness and the Notion of Adversarial Sensitivity in NLP Explanations
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866909407125700608 |
|---|---|
| author | Manna, Supriya Sett, Niladri |
| author_facet | Manna, Supriya Sett, Niladri |
| contents | Faithfulness is arguably the most critical metric to assess the reliability of explainable AI. In NLP, current methods for faithfulness evaluation are fraught with discrepancies and biases, often failing to capture the true reasoning of models. We introduce Adversarial Sensitivity as a novel approach to faithfulness evaluation, focusing on the explainer's response when the model is under adversarial attack. Our method accounts for the faithfulness of explainers by capturing sensitivity to adversarial input changes. This work addresses significant limitations in existing evaluation techniques, and furthermore, quantifies faithfulness from a crucial yet underexplored paradigm. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2409_17774 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Faithfulness and the Notion of Adversarial Sensitivity in NLP Explanations Manna, Supriya Sett, Niladri Computation and Language Artificial Intelligence Faithfulness is arguably the most critical metric to assess the reliability of explainable AI. In NLP, current methods for faithfulness evaluation are fraught with discrepancies and biases, often failing to capture the true reasoning of models. We introduce Adversarial Sensitivity as a novel approach to faithfulness evaluation, focusing on the explainer's response when the model is under adversarial attack. Our method accounts for the faithfulness of explainers by capturing sensitivity to adversarial input changes. This work addresses significant limitations in existing evaluation techniques, and furthermore, quantifies faithfulness from a crucial yet underexplored paradigm. |
| title | Faithfulness and the Notion of Adversarial Sensitivity in NLP Explanations |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2409.17774 |