Removing Spurious Correlation from Neural Network Interpretations
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866913596523413504 |
|---|---|
| author | Fotouhi, Milad Bahadori, Mohammad Taha Feyisetan, Oluwaseyi Arabshahi, Payman Heckerman, David |
| author_facet | Fotouhi, Milad Bahadori, Mohammad Taha Feyisetan, Oluwaseyi Arabshahi, Payman Heckerman, David |
| contents | The existing algorithms for identification of neurons responsible for undesired and harmful behaviors do not consider the effects of confounders such as topic of the conversation. In this work, we show that confounders can create spurious correlations and propose a new causal mediation approach that controls the impact of the topic. In experiments with two large language models, we study the localization hypothesis and show that adjusting for the effect of conversation topic, toxicity becomes less localized. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_02893 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Removing Spurious Correlation from Neural Network Interpretations Fotouhi, Milad Bahadori, Mohammad Taha Feyisetan, Oluwaseyi Arabshahi, Payman Heckerman, David Computation and Language Artificial Intelligence Machine Learning Applications Methodology The existing algorithms for identification of neurons responsible for undesired and harmful behaviors do not consider the effects of confounders such as topic of the conversation. In this work, we show that confounders can create spurious correlations and propose a new causal mediation approach that controls the impact of the topic. In experiments with two large language models, we study the localization hypothesis and show that adjusting for the effect of conversation topic, toxicity becomes less localized. |
| title | Removing Spurious Correlation from Neural Network Interpretations |
| topic | Computation and Language Artificial Intelligence Machine Learning Applications Methodology |
| url | https://arxiv.org/abs/2412.02893 |