Layered Insights: Generalizable Analysis of Authorial Style by Leveraging All Transformer Layers
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911204829560832 |
|---|---|
| author | Alshomary, Milad Varimalla, Nikhil Reddy Anand, Vishal Muresan, Smaranda McKeown, Kathleen |
| author_facet | Alshomary, Milad Varimalla, Nikhil Reddy Anand, Vishal Muresan, Smaranda McKeown, Kathleen |
| contents | We propose a new approach for the authorship attribution task that leverages the various linguistic representations learned at different layers of pre-trained transformer-based models. We evaluate our approach on three datasets, comparing it to a state-of-the-art baseline in in-domain and out-of-domain scenarios. We found that utilizing various transformer layers improves the robustness of authorship attribution models when tested on out-of-domain data, resulting in new state-of-the-art results. Our analysis gives further insights into how our model's different layers get specialized in representing certain stylistic features that benefit the model when tested out of the domain. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_00958 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Layered Insights: Generalizable Analysis of Authorial Style by Leveraging All Transformer Layers Alshomary, Milad Varimalla, Nikhil Reddy Anand, Vishal Muresan, Smaranda McKeown, Kathleen Computation and Language We propose a new approach for the authorship attribution task that leverages the various linguistic representations learned at different layers of pre-trained transformer-based models. We evaluate our approach on three datasets, comparing it to a state-of-the-art baseline in in-domain and out-of-domain scenarios. We found that utilizing various transformer layers improves the robustness of authorship attribution models when tested on out-of-domain data, resulting in new state-of-the-art results. Our analysis gives further insights into how our model's different layers get specialized in representing certain stylistic features that benefit the model when tested out of the domain. |
| title | Layered Insights: Generalizable Analysis of Authorial Style by Leveraging All Transformer Layers |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2503.00958 |