Who Wrote the Book? Detecting and Attributing LLM Ghostwriters
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866911553002930176 |
|---|---|
| author | Shetty, Anudeex Xu, Qiongkai Ohrimenko, Olga Lau, Jey Han |
| author_facet | Shetty, Anudeex Xu, Qiongkai Ohrimenko, Olga Lau, Jey Han |
| contents | In this paper, we introduce GhostWriteBench, a dataset for LLM authorship attribution. It comprises long-form texts (50K+ words per book) generated by frontier LLMs, and is designed to test generalisation across multiple out-of-distribution (OOD) dimensions, including domain and unseen LLM author. We also propose TRACE -- a novel fingerprinting method that is interpretable and lightweight -- that works for both open- and closed-source models. TRACE creates the fingerprint by capturing token-level transition patterns (e.g., word rank) estimated by another lightweight language model. Experiments on GhostWriteBench demonstrate that TRACE achieves state-of-the-art performance, remains robust in OOD settings, and works well in limited training data scenarios. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_28054 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Who Wrote the Book? Detecting and Attributing LLM Ghostwriters Shetty, Anudeex Xu, Qiongkai Ohrimenko, Olga Lau, Jey Han Computation and Language In this paper, we introduce GhostWriteBench, a dataset for LLM authorship attribution. It comprises long-form texts (50K+ words per book) generated by frontier LLMs, and is designed to test generalisation across multiple out-of-distribution (OOD) dimensions, including domain and unseen LLM author. We also propose TRACE -- a novel fingerprinting method that is interpretable and lightweight -- that works for both open- and closed-source models. TRACE creates the fingerprint by capturing token-level transition patterns (e.g., word rank) estimated by another lightweight language model. Experiments on GhostWriteBench demonstrate that TRACE achieves state-of-the-art performance, remains robust in OOD settings, and works well in limited training data scenarios. |
| title | Who Wrote the Book? Detecting and Attributing LLM Ghostwriters |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2603.28054 |