Does DetectGPT Fully Utilize Perturbation? Bridging Selective Perturbation to Fine-tuned Contrastive Learning Detector would be Better
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_ | 1866910517012987904 |
|---|---|
| author | Liu, Shengchao Liu, Xiaoming Wang, Yichen Cheng, Zehua Li, Chengzhengxu Zhang, Zhaohan Lan, Yu Shen, Chao |
| author_facet | Liu, Shengchao Liu, Xiaoming Wang, Yichen Cheng, Zehua Li, Chengzhengxu Zhang, Zhaohan Lan, Yu Shen, Chao |
| contents | The burgeoning generative capabilities of large language models (LLMs) have raised growing concerns about abuse, demanding automatic machine-generated text detectors. DetectGPT, a zero-shot metric-based detector, first introduces perturbation and shows great performance improvement. However, in DetectGPT, the random perturbation strategy could introduce noise, and logit regression depends on the threshold, harming the generalizability and applicability of individual or small-batch inputs. Hence, we propose a novel fine-tuned detector, Pecola, bridging metric-based and fine-tuned methods by contrastive learning on selective perturbation. Selective strategy retains important tokens during perturbation and weights for multi-pair contrastive learning. The experiments show that Pecola outperforms the state-of-the-art (SOTA) by 1.20% in accuracy on average on four public datasets. And we further analyze the effectiveness, robustness, and generalization of the method. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_00263 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Does DetectGPT Fully Utilize Perturbation? Bridging Selective Perturbation to Fine-tuned Contrastive Learning Detector would be Better Liu, Shengchao Liu, Xiaoming Wang, Yichen Cheng, Zehua Li, Chengzhengxu Zhang, Zhaohan Lan, Yu Shen, Chao Computation and Language The burgeoning generative capabilities of large language models (LLMs) have raised growing concerns about abuse, demanding automatic machine-generated text detectors. DetectGPT, a zero-shot metric-based detector, first introduces perturbation and shows great performance improvement. However, in DetectGPT, the random perturbation strategy could introduce noise, and logit regression depends on the threshold, harming the generalizability and applicability of individual or small-batch inputs. Hence, we propose a novel fine-tuned detector, Pecola, bridging metric-based and fine-tuned methods by contrastive learning on selective perturbation. Selective strategy retains important tokens during perturbation and weights for multi-pair contrastive learning. The experiments show that Pecola outperforms the state-of-the-art (SOTA) by 1.20% in accuracy on average on four public datasets. And we further analyze the effectiveness, robustness, and generalization of the method. |
| title | Does DetectGPT Fully Utilize Perturbation? Bridging Selective Perturbation to Fine-tuned Contrastive Learning Detector would be Better |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2402.00263 |