Spotting LLMs With Binoculars: Zero-Shot Detection of Machine-Generated Text
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910647258710016 |
|---|---|
| author | Hans, Abhimanyu Schwarzschild, Avi Cherepanova, Valeriia Kazemi, Hamid Saha, Aniruddha Goldblum, Micah Geiping, Jonas Goldstein, Tom |
| author_facet | Hans, Abhimanyu Schwarzschild, Avi Cherepanova, Valeriia Kazemi, Hamid Saha, Aniruddha Goldblum, Micah Geiping, Jonas Goldstein, Tom |
| contents | Detecting text generated by modern large language models is thought to be hard, as both LLMs and humans can exhibit a wide range of complex behaviors. However, we find that a score based on contrasting two closely related language models is highly accurate at separating human-generated and machine-generated text. Based on this mechanism, we propose a novel LLM detector that only requires simple calculations using a pair of pre-trained LLMs. The method, called Binoculars, achieves state-of-the-art accuracy without any training data. It is capable of spotting machine text from a range of modern LLMs without any model-specific modifications. We comprehensively evaluate Binoculars on a number of text sources and in varied situations. Over a wide range of document types, Binoculars detects over 90% of generated samples from ChatGPT (and other LLMs) at a false positive rate of 0.01%, despite not being trained on any ChatGPT data. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_12070 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Spotting LLMs With Binoculars: Zero-Shot Detection of Machine-Generated Text Hans, Abhimanyu Schwarzschild, Avi Cherepanova, Valeriia Kazemi, Hamid Saha, Aniruddha Goldblum, Micah Geiping, Jonas Goldstein, Tom Computation and Language Artificial Intelligence Machine Learning Detecting text generated by modern large language models is thought to be hard, as both LLMs and humans can exhibit a wide range of complex behaviors. However, we find that a score based on contrasting two closely related language models is highly accurate at separating human-generated and machine-generated text. Based on this mechanism, we propose a novel LLM detector that only requires simple calculations using a pair of pre-trained LLMs. The method, called Binoculars, achieves state-of-the-art accuracy without any training data. It is capable of spotting machine text from a range of modern LLMs without any model-specific modifications. We comprehensively evaluate Binoculars on a number of text sources and in varied situations. Over a wide range of document types, Binoculars detects over 90% of generated samples from ChatGPT (and other LLMs) at a false positive rate of 0.01%, despite not being trained on any ChatGPT data. |
| title | Spotting LLMs With Binoculars: Zero-Shot Detection of Machine-Generated Text |
| topic | Computation and Language Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2401.12070 |