Back
AI-generated text detection survey
paperAuthors
Tang, Ruixiang·Chuang, Yu-Neng·Hu, Xia
Credibility Rating
3/5
Good(3)Good quality. Reputable source with community review or editorial standards, but less rigorous than peer-reviewed venues.
Rating inherited from publication venue: arXiv
A survey of detection methods for LLM-generated text, addressing a critical AI safety concern about distinguishing synthetic from human content and analyzing both technical detection approaches and their limitations.
Paper Details
Citations
11
4 influential
Year
2023
Methodology
peer-reviewed
Categories
Advances in Machine Learning & Artificial Inte
Metadata
arxiv preprintanalysis
Summary
This comprehensive survey examines current approaches for detecting large language model (LLM) generated text, analyzing black-box and white-box detection techniques. The research highlights the challenges and potential solutions for distinguishing between human and AI-authored content.
Key Points
- •Black-box detection relies on collecting and analyzing text samples from human and machine sources
- •White-box detection involves embedding watermarks directly into language model outputs
- •Current detection methods face challenges with evolving language model capabilities
Review
The survey provides a comprehensive overview of LLM-generated text detection, addressing a critical challenge in the era of advanced language models. The authors systematically break down detection methods into black-box and white-box approaches, exploring techniques such as statistical disparities, linguistic pattern analysis, and watermarking strategies. The research emphasizes the evolving nature of detection methods, acknowledging that as language models improve, current detection techniques may become less effective. Key contributions include detailed analysis of data collection strategies, feature selection techniques, and the potential limitations of existing approaches. The authors critically examine challenges such as dataset bias, confidence calibration, and the emerging threats from open-source language models, providing a nuanced perspective on the field's current state and future research directions.
Cited by 1 page
| Page | Type | Quality |
|---|---|---|
| Authentication Collapse | Risk | 57.0 |
Cached Content Preview
HTTP 200Fetched Apr 7, 202655 KB
[2303.07205] The Science of Detecting LLM-Generated Texts
The Science of Detecting LLM-Generated Texts
Ruixiang Tang, Yu-Neng Chuang, Xia Hu
Department of Computer Science, Rice University 6100 Main St Houston USA
rt39, ynchuang, xia.hu @rice.edu
Abstract.
The emergence of large language models (LLMs) has resulted in the production of LLM-generated texts that is highly sophisticated and almost indistinguishable from texts written by humans. However, this has also sparked concerns about the potential misuse of such texts, such as spreading misinformation and causing disruptions in the education system. Although many detection approaches have been proposed, a comprehensive understanding of the achievements and challenges is still lacking. This survey aims to provide an overview of existing LLM-generated text detection techniques and enhance the control and regulation of language generation models. Furthermore, we emphasize crucial considerations for future research, including the development of comprehensive evaluation metrics and the threat posed by open-source LLMs, to drive progress in the area of LLM-generated text detection.
† † copyright: none
Figure 1. An overview of the LLM-generated text detection.
1. Introduction
Recent advancements in natural language generation (NLG) technology have significantly improved the diversity, control, and quality of LLM-generated texts. A notable example is OpenAI’s ChatGPT, which demonstrates exceptional performance in tasks such as answering questions, composing emails, essays, and codes. However, this newfound capability to produce human-like texts at high efficiency also raises concerns about detecting and preventing misuse of LLMs in tasks such as phishing, disinformation, and academic dishonesty. For instance, many schools banned ChatGPT due to concerns over cheating in assignments (Elsen-Rooney, 2023 ) , and media outlets have raised the alarm over fake news generated by LLMs (Floridi and Chiriatti, 2020 ) . These concerns about the misuse of LLMs have hindered the NLG application in important domains such as media and education.
The ability to accurately detect LLM-generated texts is critical for realizing the full potential of NLG while minimizing serious consequences. From the perspective of end-users, LLM-generated text detection could increase trust in NLG systems and encourage adoption. For machine learning system developers and researchers, the detector can aid in tracing generated texts and preventing unauthorized use. Given its significance, there has been a growing interest in academia and industry to pursue research on LLM-generated text detection and to deepen our understanding of its underlying mechanisms.
While there is a rising discussion on whether LLM-generat-ed texts could be properly detected and how this can be done, we provide a comprehensive technical introduction of existing detectio
... (truncated, 55 KB total)Resource ID:
50a941dd05ba5219 | Stable ID: sid_4SUO5JVPP7