Advacheck System for Detecting AI-Generated Text: A Multi-Task Learning Approach
The Advacheck system effectively detects AI-generated text by employing a multi-task learning architecture, achieving superior performance compared to single-task models and demonstrating robustness across diverse datasets.
Multi-Head Span-Based Detector for Identifying AI-Generated Text in Scientific Papers
This paper proposes a novel multi-task learning architecture for detecting AI-generated text fragments in scientific papers, achieving improved accuracy by leveraging binary and multi-class classification heads within a single model.
DeTeCtive: A Novel Approach to Detecting AI-Generated Text by Identifying Unique Writing Styles
Instead of traditional binary classification, DeTeCtive reframes AI-generated text detection as a nuanced analysis of writing styles, leveraging multi-level contrastive learning to distinguish subtle differences between human and machine authors, and even among different AI models.
LLM-DetectAIve: A New Tool for Detecting and Classifying AI-Generated Text
LLM-DetectAIve is a new system designed to detect and classify different types of machine-generated text, moving beyond simple binary classification to identify subtle differences in human-machine collaboration.
Evaluating the Quality of AI-Generated Text Detection Datasets
The rapid advancement of AI text generation necessitates reliable detection methods, but the quality of current evaluation datasets raises concerns about the true performance of these detectors in real-world scenarios.
Fine-Tuning RoBERTa for Enhanced Detection of AI-Generated Text Across Multiple Domains
A fine-tuned RoBERTa model, trained on a diverse dataset and leveraging a human-LLM similarity ratio, demonstrates significant improvement in detecting AI-generated text across various domains, outperforming existing tools like DetectGPT and GPTZero.
如何偵測 AI 生成的論文審查？
大型語言模型 (LLM) 如 ChatGPT 的使用日益普及，引發了人們對其潛在濫用的擔憂，特別是在學術同行評審過程中。
Detecting AI-Generated Peer Reviews: A New Challenge for Scientific Integrity
The increasing use of large language models (LLMs) like ChatGPT in academic writing raises concerns about the potential for AI-generated peer reviews, necessitating the development of effective detection methods to maintain the integrity of the peer-review process.
Improving the Robustness of AI-Generated Text Detection Using Restricted Embeddings
Restricting the feature space of AI-generated text detectors by removing specific components from text embeddings, such as attention heads or embedding coordinates, can significantly improve their robustness and ability to generalize to unseen domains and generation models.
HC3 Plus：一個語義不變的人類與 ChatGPT 文本比較語料庫
針對當前 AI 生成文本檢測器在處理語義不變任務（如翻譯、摘要和改寫）方面的不足，本文提出了一個更廣泛、更全面的數據集 HC3 Plus，並使用指令微調模型訓練了一個更強大的檢測器。
© 2024 by Linnk AI