LLMs used for authorship attribution and verification in new research

Researchers have developed a novel unsupervised framework that leverages the log-probabilities of large language models (LLMs) for authorship attribution and verification. This method utilizes the extensive pre-training and one-shot capabilities of LLMs to measure style transferability between texts, outperforming existing unsupervised baselines and showing competitive results with contrastive methods. The framework's effectiveness increases with model scale and demonstrates strong performance across multiple languages, with an optional mechanism to enhance accuracy at the cost of increased computation. AI

IMPACT This research could enhance the accuracy and efficiency of detecting plagiarism and verifying authorship in academic and professional writing.

RANK_REASON Academic paper detailing a new methodology for authorship attribution using LLMs. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Litmaps
• Pablo Miralles-Gonzalez
• ScienceCast
• scite Smart Citations

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →