New research tackles multilingual authorship attribution for AI-generated text

Researchers have introduced the problem of Multilingual Authorship Attribution (MAA) to address the challenge of distinguishing machine-generated text from human-written content across various languages. The study investigated the effectiveness of adapting monolingual authorship attribution methods to multilingual settings, focusing on 18 languages and 8 generators, including 7 large language models and human authors. Findings indicate that while some monolingual methods show promise for multilingual transfer, significant limitations persist, especially when transferring across diverse language families, highlighting the need for more robust approaches. AI

IMPACT This research highlights the growing difficulty in distinguishing AI-generated text from human writing across multiple languages, necessitating more advanced attribution methods.

RANK_REASON The cluster contains a research paper published on arXiv detailing a new problem and methodology in AI text analysis. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Andrea Tagarelli
• arXiv
• English
• Hugging Face
• large-language models
• Machine-Generated Text
• Multilingual Authorship Attribution

• paper
• other

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