Reimagining service delivery in the agentic era with Google Public Sector
cloud.google.com·1d ago
AI Research
arXiv CS.AI·Jun 12·10 min read
Authorship Attribution in Multilingual Machine-Generated Texts
TL;DR
As Large Language Models (LLMs) become more sophisticated, distinguishing between machine-generated text and human-written content is increasingly challenging. This study introduces a method for Multilingual Authorship Attribution (AA), which identifies the source of text across 18 languages and 8 generators, including both LLMs and human authors.
✦ Why It Matters
Engineers and researchers can enhance authorship detection systems by considering multilingual capabilities and adapting existing methods for broader applications.
Key Takeaways
- Explore multilingual authorship attribution for diverse language applications.
- Adapt existing monolingual methods for multilingual contexts cautiously.
- Assess the performance of different LLMs in generating text.
- Address limitations in current methods to improve attribution accuracy.
How It Works
The study evaluates existing monolingual authorship attribution methods for their applicability in multilingual settings. It involves testing these methods across 18 languages and analyzing their ability to accurately attribute texts to either human authors or specific LLMs.
The researchers focus on the cross-lingual transferability of these methods, examining how well they perform when applied to languages outside their original training context.