Authorship Attribution in Multilingual Machine-Generated Texts

This work introduces the concept of Multilingual Authorship Attribution (AA), addressing the challenge of identifying the generator of machine-generated texts across 18 languages and 8 generators, including 7 LLMs and human authors. The study evaluates the effectiveness of existing monolingual AA methods in multilingual contexts, revealing limitations in cross-lingual transferability and highlighting the complexities involved in accurately attributing authorship across diverse language families. This research is significant for practitioners as it underscores the necessity for more advanced techniques to handle the intricacies of multilingual text generation and attribution, which is increasingly relevant in the deployment of LLMs.