Beyond Via: Analysis and Estimation of the Impact of Large Language Models in Academic Papers

A recent analysis of academic papers on arXiv reveals significant shifts in word usage patterns, likely driven by the adoption of large language models (LLMs)¹. Notably, the frequency of words like "beyond" and "via" in titles has increased, while the usage of common words like "the" and "of" in abstracts has decreased. These changes suggest that LLMs are having a profound impact on the way researchers express themselves in academic writing. The similarities among different LLMs also pose challenges for current classifiers, which struggle to distinguish between human-written and AI-generated content. As LLMs continue to influence academic writing, it matters to practitioners because the increasing reliance on AI-generated content can have far-reaching implications for the validity and credibility of research, ultimately affecting the integrity of the scientific record.