Over 30% of new ArXiv submissions now read as AI-written

Explain Like I'm Normal

A statistical analysis of ArXiv pre-prints indicates a massive surge in LLM-associated vocabulary like 'delve' and 'showcase' since 2023. This suggests nearly a third of all scientific submissions are now generated or heavily polished by AI, raising concerns about peer review collapse and model collapse. While it helps non-native speakers, the homogenizing effect on scientific literature is accelerating.