DSIPA: Detecting LLM-Generated Texts via Sentiment-Invariant Patterns Divergence Analysis

arXiv:2604.26328v1 Announce Type: cross Abstract: The rapid advancement of large language models (LLMs) presents new security challenges, particularly in detecting machine-generated text used for misinformation, impersonation, and content forgery. Most existing detection approaches struggle with...