Argus: Academic Integrity in the Era of Generative AI

Research from arXiv highlights significant challenges to academic integrity in programming education due to the proliferation of large language models (LLMs). A system called Argus, designed to detect LLM-assisted student work in C programming, analyzed six years of data from a large university course. It found that 45% of enrolled students exhibited patterns consistent with LLM assistance, indicating a widespread issue.

The findings underscore a fundamental shift in educational integrity, driven by advancements in generative AI. This necessitates a strategic re-evaluation of current assessment methods, instructional design, and institutional policies to maintain the value and credibility of educational outcomes in a technology-rich environment.

The rapid proliferation of large language models (LLMs) poses significant challenges to academic integrity, particularly in programming courses.

Forward consideration, not a verified fact.