Your CS Degree May Be Missing 50% of What You Need — New AI Pipeline Reveals the Gap

A team led by Sherzod Turaev from UAE University has introduced a novel framework that uses semantic retrieval and human validation to quantify how well undergraduate computer science programs align with evolving curricular guidelines. The pipeline, tested against the ACM/IEEE Computer Science Curricula 2013 (CS2013) and 2023 (CS2023), first represents both the program and the guideline as structured corpora. It then employs a reciprocal-rank-fusion ensemble of seven retrievers to match courses to knowledge units, with human judges confirming matches under a strict coverage definition. The researchers found that a small sentence model outperformed a reputed long-context model, underscoring the importance of retriever selection. Inter-rater reliability was strong (Cohen's kappa of 0.64 for CS2023 and 0.69 for CS2013).

Applying the framework longitudinally, the program covered 49.7% of CS2023 knowledge units and 50.9% of CS2013—a nearly constant level over a decade. While competency articulation remained high at ~88% for covered units under both guidelines, the cognitive depth at which that competency was delivered dropped from 95% under CS2013 to 76% under CS2023, driven entirely by the newer guideline's heightened expectations. The analysis identified persistent structural gaps (parallel and distributed computing, foundations of programming languages, systems fundamentals) that appear across both guidelines and the ABET accreditation standard. The authors stress that their instrument is reusable and available upon request, offering departments a data-driven method to diagnose curriculum deficiencies.