New framework measures alignment between computer science curricula and program coverage
A human-in-the-loop pipeline quantifies how undergraduate CS programs align with international curricular guidelines, revealing persistent gaps and the impact of guideline updates.
1 source · cross-referenced
- A new arXiv paper proposes a reusable pipeline to measure how completely undergraduate computer science programs cover international curricular guidelines.
- The framework was applied longitudinally to a single accredited BSc program against CS2013 and CS2023, finding near-constant coverage (49.7% vs. 50.9%) over a decade.
- The study also evaluates competency articulation and cognitive depth, showing a drop in delivery depth under CS2023 (76%) compared to CS2013 (95%).
- Persistent structural gaps were identified in parallel and distributed computing, foundations of programming languages, and systems fundamentals.
A new arXiv preprint introduces a human-in-the-loop pipeline designed to measure how completely an undergraduate computer science program covers external curricular guidelines, addressing a longstanding need for reliable, reproducible assessment tools in higher education.
The authors apply their framework to an accredited Bachelor of Science in Computer Science, comparing its coverage against two iterations of international curricular guidelines: Computer Science Curricula 2013 (CS2013) and 2023 (CS2023). The pipeline represents both the program and the guideline as structured corpora, uses semantic retrieval to generate candidate course-to-knowledge-unit matches, and confirms these matches through human judgment under an explicit coverage definition.
Among seven benchmarked retrievers, a reciprocal-rank-fusion ensemble performed best, and a reputed long-context model underperformed a small sentence model, indicating that retriever choice materially affects outcomes and must be empirically validated.
The alignment maps were validated by an independent second rater, yielding Cohen’s kappa values of 0.64 for CS2023 and 0.69 for CS2013, demonstrating moderate to substantial agreement.
The program’s coverage was quantified as 49.7% of CS2023 knowledge units and 50.9% of CS2013 knowledge units, a near-constant figure across the decade separating the two guidelines.
Extending the framework to competency articulation and cognitive depth, the authors find that the program articulates the competency for approximately 88% of covered units under both guidelines, but delivers the material at the recommended depth for 76% of present units under CS2023 compared to 95% under CS2013. The authors attribute this gap to the newer guideline’s raised expectations rather than changes in the program itself.
The longitudinal comparison identifies persistent structural gaps—parallel and distributed computing, foundations of programming languages, and systems fundamentals—that remain uncovered against both guidelines and ABET standards, distinguishing them from differences attributable to the evolution of the standards.
The instrument is presented as reusable and available from the authors on request, offering a practical tool for other institutions and accreditors.
- Jun 19, 2026 · arXiv cs.CL
No evidence of Semitic-specific cross-lingual transfer in large language models
Trust79 - Jun 19, 2026 · arXiv cs.CL
LLM ensemble achieves 0.74 F1-score in automating EQ-5D study detection from PubMed abstracts
Trust79 - Jun 19, 2026 · arXiv cs.CL
Frontier LLMs hit ceiling on VerilogEval hardware-coding benchmark, study finds
Trust79