Average H-scores rose from near-zero in pre-LLM offerings to 11.9 in Spring 2026, with maximum scores reaching 81¶

Claim · All claims
Evidence · 1 study · associational r1 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: local conditions change quickly

Subclaims¶

q2 i? Score distributions were heavily right-skewed, with means rising from 1.1 (Spring 2020) and 0.5 (Spring 2022) to 4.2 (Spring 2025) and 11.9 (Spring 2026). → David Racovan 2026

Evidence¶

David Racovan 2026¶

David Racovan, Ajay Rawat, Christopher K. May, Jeffrey A. Turkstra. (2026). Argus: Academic Integrity in the Era of Generative AI. https://arxiv.org/abs/2609.36073 doi:10.48550/arxiv.2609.36073

q2 · i? · associational · r1

Distributional analysis of 3,893 student-semester pairs and 42,695 homework submissions across six offerings, shown in Figure 2; the text reports "Spring 2026 (mean =11.9)". Maximum H-scores reached 81 in Spring 2026 versus 40–45 earlier. No effect size is printed.

"A more pronounced shift appears in Spring 2025 (mean =4.2), and Spring 2026 (mean =11.9). This suggests a large change in student behavior towards the use of LLM-based assistance, rather than a change driven by a small number of outliers."

Discussion¶

Related Claims¶

• In Spring 2026, 45.7% of students in a large CS2 course were flagged as LLM-assisted following structured manual review — related
• Average homework H-scores correlate negatively with proctored exam performance, with the correlation strengthening from Spring 2024 to Spring 2026 — related
• More students embraced LLMs for debugging from Spring to Fall 2025 despite declining course enrollment — related
• MOSS detected far fewer suspicious cases in Spring 2026 than in prior offerings, as LLM-assisted submissions lack similarity signals — related
• In an unrestricted university course, only a minority of students (23 of 68, 34%) reported using LLMs, averaging 12% of discussion-point and 11% of critical-summary submissions — related
• All four evaluated LLMs produce strongly bimodal item-level score distributions on bash exams, and rubric-enhanced prompts amplify the near-perfect-score peak — related
• Authorship scores were highly concentrated near the maximum of the rubric scale — related
• Flagged Spring 2026 students scored lower on all three exams than non-flagged peers, with performance declining as flagged assignments increased — related