Student solutions became less varied on four of five selected functions by 2025, with exact-form declines of 18.4–51.1%, while the open-ended loopy_madness did not decline¶

Claim · All claims
Evidence · 1 study · associational r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2031: baselines and conditions change

Subclaims¶

q2 i? Relative to 2021, distinct exact AST forms per 1,000 files in 2025 were lower by 51.1% for my_and, 25.4% for exists_triangle, 22.9% for sum_string, and 18.4% for is_square, but 5.1% higher for loopy_madness with an interval including zero. → Runlong Ye 2026

Evidence¶

Runlong Ye 2026¶

Runlong Ye, Jing Fan, Angela Zavaleta Bernuy, Oscar Karnalim, Paul Denny, Juho Leinonen, and Michael Liut. (2026). Correctness, Convergence, and AI-Generated Code Detection: A Longitudinal Study of Student and Large Language Model Code in Introductory Programming. https://arxiv.org/abs/2610.00863 doi:10.48550/arxiv.2610.00863

q2 · i? · associational · r2

Cohort comparison of distinct serialized AST forms across 2021, 2023, and 2025 using each cohort's full eligible pool, with Poisson 95% intervals and an exact interval on the 2021-to-2025 rate ratio (Figure 3). The four declines exclude zero; the loopy_madness interval does not.

"Relative to 2021, the number of exact forms per 1,000 files in 2025 was lower by 51.1% for my_and (95% CI 35–64%), 25.4% for exists_triangle (16–34%), 22.9% for sum_string (14–31%), and 18.4% for is_square (7–28%)."

Discussion¶

Related Claims¶

• On tightly specified functions, LLMs converged on a few exact AST forms (1–14 per 1,000 files) while an open-ended function remained diverse, and source was associated with exact form for all five functions — related