Solutions from different LLM providers converged broadly on most CS1 labs, with cross-LLM MOSS match rates of about 88% pooled across labs but only 13.18–22.11% on Lab 10¶
Claim · All claims
Evidence · 1 study · associationalr2·q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly
Subclaims¶
Reading the evidence codes
q is the design tier, i the impact, n the sample, and the kind with its r
the rigour judged within that kind: what each code means.
This claim: no standardized effect is recorded in observations/ for it yet, so no benchmark reading is possible.
q2 i? Pooled across the ten labs, 87.79–88.14% of cross-LLM pairs matched after docstring removal, only 3.40–3.61 percentage points below within-LLM rates; across Labs 1–9 cross-LLM rates were 95.48–96.08%, but on Lab 10 they were 13.18–22.11%. → 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
MOSS similarity analysis of 3,000 attempts per cohort (1,000 per model) with starter files registered as base files and docstrings removed. The article reports cross-LLM rates of 95.48–96.08% on Labs 1–9 versus 13.18–22.11% on Lab 10 (Table 3).
"Using all 1,000 attempts from each LLM and pooling the ten labs within each cohort, 87.79–88.14% of cross-LLM pairs matched after docstring removal. The corresponding within-LLM rates were 91.35–91.72%, only 3.40–3.61 percentage points higher (Table 3)."
Discussion¶
Related Claims¶
- Frontier LLMs produced near-perfect solutions to CS1 labs, with mean hidden-instructor-test scores of 98.93–99.99% across model–year combinations — related
- Human-LLM agreement on a complex multi-label codebook falls well below human-human agreement, while LLM-LLM agreement is comparable to human-human agreement — related
- 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
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
cross-llm-convergence-task-dependent |
title |
Solutions from different LLM providers converged broadly on most CS1 labs, with cross-LLM MOSS match rates of about 88% pooled across labs but only 13.18–22.11% on Lab 10 |
description |
Solutions from different LLM providers converged broadly on most CS1 labs, with cross-LLM MOSS match rates of about 88% pooled across labs but only 13.18–22.11% on Lab 10 |
status |
review |
evidence_strength |
moderate |
newest_evidence |
2026 |
recheck_by |
2029 |
generated |
by process:wiki-ingest on 2026-10-11 |
trust |
unverified — no one has confirmed this page's content |
sources |
1 — runlong-ye-2026 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.