Full-scale LLM versions consistently achieve higher agreement with human grades than their mini and nano counterparts¶

Claim · All claims
Evidence · 1 study · associational r2 · q2
Currency · newest study 2025 · not yet checked against its sources · not reviewed by a person · recheck by 2028: tool and model accuracy changes quickly

Subclaims¶

q2 i? Full-scale models outperform their simplified counterparts on human-agreement ICC: gpt-4.1 (0.421) vs gpt-4.1-nano (0.204), gpt-5 (0.378) vs gpt-5-nano (0.314), gpt-4o (0.438) vs gpt-4o-mini (0.365). → Marcin Jukiewicz 2025

Evidence¶

Marcin Jukiewicz 2025¶

Marcin Jukiewicz. (2025). A systematic comparison of Large Language Models for automated assignment assessment in programming education: Exploring the importance of architecture and vendor. arXiv preprint. https://arxiv.org/abs/2509.26483 doi:10.48550/arxiv.2509.26483

q2 · i? · associational · r2

ICC(2,1) comparison with human grades across model size variants (Table 5). The study reports "the full gpt-4.1 (0.421) significantly outperforms gpt-4.1-nano (0.204)" and similar patterns for gpt-5 and gpt-4o families, suggesting reduced model complexity diminishes consistency.

"For instance, the full gpt-4.1 (0.421) significantly outperforms gpt-4.1-nano (0.204). Similarly, gpt-5 (0.378) performs better than gpt-5-nano (0.314), and gpt-4o (0.438) outperforms gpt-4o-mini (0.365)."

Discussion¶

Related Claims¶

• LLMs agree strongly with their own consensus grades (most ICC > 0.8), far more than with human teachers' grades — related
• Even the best LLM grader reaches only moderate agreement with human teachers, below the threshold for good reliability — related
• LLM ratings of classroom transcripts are more correlated with each other than with expert human ratings, across the same and different tasks — related
• GPT-4o mini achieved the lowest ACT balance MAE (6.12) in replicating human supervisor ratings across 49 full transcripts — related