Ranking Consistency Analysis on MMLU-Pro Medical genetics health
0.42Spearman CorrelationSCICUSTOM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SCICUSTOMConstruction Method=Grounded data and ontology-driven2026.05 | 0.42 | 0.33 | |
| MedQAConstruction Method=Domain-specific comparison2026.05 | 0.35 | 0.23 | |
| GPT-5Construction Method=Synthetic benchmark generation2026.05 | 0.09 | 0.06 | |
| MMLUConstruction Method=Domain-specific comparison2026.05 | 0.04 | 0 | |
| SimpleQAConstruction Method=General-purpose benchmark2026.05 | 0 | 0 | |
| GPQAConstruction Method=General scientific benchmark2026.05 | -0.09 | -0.11 | |
| EmbeddingConstruction Method=Embedding-based k-NN retrieval2026.05 | -0.21 | -0.14 | |
| IfBenchConstruction Method=General-purpose benchmark2026.05 | -0.64 | -0.43 |