Calibration
HMMT Feb 2025 → MMAnswerBench
MMAnswerBench is estimated from HMMT Feb 2025 with a Michaelis–Menten + offset curve fitted on 7 models measured on both: y = 0.0001 + 1.6871·x / (1.00397 + x), R² = 0.33, cross-validated error 2.0 pp. It is used for 3 estimates.
| Estimated model | HMMT Feb 2025 | MMAnswerBench | Source |
|---|---|---|---|
| Granite 4.2 30B | 89.2% | 79.4% | estimated ± 2.0 pp, low confidence |
| Granite 4.2 3B | 66.7% | 67.3% | estimated ± 2.0 pp, low confidence |
| Granite 4.2 8B | 78.3% | 73.9% | estimated ± 2.0 pp, low confidence |