Calibration
MMAnswerBench → HMMT Nov 2025
HMMT Nov 2025 is estimated from MMAnswerBench with a offset logistic curve fitted on 9 models measured on both: y = 0.8903 + (0.9443 − 0.8903) / (1 + exp(−164.65·(x − 0.8102))), R² = 0.64, cross-validated error 2.0 pp. It is used for 1 estimate.
| Estimated model | MMAnswerBench | HMMT Nov 2025 | Source |
|---|---|---|---|
| Kimi K2.6 | 86.0% | 94.4% | estimated ± 2.0 pp, high confidence |