Calibration
GPQA Diamond → SuperGPQA
SuperGPQA is estimated from GPQA Diamond with a Hill curve fitted on 7 models measured on both: y = 0.2193 + (1.2000 − 0.2193)·x^6.00 / (0.86722^6.00 + x^6.00), R² = 0.85, cross-validated error 13.9 pp. It is used for 4 estimates.
| Estimated model | GPQA Diamond | SuperGPQA | Source |
|---|---|---|---|
| Grok 4.20 | 88.5% | 73.9% | estimated ± 13.9 pp, low confidence |
| LLaDA2.2-mini | 44.4% | 23.7% | estimated ± 13.9 pp, low confidence |
| Pareto 26.10 Preview | 92.4% | 80.2% | estimated ± 13.9 pp, low confidence |
| Beam | 90.5% | 77.2% | estimated ± 13.9 pp, low confidence |