Calibration
SuperGPQA → AA-GPQA Diamond
AA-GPQA Diamond is estimated from SuperGPQA with a offset logistic curve fitted on 16 models measured on both: y = 0.7020 + (0.8630 − 0.7020) / (1 + exp(−37.01·(x − 0.5912))), R² = 0.60, cross-validated error 5.0 pp. It is used for 4 estimates.
| Estimated model | SuperGPQA | AA-GPQA Diamond | Source |
|---|---|---|---|
| MiniCPM5-1B | 23.1% | 70.2% | estimated ± 5.0 pp, low confidence |
| Qwen3 235B 2507 | 62.6% | 82.8% | estimated ± 5.0 pp, medium confidence |
| Seed 2.1 Pro | 70.8% | 86.1% | estimated ± 5.0 pp, medium confidence |
| Seed 2.1 Turbo | 67.4% | 85.6% | estimated ± 5.0 pp, medium confidence |