Calibration
HLE w/o tools → GPQA Diamond
GPQA Diamond is estimated from HLE w/o tools with a linear curve fitted on 26 models measured on both: y = 0.3802·x + 0.7642, R² = 0.87, cross-validated error 1.9 pp. It is used for 3 estimates.
| Estimated model | HLE w/o tools | GPQA Diamond | Source |
|---|---|---|---|
| Claude Haiku 5.5 | 45.9% | 93.9% | estimated ± 1.9 pp, high confidence |
| Gemma 4 26B A4B | 8.7% | 79.7% | estimated ± 1.9 pp, high confidence |
| Pareto 26.9 | 49.0% | 95.0% | estimated ± 1.9 pp, high confidence |