Calibration
HLE → AA-GPQA Diamond
AA-GPQA Diamond is estimated from HLE with a offset logistic curve fitted on 46 models measured on both: y = 0.0000 + (0.8957 − 0.0000) / (1 + exp(−12.02·(x − -0.0178))), R² = 0.51, cross-validated error 3.4 pp. It is used for 4 estimates.
| Estimated model | HLE | AA-GPQA Diamond | Source |
|---|---|---|---|
| Agents-A1 | 47.6% | 89.3% | estimated ± 3.4 pp, high confidence |
| DeepSeek V4.1 Flash | 36.8% | 88.7% | estimated ± 3.4 pp, high confidence |
| Qwen3.8-Omni-Flash | 36.5% | 88.7% | estimated ± 3.4 pp, high confidence |
| Step 5 Preview | 46.5% | 89.3% | estimated ± 3.4 pp, high confidence |