Calibration
Vals GPQA Diamond → AA-HLE
AA-HLE is estimated from Vals GPQA Diamond with a offset logistic curve fitted on 47 models measured on both: y = 0.1067 + (0.8992 − 0.1067) / (1 + exp(−11.78·(x − 0.9516))), R² = 0.69, cross-validated error 7.4 pp. It is used for 7 estimates.
| Estimated model | Vals GPQA Diamond | AA-HLE | Source |
|---|---|---|---|
| Claude Haiku 4.5 | 72.2% | 15.6% | estimated ± 7.4 pp, medium confidence |
| Gemini 3.1 Flash-Lite | 81.1% | 23.4% | estimated ± 7.4 pp, medium confidence |
| GLM-4.5 | 72.2% | 15.6% | estimated ± 7.4 pp, medium confidence |
| Laguna M.1 | 27.0% | 10.7% | estimated ± 7.4 pp, low confidence |
| Laguna XS.2 | 55.1% | 11.4% | estimated ± 7.4 pp, medium confidence |
| MiMo-V2.5 | 81.6% | 24.0% | estimated ± 7.4 pp, medium confidence |
| Qwen3.5 Flash | 82.8% | 25.7% | estimated ± 7.4 pp, medium confidence |