Calibration
GPQA Diamond → ARC-AGI-3
ARC-AGI-3 is estimated from GPQA Diamond with a offset logistic curve fitted on 10 models measured on both: y = 0.0000 + (1.2000 − 0.0000) / (1 + exp(−200.00·(x − 0.9597))), R² = 0.99, cross-validated error 3.8 pp. It is used for 6 estimates.
| Estimated model | GPQA Diamond | ARC-AGI-3 | Source |
|---|---|---|---|
| LLaDA2.2-mini | 44.4% | 0.0% | estimated ± 3.8 pp, medium confidence |
| LongCat-Flash-Lite-Sparse | 69.5% | 0.0% | estimated ± 3.8 pp, medium confidence |
| MiniCPM5-1B | 26.3% | 0.0% | estimated ± 3.8 pp, medium confidence |
| Pareto 26.10 Preview | 92.4% | 0.1% | estimated ± 3.8 pp, high confidence |
| Beam | 90.5% | 0.0% | estimated ± 3.8 pp, high confidence |
| Solar Open 2 | 86.3% | 0.0% | estimated ± 3.8 pp, medium confidence |