Calibration
ARC-AGI-2 → CritPt
CritPt is estimated from ARC-AGI-2 with a offset logistic curve fitted on 37 models measured on both: y = 0.0909 + (1.2000 − 0.0909) / (1 + exp(−4.58·(x − 1.2382))), R² = 0.63, cross-validated error 5.7 pp. It is used for 2 estimates.
| Estimated model | ARC-AGI-2 | CritPt | Source |
|---|---|---|---|
| Claude Sonnet 4.5 Thinking | 13.6% | 9.8% | estimated ± 5.7 pp, medium confidence |
| dots3-note Preview | 81.4% | 23.0% | estimated ± 5.7 pp, medium confidence |