Calibration
ARC-AGI-1 → ARC-AGI-2
ARC-AGI-2 is estimated from ARC-AGI-1 with a inverse Michaelis–Menten curve fitted on 34 models measured on both: y = 0.64096·(x − 0.4567) / (0.4567 + 0.8863 − x), R² = 0.95, cross-validated error 5.7 pp. It is used for 1 estimate.
| Estimated model | ARC-AGI-1 | ARC-AGI-2 | Source |
|---|---|---|---|
| GPT-5.1 | 72.8% | 28.3% | estimated ± 5.7 pp, medium confidence |