Calibration

ARC-AGI-3 → AA-HLE

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

AA-HLE is estimated from ARC-AGI-3 with a Hill curve fitted on 16 models measured on both: y = 0.3042 + (0.7696 − 0.3042)·x^0.20 / (0.49498^0.20 + x^0.20), R² = 0.82, cross-validated error 2.6 pp. It is used for 1 estimate.

Estimated model	ARC-AGI-3	AA-HLE	Source
Grok 4.20	0.1%	40.7%	estimated ± 2.6 pp, medium confidence