Calibration

HLE → ARC-AGI-1

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

ARC-AGI-1 is estimated from HLE with a Michaelis–Menten curve fitted on 15 models measured on both: y = 1.6764·x / (0.45106 + x), R² = 0.39, cross-validated error 11.9 pp. It is used for 1 estimate.

Estimated model	HLE	ARC-AGI-1	Source
Agents-A1	47.6%	86.1%	estimated ± 11.9 pp, low confidence