benchgap
Calibration

HLE-Verified → ARC-AGI-2

ARC-AGI-2 is estimated from HLE-Verified with a Michaelis–Menten + offset curve fitted on 5 models measured on both: y = 0.0000 + 2.0000·x / (0.68154 + x), R² = 0.44, cross-validated error 3.1 pp. It is used for 1 estimate.

Estimated modelHLE-VerifiedARC-AGI-2Source
Claude Sonnet 531.0%62.5%estimated ± 3.1 pp, low confidence