benchgap
Calibration

HLE → AA-Omniscience Accuracy

AA-Omniscience Accuracy is estimated from HLE with a linear curve fitted on 48 models measured on both: y = 0.6931·x + 0.0688, R² = 0.51, cross-validated error 10.5 pp. It is used for 1 estimate.

Estimated modelHLEAA-Omniscience AccuracySource
Agents-A147.6%39.9%estimated ± 10.5 pp, low confidence