Calibration

AA-Omniscience Accuracy → HLE w/o tools

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

HLE w/o tools is estimated from AA-Omniscience Accuracy with a Michaelis–Menten curve fitted on 27 models measured on both: y = 2.0000·x / (1.65366 + x), R² = 0.75, cross-validated error 7.9 pp. It is used for 0 estimates.