Calibration

AA Tau3 Banking → GDPval-AA

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

GDPval-AA is estimated from AA Tau3 Banking with a Hill curve fitted on 13 models measured on both: y = 0.2626 + (0.6937 − 0.2626)·x^6.00 / (0.42945^6.00 + x^6.00), R² = 0.78, cross-validated error 8.7 pp. It is used for 0 estimates.