Calibration

Decision Index 0.3 Knowledge → Decision Index 0.3 Language

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

Decision Index 0.3 Language is estimated from Decision Index 0.3 Knowledge with a Hill curve fitted on 112 models measured on both: y = 0.0475 + (0.8368 − 0.0475)·x^1.51 / (0.26238^1.51 + x^1.51), R² = 0.95, cross-validated error 4.6 pp. It is used for 0 estimates.