Calibration

ExploitGym → AutomationBench

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

AutomationBench is estimated from ExploitGym with a Hill curve fitted on 7 models measured on both: y = 0.5450 + (0.3680 − 0.5450)·x^6.00 / (0.18878^6.00 + x^6.00), R² = 0.65, cross-validated error 8.0 pp. It is used for 0 estimates.