Calibration

ExploitGym → Agents' Last Exam

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

Agents' Last Exam is estimated from ExploitGym with a Michaelis–Menten + offset curve fitted on 6 models measured on both: y = 0.1580 + 1.8673·x / (1.32566 + x), R² = 0.72, cross-validated error 10.2 pp. It is used for 0 estimates.