Calibration

Terminal-Bench-Science 0.1 → AA Terminal-Bench 4.0

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

AA Terminal-Bench 4.0 is estimated from Terminal-Bench-Science 0.1 with a Michaelis–Menten + offset curve fitted on 5 models measured on both: y = 0.0000 + 1.8554·x / (1.28316 + x), R² = 0.51, cross-validated error 4.1 pp. It is used for 0 estimates.