Calibration

AA ITBench → AA AutomationBench

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

AA AutomationBench is estimated from AA ITBench with a offset logistic curve fitted on 12 models measured on both: y = 0.6440 + (0.3002 − 0.6440) / (1 + exp(−39.17·(x − 0.5677))), R² = 0.48, cross-validated error 5.5 pp. It is used for 1 estimate.

Estimated model	AA ITBench	AA AutomationBench	Source
GLM-5.2	42.7%	64.3%	estimated ± 5.5 pp, low confidence