benchgap
Calibration

AA AutomationBench → AA Tau3 Banking

AA Tau3 Banking is estimated from AA AutomationBench with a linear curve fitted on 13 models measured on both: y = 0.4792·x + 0.1781, R² = 0.80, cross-validated error 7.1 pp. It is used for 1 estimate.

Estimated modelAA AutomationBenchAA Tau3 BankingSource
Claude Haiku 5.535.4%34.8%estimated ± 7.1 pp, medium confidence