Calibration
CWE-bench v1 → AA Tau3 Banking
AA Tau3 Banking is estimated from CWE-bench v1 with a offset logistic curve fitted on 5 models measured on both: y = 0.1348 + (0.4745 − 0.1348) / (1 + exp(−105.69·(x − 0.3715))), R² = 0.83, cross-validated error 10.7 pp. It is used for 1 estimate.
| Estimated model | CWE-bench v1 | AA Tau3 Banking | Source |
|---|---|---|---|
| Hy4 preview | 53.0% | 47.4% | estimated ± 10.7 pp, low confidence |