Calibration
ExploitGym → GDP.pdf
GDP.pdf is estimated from ExploitGym with a Michaelis–Menten curve fitted on 7 models measured on both: y = 0.6003·x / (0.36900 + x), R² = 0.62, cross-validated error 5.9 pp. It is used for 1 estimate.
| Estimated model | ExploitGym | GDP.pdf | Source |
|---|---|---|---|
| Claude Mythos Preview | 17.5% | 19.3% | estimated ± 5.9 pp, low confidence |