Calibration
MCP Atlas → CyberGym
CyberGym is estimated from MCP Atlas with a Michaelis–Menten curve fitted on 11 models measured on both: y = 1.3377·x / (0.59667 + x), R² = 0.63, cross-validated error 9.9 pp. It is used for 2 estimates.
| Estimated model | MCP Atlas | CyberGym | Source |
|---|---|---|---|
| LLaDA2.2-flash | 46.2% | 58.4% | estimated ± 9.9 pp, medium confidence |
| Solar Open 2 | 58.2% | 66.1% | estimated ± 9.9 pp, medium confidence |