Calibration
MCP Atlas → Toolathlon-Verified
Toolathlon-Verified is estimated from MCP Atlas with a inverse Michaelis–Menten curve fitted on 10 models measured on both: y = 1.04353·(x − 0.0000) / (0.0000 + 2.0000 − x), R² = 0.54, cross-validated error 9.3 pp. It is used for 2 estimates.
| Estimated model | MCP Atlas | Toolathlon-Verified | Source |
|---|---|---|---|
| LLaDA2.2-flash | 46.2% | 31.4% | estimated ± 9.3 pp, low confidence |
| Solar Open 2 | 58.2% | 42.8% | estimated ± 9.3 pp, medium confidence |