Calibration
AA EnterpriseOps-Gym → AA-AnalystAgent
AA-AnalystAgent is estimated from AA EnterpriseOps-Gym with a linear curve fitted on 6 models measured on both: y = 1.8698·x + -0.4849, R² = 0.97, cross-validated error 3.6 pp. It is used for 4 estimates.
| Estimated model | AA EnterpriseOps-Gym | AA-AnalystAgent | Source |
|---|---|---|---|
| Gemini 3.5 Flash | 50.1% | 45.2% | estimated ± 3.6 pp, medium confidence |
| Grok 4.6 | 48.3% | 41.8% | estimated ± 3.6 pp, medium confidence |
| Muse Glimmer 30B | 34.7% | 16.4% | estimated ± 3.6 pp, medium confidence |
| Qwen3.8-27B | 44.2% | 34.2% | estimated ± 3.6 pp, medium confidence |