Calibration
AA-AnalystAgent → τ³-Bench Banking
τ³-Bench Banking is estimated from AA-AnalystAgent with a Michaelis–Menten curve fitted on 7 models measured on both: y = 0.7615·x / (0.33176 + x), R² = 0.94, cross-validated error 4.5 pp. It is used for 5 estimates.
| Estimated model | AA-AnalystAgent | τ³-Bench Banking | Source |
|---|---|---|---|
| Claude Sonnet 5.5 (max with fallback) | 57.5% | 48.3% | estimated ± 4.5 pp, medium confidence |
| GPT-6.1 Sol (max) | 50.0% | 45.8% | estimated ± 4.5 pp, medium confidence |
| Claude Fable 5 (with fallback) | 48.8% | 45.3% | estimated ± 4.5 pp, medium confidence |
| Gemini 3.7 Flash (high) | 60.0% | 49.0% | estimated ± 4.5 pp, low confidence |
| Claude Opus 5 (max) | 53.8% | 47.1% | estimated ± 4.5 pp, medium confidence |