Calibration
APEX-Agents-AA → ApprenticeBench
ApprenticeBench is estimated from APEX-Agents-AA with a linear curve fitted on 5 models measured on both: y = 1.1891·x + -0.3007, R² = 0.46, cross-validated error 5.0 pp. It is used for 10 estimates.
| Estimated model | APEX-Agents-AA | ApprenticeBench | Source |
|---|---|---|---|
| Apodex 1.1 | 31.2% | 7.0% | estimated ± 5.0 pp, low confidence |
| Apodex 1.1 Mini | 31.2% | 7.0% | estimated ± 5.0 pp, low confidence |
| Claude Opus 4.6 (Adaptive) | 33.0% | 9.2% | estimated ± 5.0 pp, low confidence |
| GLM-5.2 | 33.7% | 10.0% | estimated ± 5.0 pp, low confidence |
| GPT-5.4 mini | 28.2% | 3.5% | estimated ± 5.0 pp, low confidence |
| GPT-5.4 nano | 24.9% | 0.0% | estimated ± 5.0 pp, low confidence |
| GPT-OSS 20B | 0.7% | 0.0% | estimated ± 5.0 pp, low confidence |
| Nemotron 3 Super 100B | 1.8% | 0.0% | estimated ± 5.0 pp, low confidence |
| Qwen3.7 Plus | 22.4% | 0.0% | estimated ± 5.0 pp, low confidence |
| Qwen3.8 Max Preview | 42.4% | 20.3% | estimated ± 5.0 pp, low confidence |