Calibration
NL2Repo → AA Coding Index
AA Coding Index is estimated from NL2Repo with a Hill curve fitted on 13 models measured on both: y = 0.3535 + (0.7549 − 0.3535)·x^6.00 / (0.38805^6.00 + x^6.00), R² = 0.83, cross-validated error 4.6 pp. It is used for 9 estimates.
| Estimated model | NL2Repo | AA Coding Index | Source |
|---|---|---|---|
| Claude Opus 4.5 | 43.2% | 61.7% | estimated ± 4.6 pp, high confidence |
| dots3-note Preview | 49.8% | 68.1% | estimated ± 4.6 pp, high confidence |
| Ornith-1.0-35B | 34.6% | 48.8% | estimated ± 4.6 pp, high confidence |
| Ornith-1.0-397B | 48.2% | 66.9% | estimated ± 4.6 pp, high confidence |
| Ornith-1.0-9B | 27.2% | 39.6% | estimated ± 4.6 pp, medium confidence |
| Ornith-1.5-9B | 32.4% | 45.5% | estimated ± 4.6 pp, high confidence |
| Qwen 3.6 Max (preview) | 42.9% | 61.3% | estimated ± 4.6 pp, high confidence |
| Seed 2.1 Pro | 47.0% | 65.8% | estimated ± 4.6 pp, high confidence |
| Seed 2.1 Turbo | 43.7% | 62.3% | estimated ± 4.6 pp, high confidence |