Calibration
VITA-Bench → QwenClawBench
QwenClawBench is estimated from VITA-Bench with a offset logistic curve fitted on 6 models measured on both: y = 0.5204 + (0.6396 − 0.5204) / (1 + exp(−200.00·(x − 0.4465))), R² = 0.96, cross-validated error 2.0 pp. It is used for 6 estimates.
| Estimated model | VITA-Bench | QwenClawBench | Source |
|---|---|---|---|
| Agents-A1 | 38.8% | 52.0% | estimated ± 2.0 pp, medium confidence |
| Agents-A1-4B | 40.3% | 52.0% | estimated ± 2.0 pp, medium confidence |
| Claude Sonnet 4.5 | 17.0% | 52.0% | estimated ± 2.0 pp, low confidence |
| DeepSeek V3.2 | 18.5% | 52.0% | estimated ± 2.0 pp, low confidence |
| GLM-4.7 | 15.5% | 52.0% | estimated ± 2.0 pp, low confidence |
| LongCat-Flash-Lite-Sparse | 21.7% | 52.0% | estimated ± 2.0 pp, low confidence |