Calibration
DeepPlanning → τ³-bench results
τ³-bench results is estimated from DeepPlanning with a Michaelis–Menten curve fitted on 6 models measured on both: y = 0.7234·x / (0.01472 + x), R² = 0.73, cross-validated error 1.4 pp. It is used for 1 estimate.
| Estimated model | DeepPlanning | τ³-bench results | Source |
|---|---|---|---|
| Qwen3.7 Plus | 62.3% | 70.7% | estimated ± 1.4 pp, low confidence |