benchgap
Calibration

CursorBench 3.2 → DeepSWE

DeepSWE is estimated from CursorBench 3.2 with a Michaelis–Menten curve fitted on 10 models measured on both: y = 2.0000·x / (1.34669 + x), R² = 0.32, cross-validated error 8.4 pp. It is used for 0 estimates.