Calibration

CursorBench 3.1 → Vals LiveCodeBench

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

Vals LiveCodeBench is estimated from CursorBench 3.1 with a linear curve fitted on 6 models measured on both: y = 0.1794·x + 0.7657, R² = 0.36, cross-validated error 2.8 pp. It is used for 0 estimates.