Calibration

C-Eval → HLE

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

HLE is estimated from C-Eval with a linear curve fitted on 5 models measured on both: y = 2.5076·x + -2.0391, R² = 0.74, cross-validated error 2.5 pp. It is used for 1 estimate.

Estimated model	C-Eval	HLE	Source
LongCat-Flash-Lite-Sparse	85.8%	11.1%	estimated ± 2.5 pp, low confidence