benchgap
Calibration

HLE → Vals MMLU-Pro

Vals MMLU-Pro is estimated from HLE with a linear curve fitted on 28 models measured on both: y = 0.1511·x + 0.8001, R² = 0.37, cross-validated error 2.5 pp. It is used for 1 estimate.

Estimated modelHLEVals MMLU-ProSource
Agents-A147.6%87.2%estimated ± 2.5 pp, medium confidence