Calibration

V* → CharXiv

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

CharXiv is estimated from V* with a offset logistic curve fitted on 8 models measured on both: y = 0.0000 + (0.8026 − 0.0000) / (1 + exp(−168.32·(x − 0.6595))), R² = 0.86, cross-validated error 4.9 pp. It is used for 1 estimate.

Estimated model	V*	CharXiv	Source
Step 3.7 Flash	95.3%	80.3%	estimated ± 4.9 pp, high confidence