Calibration

BioMysteryBench (human-difficult) → Artificial Analysis Intelligence Index

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

Artificial Analysis Intelligence Index is estimated from BioMysteryBench (human-difficult) with a offset logistic curve fitted on 6 models measured on both: y = 0.2616 + (0.5142 − 0.2616) / (1 + exp(−200.00·(x − 0.4336))), R² = 0.73, cross-validated error 12.9 pp. It is used for 0 estimates.