Calibration

MCP-Tasks → Claw-Eval

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

Claw-Eval is estimated from MCP-Tasks with a offset logistic curve fitted on 5 models measured on both: y = 0.2098 + (0.5840 − 0.2098) / (1 + exp(−136.71·(x − 0.5790))), R² = 0.87, cross-validated error 3.4 pp. It is used for 0 estimates.