jev-benchmarks

来自 readme

jev-benchmarks Probability-aware evaluation for typed decision models. jev-benchmarks measures more than whether a model selects the right label. It evaluates whether the reported probabilities are calibrated enough to support automation, how much work can be accepted at a fixed error budget, what resources each decision uses, and how long it takes end to …