AbdelStark / jev-benchmarks

Probability-aware evaluation for typed decision models: calibration, selective risk, latency, and reproducible benchmarks.