Jev is a useful zero-shot classifier, but its probabilities can't be calibrated for your data. Calibration depends on your data distribution, which Jev never sees, so treat its outputs as scores and recalibrate them on your own labeled examples.