KNOWN-RATE SYNTHETIC CALIBRATION · NO MODEL CALLS
Does the estimate track risk?
Set a known failure probability. Add synthetic evidence. Compare the estimate with the truth and measure interval coverage across independent batches.
Known synthetic probabilityUnavailable
Estimated probabilityUnavailable
Observed failuresUnavailable
Orange: known probability. Blue: Bayesian estimate and 95% range. Navy: exact 95% confidence interval for the fixed sample. Add observations to compare the estimates and interval coverage.
Exact-interval coverage · 500 independent batchesUnavailable
95% interval for simulated coverageUnavailable
Mean exact interval widthUnavailable