v0.2.0

Added

• computeBayesianScore(): v2 Bayesian log-odds scoring with per-layer sensitivity/specificity
• LayerBayesianParams type: { sensitivity, specificity } per layer
• BayesianLayerConfig type: extends LayerConfig with a bayesian field
• prior field on DomainConfig: P(reference is real given domain)
• bayesianThreshold field on DomainConfig: minimum posterior to VERIFY
• Domain priors: ACADEMIC=0.72, NEWS=0.75, GOVERNMENT=0.82, GENERAL=0.45
• Bayesian thresholds: ACADEMIC=0.82, NEWS=0.65, GOVERNMENT=0.72, GENERAL=0.68
• 17 new Bayesian tests (126 total)

Changed

• DomainConfig.layers type widened from LayerConfig[] to BayesianLayerConfig[]
• DOMAIN_CONFIGS updated with calibrated sensitivity/specificity for all layers

Notes

• The v1 API (computeDomainAwareScore, DomainConfig.threshold) is fully backward-compatible.