{"base_learner_scores":{"dbscan_is_anom":0,"dbscan_score":0.0,"n_measurements":0,"p_classifier":0.2306,"p_corroboration":0.0004,"p_measurement":0.0},"contributions":{"dbscan_is_anom":{"coefficient":0.3243,"log_odds_contribution":0.0,"value":0.0},"dbscan_score":{"coefficient":0.1359,"log_odds_contribution":0.0,"value":0.0},"p_classifier":{"coefficient":1.9479,"log_odds_contribution":0.4492,"value":0.2306},"p_corroboration":{"coefficient":0.0007,"log_odds_contribution":0.0,"value":0.0004},"p_measurement":{"coefficient":0.1215,"log_odds_contribution":0.0,"value":0.0}},"country":"IR","date":"2026-10-04","honest_caveats":["p_classifier is IN-SAMPLE from v3.3, which itself was trained on the same labels. Proper stacked CV would use out-of-fold v3.3 scores; this baseline is an honest starting point \u2014 any genuine lift here is over a self-confident base learner.","Per-measurement classifier has AUC ~1.0 by reconstructing the labeling rule, so its contribution to the meta-learner may be inflated; weight magnitudes printed below.","DBSCAN scores cover 92.6% of training rows; the remaining 7.4% fall back to 0."],"intercept":-0.3586,"meta_learner_probability":0.2953,"passed_promote_gates":false,"schema":"voidly-stacking-score/v1","winner":"logistic_regression"}