uid stringclasses 10
values | dataset_id stringclasses 1
value | flag_rate stringclasses 3
values | confirmed_fraud_rate stringclasses 4
values | false_positive_rate stringclasses 3
values | customer_complaints stringclasses 3
values | loss_trend stringclasses 2
values | model_confidence_alignment stringclasses 3
values | case_summary stringclasses 9
values | question stringclasses 1
value | ground_truth_label stringclasses 2
values | ground_truth_rationale stringclasses 10
values | outcome_signal stringclasses 7
values | source_citation stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
BNKFR-001 | banking-fraud-detection-coherence-risk-v0.1 | normal | aligned | low | low | stable | aligned | Flags match real fraud. | Is fraud-detection coherence intact. Answer coherent or incoherent. | coherent | Detection matches confirmed fraud. | stable_losses | Synthetic |
BNKFR-002 | banking-fraud-detection-coherence-risk-v0.1 | high | low | high | high | stable | misaligned | Many flags but little fraud. | Is fraud-detection coherence intact. Answer coherent or incoherent. | incoherent | High false positives. | customer_churn_risk | Synthetic |
BNKFR-003 | banking-fraud-detection-coherence-risk-v0.1 | low | high | low | low | upward | misaligned | Fraud rising but few flags. | Is fraud-detection coherence intact. Answer coherent or incoherent. | incoherent | Missed fraud. | loss_growth | Synthetic |
BNKFR-004 | banking-fraud-detection-coherence-risk-v0.1 | normal | aligned | low | low | stable | aligned | Balanced. | Is fraud-detection coherence intact. Answer coherent or incoherent. | coherent | Aligned signals. | stable | Synthetic |
BNKFR-005 | banking-fraud-detection-coherence-risk-v0.1 | normal | aligned | low | low | stable | aligned | Stable. | Is fraud-detection coherence intact. Answer coherent or incoherent. | coherent | No mismatch. | stable | Synthetic |
BNKFR-006 | banking-fraud-detection-coherence-risk-v0.1 | high | moderate | moderate | moderate | stable | partial | Flagging slightly high. | Is fraud-detection coherence intact. Answer coherent or incoherent. | incoherent | Friction rising without fraud benefit. | customer_friction | Synthetic |
BNKFR-007 | banking-fraud-detection-coherence-risk-v0.1 | low | high | low | low | upward | misaligned | Fraud under-detected. | Is fraud-detection coherence intact. Answer coherent or incoherent. | incoherent | Model missing fraud. | loss_risk | Synthetic |
BNKFR-008 | banking-fraud-detection-coherence-risk-v0.1 | normal | aligned | low | low | stable | aligned | Aligned. | Is fraud-detection coherence intact. Answer coherent or incoherent. | coherent | Balanced system. | stable | Synthetic |
BNKFR-009 | banking-fraud-detection-coherence-risk-v0.1 | normal | aligned | low | low | stable | aligned | Stable. | Is fraud-detection coherence intact. Answer coherent or incoherent. | coherent | Aligned. | stable | Synthetic |
BNKFR-010 | banking-fraud-detection-coherence-risk-v0.1 | high | low | high | high | stable | misaligned | Over-flagging. | Is fraud-detection coherence intact. Answer coherent or incoherent. | incoherent | Too many false alerts. | customer_churn | Synthetic |
What this repo is for
Detect when fraud systems drift out of balance.
Focus
flags vs confirmed fraud
false positives
missed fraud
Why it matters Fraud models fail slowly, then suddenly.
- Downloads last month
- 18