Dataset Viewer
Auto-converted to Parquet Duplicate
uid
stringclasses
10 values
dataset_id
stringclasses
1 value
line_of_business
stringclasses
4 values
claim_type
stringclasses
9 values
fraud_flags_count
int64
1
5
flag_strength
stringclasses
3 values
claim_evidence_strength
stringclasses
3 values
investigation_action
stringclasses
2 values
final_fraud_outcome
stringclasses
2 values
triage_summary
stringclasses
10 values
question
stringclasses
1 value
ground_truth_label
stringclasses
2 values
ground_truth_rationale
stringclasses
10 values
outcome_signal
stringclasses
5 values
source_citation
stringclasses
1 value
INSF-001
insurance-fraud-flag-coherence-risk-v0.1
auto
injury
1
low
strong
fast_track
not_fraud
Few weak flags with strong evidence.
Is fraud-flag coherence intact. Answer coherent or incoherent.
coherent
Action matched low suspicion.
low_false_positive
Synthetic
INSF-002
insurance-fraud-flag-coherence-risk-v0.1
auto
injury
4
high
weak
investigate
fraud
Many strong flags and weak evidence.
Is fraud-flag coherence intact. Answer coherent or incoherent.
coherent
Investigation justified and confirmed.
fraud_caught
Synthetic
INSF-003
insurance-fraud-flag-coherence-risk-v0.1
property
storm
3
medium
medium
investigate
not_fraud
Moderate flags but legitimate loss.
Is fraud-flag coherence intact. Answer coherent or incoherent.
incoherent
Investigation over-triggered.
false_positive_cost
Synthetic
INSF-004
insurance-fraud-flag-coherence-risk-v0.1
property
fire
1
low
weak
fast_track
fraud
Weak flag but weak evidence and fraud later confirmed.
Is fraud-flag coherence intact. Answer coherent or incoherent.
incoherent
Triage missed fraud risk.
fraud_missed
Synthetic
INSF-005
insurance-fraud-flag-coherence-risk-v0.1
health
procedure
2
low
strong
fast_track
not_fraud
Routine claim with mild flags.
Is fraud-flag coherence intact. Answer coherent or incoherent.
coherent
Fast track appropriate.
low_false_positive
Synthetic
INSF-006
insurance-fraud-flag-coherence-risk-v0.1
commercial
liability
5
high
weak
investigate
fraud
High flags and weak support.
Is fraud-flag coherence intact. Answer coherent or incoherent.
coherent
High suspicion acted on.
fraud_caught
Synthetic
INSF-007
insurance-fraud-flag-coherence-risk-v0.1
auto
theft
3
medium
strong
investigate
not_fraud
Strong evidence but investigated heavily.
Is fraud-flag coherence intact. Answer coherent or incoherent.
incoherent
Flags not strong enough to justify delay.
customer_friction
Synthetic
INSF-008
insurance-fraud-flag-coherence-risk-v0.1
property
water_damage
4
high
medium
fast_track
fraud
High flags but fast-tracked.
Is fraud-flag coherence intact. Answer coherent or incoherent.
incoherent
Action failed to match strong suspicion.
fraud_missed
Synthetic
INSF-009
insurance-fraud-flag-coherence-risk-v0.1
health
disability
4
high
weak
investigate
fraud
High suspicion and weak support.
Is fraud-flag coherence intact. Answer coherent or incoherent.
coherent
Investigation aligned.
fraud_caught
Synthetic
INSF-010
insurance-fraud-flag-coherence-risk-v0.1
auto
collision
2
low
strong
fast_track
not_fraud
Low suspicion and strong evidence.
Is fraud-flag coherence intact. Answer coherent or incoherent.
coherent
No over-triage.
low_false_positive
Synthetic

What this repo is for

Detect fraud triage errors.

Focus

flags vs evidence

investigate vs fast track choice

false positives

missed fraud

Why it matters Fraud cost grows quietly. Over-triage also burns trust and time.

Downloads last month
16