uid stringclasses 10
values | dataset_id stringclasses 1
value | line_of_business stringclasses 4
values | claim_type stringclasses 10
values | reported_loss_severity stringclasses 4
values | objective_evidence_strength stringclasses 3
values | adjuster_initial_reserve stringclasses 4
values | coverage_clarity stringclasses 2
values | settlement_recommendation stringclasses 4
values | payout_trend stringclasses 2
values | claims_summary stringclasses 10
values | question stringclasses 1
value | ground_truth_label stringclasses 2
values | ground_truth_rationale stringclasses 10
values | outcome_signal stringclasses 6
values | source_citation stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
INSC-001 | insurance-claim-assessment-coherence-risk-v0.1 | auto | collision | moderate | strong | adequate | clear | fair | stable | Evidence supports moderate payout. | Is claim-assessment coherence intact. Answer coherent or incoherent. | coherent | Reserve and payout match evidence. | efficient_settlement | Synthetic |
INSC-002 | insurance-claim-assessment-coherence-risk-v0.1 | property | water_damage | high | weak | high | unclear | deny | upward | High reserve set on weak evidence. | Is claim-assessment coherence intact. Answer coherent or incoherent. | incoherent | Reserve does not match evidence. | over_reserving_risk | Synthetic |
INSC-003 | insurance-claim-assessment-coherence-risk-v0.1 | health | procedure | low | strong | low | clear | approve | stable | Low claim with strong evidence. | Is claim-assessment coherence intact. Answer coherent or incoherent. | coherent | Decision matches evidence. | efficient | Synthetic |
INSC-004 | insurance-claim-assessment-coherence-risk-v0.1 | commercial | liability | high | medium | low | clear | low_offer | upward | Low reserve despite high exposure. | Is claim-assessment coherence intact. Answer coherent or incoherent. | incoherent | Under-reserving given exposure. | reserve_shortfall_risk | Synthetic |
INSC-005 | insurance-claim-assessment-coherence-risk-v0.1 | auto | injury | medium | medium | medium | clear | fair | stable | Reserve matches injury evidence. | Is claim-assessment coherence intact. Answer coherent or incoherent. | coherent | Assessment consistent. | stable | Synthetic |
INSC-006 | insurance-claim-assessment-coherence-risk-v0.1 | property | fire | high | strong | low | clear | low_offer | upward | Strong evidence but low reserve. | Is claim-assessment coherence intact. Answer coherent or incoherent. | incoherent | Reserve too low for facts. | reserve_shortfall_risk | Synthetic |
INSC-007 | insurance-claim-assessment-coherence-risk-v0.1 | health | disability | medium | strong | medium | clear | approve | stable | Approval matches evidence. | Is claim-assessment coherence intact. Answer coherent or incoherent. | coherent | Reserve aligned. | stable | Synthetic |
INSC-008 | insurance-claim-assessment-coherence-risk-v0.1 | commercial | workers_comp | medium | weak | medium | unclear | deny | upward | Denial with unclear coverage and weak evidence triage. | Is claim-assessment coherence intact. Answer coherent or incoherent. | incoherent | Coverage unclear yet denial chosen. | litigation_risk | Synthetic |
INSC-009 | insurance-claim-assessment-coherence-risk-v0.1 | auto | theft | low | strong | low | clear | approve | stable | Straightforward theft claim. | Is claim-assessment coherence intact. Answer coherent or incoherent. | coherent | Clear coverage and evidence. | efficient | Synthetic |
INSC-010 | insurance-claim-assessment-coherence-risk-v0.1 | property | storm | medium | medium | high | clear | fair | stable | Reserve conservative but evidence supports. | Is claim-assessment coherence intact. Answer coherent or incoherent. | coherent | Reserve within reasonable band. | stable | Synthetic |
What this repo is for
Detect claim handling drift.
Focus
evidence strength vs reserve
coverage clarity vs deny or pay choice
under-reserving and over-reserving risk
Why it matters Bad assessment creates two costs. Overpay. Or litigation.
- Downloads last month
- 29