scenario_id string | scenario_text string | claim string | label int64 |
|---|---|---|---|
train_001 | One server failure causes service outage across the entire platform. | The system shows fragility exposure. | 1 |
train_002 | One server failure is absorbed by automatic failover. | The system shows fragility exposure. | 0 |
train_003 | One staff absence causes the clinic to cancel appointments for multiple days. | The clinic shows fragility exposure. | 1 |
train_004 | One staff absence is absorbed through reserve coverage. | The clinic shows fragility exposure. | 0 |
train_005 | One supplier delay causes stockouts across all locations. | The supply chain shows fragility exposure. | 1 |
train_006 | One supplier delay is absorbed through backup suppliers and inventory buffer. | The supply chain shows fragility exposure. | 0 |
train_007 | One database failure prevents all customer transactions. | The platform shows fragility exposure. | 1 |
train_008 | One database failure affects only a limited subsystem. | The platform shows fragility exposure. | 0 |
train_009 | One project blocker prevents all downstream work. | The project shows fragility exposure. | 1 |
train_010 | One project blocker is isolated and alternative work continues. | The project shows fragility exposure. | 0 |
train_011 | One bridge support degradation creates immediate structural risk. | The bridge shows fragility exposure. | 1 |
train_012 | One bridge support degradation is contained within design tolerance. | The bridge shows fragility exposure. | 0 |
train_013 | One customer payment delay creates payroll risk. | The organization shows fragility exposure. | 1 |
train_014 | One customer payment delay is absorbed through reserves. | The organization shows fragility exposure. | 0 |
train_015 | One cooling fan failure causes repeated machine shutdowns. | The machine shows fragility exposure. | 1 |
train_016 | One cooling fan failure is absorbed by redundant cooling. | The machine shows fragility exposure. | 0 |
train_017 | One moderator absence causes platform abuse reports to accumulate rapidly. | The platform shows fragility exposure. | 1 |
train_018 | One moderator absence has little effect because workload is distributed. | The platform shows fragility exposure. | 0 |
train_019 | One missed clinical review leads directly to patient harm. | The care pathway shows fragility exposure. | 1 |
train_020 | One missed clinical review triggers secondary review safeguards. | The care pathway shows fragility exposure. | 0 |
What this dataset does
This dataset tests whether a model can detect fragility exposure.
The task is simple:
Given a scenario and a fragility-exposure claim, predict whether the claim is supported.
Core stability idea
Fragility exposure measures how much damage can result from a small disturbance.
Fragile systems show:
- single points of failure
- low redundancy
- narrow margins
- disproportionate consequences
- poor fault isolation
Robust systems absorb local failures without widespread disruption.
Prediction target
Binary label:
- 1 = fragility exposure is present
- 0 = fragility exposure is not present
Row structure
Each row contains:
- scenario_id
- scenario_text
- claim
- label
Files
- data/train.csv
- data/test.csv
- scorer.py
- README.md
Evaluation
Create a predictions CSV with:
scenario_id,prediction
test_001,1
test_002,0
Run:
python scorer.py --predictions predictions.csv --truth data/test.csv
Structural Note
This dataset is intentionally small.
Its purpose is to test whether a model can identify systems where small perturbations create disproportionately large consequences.
The hidden value is in detecting single-point dependency, missing redundancy, poor isolation, and amplified vulnerability.
License
MIT
- Downloads last month
- 35