Datasets:
scenario_id string | pressure float64 | coupling_strength float64 | drift_gradient float64 | propagation_factor float64 | boundary_distance float64 | label_coupling_explosion int64 |
|---|---|---|---|---|---|---|
CE01 | 0.81 | 0.92 | 0.46 | 0.88 | 0.21 | 1 |
CE02 | 0.78 | 0.89 | 0.43 | 0.85 | 0.24 | 1 |
CE03 | 0.74 | 0.87 | 0.41 | 0.82 | 0.27 | 1 |
CE04 | 0.71 | 0.84 | 0.38 | 0.79 | 0.3 | 1 |
CE05 | 0.68 | 0.82 | 0.35 | 0.76 | 0.33 | 1 |
CE06 | 0.52 | 0.41 | -0.1 | 0.38 | 0.66 | 0 |
CE07 | 0.49 | 0.37 | -0.15 | 0.34 | 0.7 | 0 |
CE08 | 0.45 | 0.33 | -0.19 | 0.3 | 0.74 | 0 |
CE09 | 0.42 | 0.29 | -0.23 | 0.26 | 0.79 | 0 |
CE10 | 0.38 | 0.25 | -0.28 | 0.22 | 0.83 | 0 |
CE — Coupling Explosion Probe
This dataset is a periodic-table probe designed to detect coupling explosion in complex systems.
Coupling explosion occurs when subsystem interactions become so strong that disturbances rapidly propagate through the entire system. Small perturbations can amplify quickly, producing system-wide instability.
This probe evaluates whether models can detect the structural conditions that enable this amplification.
Geometry Symbol
CE
Coupling Explosion
Family: Interaction Amplification Failures
Mechanism
Coupling explosion emerges when subsystem coupling strength becomes high enough that disturbances propagate faster than the system can absorb them.
Typical characteristics include:
high subsystem coupling
positive drift toward instability
strong propagation factors
shrinking boundary distance
These conditions create rapid system-wide instability.
Signals
pressure
overall system stress level
coupling_strength
degree of interaction between subsystems
drift_gradient
direction of system trajectory
propagation_factor
how strongly disturbances spread
boundary_distance
distance to instability threshold
Prediction Target
label_coupling_explosion
1 = coupling-driven instability forming
0 = system interactions remain contained
Row Structure
Each row represents a snapshot of system state.
Signals are normalized between 0 and 1.
Higher values generally indicate stronger instability forces.
Files
data/train.csv
data/tester.csv
scorer.py
Evaluation
Predictions must include:
scenario_id
prediction
Metrics reported:
accuracy
precision
recall
f1
Instability Geometry Probes
This dataset is part of the instability geometry probe library.
CC — Cascade Collapse
BE — Buffer Exhaustion
DF — Delayed Feedback
OE — Oscillatory Instability
BD — Boundary Drift Collapse
CE — Coupling Explosion
These probes isolate fundamental instability geometries across complex systems.
License
MIT
- Downloads last month
- 42