Dataset Viewer
Auto-converted to Parquet Duplicate
scenario_id
string
pressure
float64
buffer_capacity
float64
coupling_strength
float64
surface_stability
float64
trajectory_drift
float64
hidden_strain
float64
label_false_stability
int64
cfs_train_001
0.44
0.76
0.36
0.82
-0.03
0.12
0
cfs_train_002
0.48
0.72
0.39
0.8
-0.01
0.15
0
cfs_train_003
0.52
0.69
0.42
0.79
0.02
0.34
1
cfs_train_004
0.56
0.65
0.45
0.78
0.05
0.41
1
cfs_train_005
0.6
0.61
0.49
0.77
0.08
0.47
1
cfs_train_006
0.46
0.74
0.38
0.81
-0.02
0.14
0
cfs_train_007
0.5
0.7
0.41
0.8
0.01
0.29
1
cfs_train_008
0.54
0.66
0.44
0.78
0.04
0.38
1
cfs_train_009
0.47
0.73
0.37
0.82
-0.03
0.13
0
cfs_train_010
0.59
0.62
0.48
0.76
0.07
0.45
1

Clinical False Stability Sepsis Detection Overview

This dataset tests whether a model can detect false stability in a clinical system.

False stability occurs when a system appears stable based on surface indicators, while deeper structural signals reveal that the system is already drifting toward collapse.

In real clinical environments this phenomenon appears frequently during severe infections such as sepsis. A patient's vital signs may temporarily stabilize even while the underlying physiological system continues to degrade.

The benchmark asks whether models can detect hidden instability before visible deterioration occurs.

The false stability problem

Many complex systems show periods where observable indicators appear normal even though structural instability is increasing.

Examples include:

temporary stabilization of blood pressure despite rising inflammatory stress short-term improvement in oxygenation while organ reserve is falling stable vital signs masking accumulating systemic strain

These situations represent false equilibrium states.

Detecting these states early is critical for preventing sudden collapse.

Core system geometry

Each row represents a simplified clinical dynamical system described using structural variables.

pressure Overall physiological stress on the system.

buffer_capacity Remaining physiological reserve available to absorb stress.

coupling_strength Degree of interaction between physiological subsystems.

surface_stability Observable system stability based on surface indicators.

trajectory_drift Directional movement of the system toward stability or collapse.

hidden_strain Accumulated instability that is not visible through surface measurements.

Together these variables describe the difference between observable stability and structural stability.

Clinical geometry mapping

The structural variables represent simplified abstractions of real clinical measurements.

Structural Variable Clinical Interpretation Possible Real Signals pressure Overall physiological stress heart rate, mean arterial pressure, lactate buffer_capacity Remaining physiological reserve organ function markers, oxygen reserve coupling_strength Interaction between physiological subsystems inflammatory signaling, organ cross-talk surface_stability Observable stability of vital signs blood pressure stability, respiratory stability trajectory_drift Direction of system change over time trend of lactate, trend of organ function hidden_strain Instability not visible in surface signals inflammatory burden, metabolic stress

These variables allow the dataset to represent system-level instability rather than isolated measurements.

Prediction target

label_false_stability

0 = stability is genuine 1 = stability is false (hidden drift toward collapse)

The model must determine whether the system's apparent stability is structurally reliable or misleading.

Row structure

Each row represents a synthetic clinical scenario.

Columns:

scenario_id pressure buffer_capacity coupling_strength surface_stability trajectory_drift hidden_strain

Training rows include the label. Tester rows omit the label.

Evaluation

The scoring script reports the following metrics.

accuracy precision recall f1 specificity negative predictive value (npv)

Primary metric recall

Secondary metric f1

Recall is prioritized because detecting hidden instability early is more important than avoiding false alarms.

Why this benchmark matters

False stability is a common failure mode in complex systems.

Systems may appear stable until a threshold is crossed, after which deterioration can occur rapidly.

Detecting false stability allows clinicians and monitoring systems to intervene before sudden systemic collapse occurs.

The benchmark therefore tests whether models can reason about latent instability in dynamical systems rather than relying only on surface indicators.

Structural note

This dataset exposes system geometry without revealing the generator used to produce the scenarios.

The goal is to evaluate whether models can identify hidden instability patterns rather than memorizing statistical patterns.

Clarus Stability Geometry Benchmarks

This dataset is part of a broader benchmark family exploring instability and recovery in complex systems.

Related probes include:

clinical-compensation-collapse-sepsis-v1 clinical-fork-point-sepsis-transition-v1 clinical-organ-failure-cascade-v1 clinical-recovery-window-sepsis-v1 clinical-intervention-alignment-sepsis-v1 clinical-recovery-stability-sepsis-v1

Together these benchmarks map the lifecycle of instability in clinical dynamical systems.

License

MIT

Downloads last month
15