File size: 4,508 Bytes
df2b12f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
---
language: en
license: mit
task_categories:
- text-classification
tags:
- clinical-trials
- five-node-cascade
size_categories:
- 1K<n<10K
pretty_name: Clarus Clinical Five Node Shock Cascade Boundary v0.2
---

# What this repo does

This dataset evaluates whether machine learning models can detect shock cascade boundary transitions using both system state and system trajectory.

Earlier Clarus datasets in the v0.1 series tested whether models could classify system state alone.

Clarus v0.2 datasets add a trajectory signal so models must determine not only where the system is, but where it is moving inside state space.

This benchmark therefore tests whether models can read trajectory inside state space and detect approaching instability even when the current state still appears locally stable.

# Five-node core

shock_pressure  
physiological_buffer  
intervention_delay  
organ_coupling  
hemodynamic_reserve

These five interacting variables represent a richer shock-state model for boundary-sensitive collapse detection.

All variables are normalized between 0 and 1.

Together they define the current position of the system inside the stability manifold.

# Trajectory Layer

This dataset includes a trajectory variable called drift_gradient.

drift_gradient measures the directional alignment between the system’s motion vector and the instability gradient of the system’s potential field.

Positive values indicate motion toward cascade.  
Negative values indicate motion toward recovery.

This converts the dataset from a static boundary classifier into a trajectory-aware cascade boundary detection benchmark.

The core test is whether models can read trajectory inside state space rather than relying only on static variable levels.

# Drift definition

Let the system state be represented by vector x(t).

The previous system state is x(t−1).

System motion

Δx = x(t) − x(t−1)

Instability gradient

g = ∇Φ(x)

Drift alignment

drift_gradient = (Δx · g) / (||Δx|| ||g||)

Interpretation

+1 system moving directly toward cascade  
0 neutral movement  
−1 system moving toward recovery

Operational meaning

A positive drift_gradient such as +0.65 means the physiological trajectory is strongly aligned toward the shock boundary even if the current state appears only moderately stressed.

This is the false stability detection problem encoded by v0.2 datasets.

# Prediction target

label_shock_boundary_transition

The label identifies whether the system is undergoing a shock cascade boundary transition.

High drift toward cascade can trigger a positive label even when state variables remain within nominal range.

This lets the dataset test whether models can identify instability from state plus trajectory, not state alone.

# Binary simplification note

Shock boundary progression is continuous in clinical reality.

For benchmarking purposes the outcome is simplified to binary classification:

0 stable regime  
1 shock boundary transition

This makes the benchmark suitable for evaluating early warning performance.

# Row structure

scenario_id  
shock_pressure  
physiological_buffer  
intervention_delay  
organ_coupling  
hemodynamic_reserve  
drift_gradient  
label_shock_boundary_transition

State variables range between 0 and 1.

drift_gradient ranges between -1 and +1.

# Files

data/train.csv  
data/tester.csv  
scorer.py  
README.md

# Evaluation

Binary classification metrics

accuracy  
precision  
recall_cascade_detection  
false_safe_rate  
f1  
confusion_matrix

Primary metric

recall_cascade_detection

This prioritizes whether a model can detect cascade onset before visible deterioration.

false_safe_rate tracks the proportion of true cascade cases incorrectly predicted as safe.

# License

MIT

# Structural Note

Clarus datasets explore instability dynamics in coupled systems.

The v0.2 series adds trajectory awareness to test whether models can detect movement toward failure boundaries rather than only classify static state snapshots.

# Production Deployment

These datasets support experimentation and benchmarking for early instability detection across domains including:

clinical deterioration  
infrastructure instability  
financial contagion  
AI multi-agent systems  
supply chain collapse

# Enterprise & Research Collaboration

Organizations interested in trajectory-aware cascade detection, false stability benchmarking, or operational early warning systems may contact the Clarus project for collaboration.