ClarusC64/finance-latent-cross-coupling-liquidity-collapse-v0.1

What this repo does

This repository introduces a Clarus dataset for detecting latent instability under cross-coupled conditions in financial systems.

The goal is to identify institutions, markets, or portfolios that may still appear outwardly stable or only mildly abnormal but already contain hidden internal degradation that may activate into overt liquidity collapse once interacting pressures exceed containment.

Core structure

This dataset models a pre-failure geometry built from:

• —latent instability
• —cross-coupling intensity
• —hidden state accumulation
• —activation threshold distance
• —susceptibility and amplification dynamics

Prediction target

The target is binary:

• —1 means hidden instability plus interacting pressures are sufficient to make liquidity collapse likely
• —0 means latent instability remains contained or below meaningful activation threshold

Target column used in this repo:

• —label_liquidity_collapse

Row structure

Each row represents a financial system state described by:

• —observable state
• —latent instability score
• —cross-coupling intensity
• —hidden state index
• —activation threshold distance
• —time under exposure
• —susceptibility factor
• —amplification pressure
• —stabilization buffer

Column meanings

observable_state

What the system appears to show at the surface level.

Examples:

• —stable
• —mild-anomaly
• —no-visible-failure
• —surface-normal

latentinstabilityscore

How much hidden financial instability exists beneath visible conditions.

Range:

0.00 to 1.00

crosscouplingintensity

Strength of interaction between destabilizing variables such as funding fragility, leverage stress, collateral deterioration, redemption pressure, counterparty exposure, and concentration risk.

Range:

0.00 to 1.00

hiddenstateindex

Composite measure of concealed financial degradation or unseen vulnerability.

Range:

0.00 to 1.00

activationthresholddistance

Distance from hidden instability becoming overt liquidity collapse.

Lower means closer to activation.

Range:

0.00 to 1.00

timeunderexposure

Normalized duration score for how long destabilizing conditions have been present.

Range:

0.00 to 1.00

susceptibility_factor

How vulnerable the institution, market, or portfolio is to hidden degradation.

Examples include low cash flexibility, asset mismatch, refinancing dependence, crowding, and fragile funding structure.

Range:

0.00 to 1.00

amplification_pressure

External or internal force increasing the chance that hidden instability will activate.

Examples include spread widening, collateral calls, outflow pressure, volatility spikes, adverse news flow, and market-wide stress propagation.

Range:

0.00 to 1.00

stabilization_buffer

Capacity resisting activation.

Examples include central bank access, cash reserves, unencumbered collateral, diversified funding, hedging support, and strong balance sheet resilience.

Range:

0.00 to 1.00

Default label logic

Standard rule used for this dataset family:

label = 1 if latent_instability_score >= 0.60 AND cross_coupling_intensity >= 0.60 AND hidden_state_index >= 0.60 AND activation_threshold_distance <= 0.35 AND amplification_pressure > stabilization_buffer else 0

Files

• —data/train.csv — labeled examples
• —data/tester.csv — unlabeled evaluation examples
• —scorer.py — finance-hardened production scorer
• —README.md — dataset card

Evaluation

Primary metric:

• —missedlatentactivation_rate

Secondary metric:

• —falseactivationrate

Additional reported metrics:

• —accuracy
• —precision
• —recall
• —f1

The scorer expects binary predictions only.

No score threshold is applied.

The scorer is deterministic and includes audit metadata:

• —scorer version
• —scorer id
• —UTC evaluation timestamp
• —SHA-256 hash of reference file
• —SHA-256 hash of predictions file

Example scorer call

python scorer.py reference.csv predictions.csv
---
language:
- en
license: mit
task_categories:
- text-classification
tags:
- clarus
- latent-instability
- cross-coupling
- hidden-state
- pre-failure-geometry
- finance
- liquidity
size_categories:
- 1K<n<10K
pretty_name: Finance Latent Cross-Coupling Liquidity Collapse v0.1
---

# What this repo does

This repository introduces a Clarus dataset for detecting latent instability under cross-coupled conditions in financial systems.

The goal is to identify institutions, markets, or portfolios that may still appear outwardly stable or only mildly abnormal but already contain hidden internal degradation that may activate into overt liquidity collapse once interacting pressures exceed containment.

# Core structure

This dataset models a pre-failure geometry built from:

- latent instability
- cross-coupling intensity
- hidden state accumulation
- activation threshold distance
- susceptibility and amplification dynamics

# Prediction target

The target is binary:

- `1` means hidden instability plus interacting pressures are sufficient to make liquidity collapse likely
- `0` means latent instability remains contained or below meaningful activation threshold

Target column used in this repo:

- `label_liquidity_collapse`

# Row structure

Each row represents a financial system state described by:

- observable state
- latent instability score
- cross-coupling intensity
- hidden state index
- activation threshold distance
- time under exposure
- susceptibility factor
- amplification pressure
- stabilization buffer

## Column meanings

### observable_state

What the system appears to show at the surface level.

Examples:

- stable
- mild-anomaly
- no-visible-failure
- surface-normal

### latent_instability_score

How much hidden financial instability exists beneath visible conditions.

Range:

`0.00 to 1.00`

### cross_coupling_intensity

Strength of interaction between destabilizing variables such as funding fragility, leverage stress, collateral deterioration, redemption pressure, counterparty exposure, and concentration risk.

Range:

`0.00 to 1.00`

### hidden_state_index

Composite measure of concealed financial degradation or unseen vulnerability.

Range:

`0.00 to 1.00`

### activation_threshold_distance

Distance from hidden instability becoming overt liquidity collapse.

Lower means closer to activation.

Range:

`0.00 to 1.00`

### time_under_exposure

Normalized duration score for how long destabilizing conditions have been present.

Range:

`0.00 to 1.00`

### susceptibility_factor

How vulnerable the institution, market, or portfolio is to hidden degradation.

Examples include low cash flexibility, asset mismatch, refinancing dependence, crowding, and fragile funding structure.

Range:

`0.00 to 1.00`

### amplification_pressure

External or internal force increasing the chance that hidden instability will activate.

Examples include spread widening, collateral calls, outflow pressure, volatility spikes, adverse news flow, and market-wide stress propagation.

Range:

`0.00 to 1.00`

### stabilization_buffer

Capacity resisting activation.

Examples include central bank access, cash reserves, unencumbered collateral, diversified funding, hedging support, and strong balance sheet resilience.

Range:

`0.00 to 1.00`

# Default label logic

Standard rule used for this dataset family:

`label = 1 if latent_instability_score >= 0.60 AND cross_coupling_intensity >= 0.60 AND hidden_state_index >= 0.60 AND activation_threshold_distance <= 0.35 AND amplification_pressure > stabilization_buffer else 0`

# Files

- `data/train.csv` — labeled examples
- `data/tester.csv` — unlabeled evaluation examples
- `scorer.py` — finance-hardened production scorer
- `README.md` — dataset card

# Evaluation

Primary metric:

- missed_latent_activation_rate

Secondary metric:

- false_activation_rate

Additional reported metrics:

- accuracy
- precision
- recall
- f1

The scorer expects binary predictions only.

No score threshold is applied.

The scorer is deterministic and includes audit metadata:

- scorer version
- scorer id
- UTC evaluation timestamp
- SHA-256 hash of reference file
- SHA-256 hash of predictions file

## Example scorer call

python scorer.py reference.csv predictions.csv Where:

reference.csv contains a label_... target column

predictions.csv contains one of: prediction, pred, label, or output

Why this matters

Most benchmark datasets detect visible market stress after it has already surfaced.

This dataset class targets hidden instability before overt liquidity collapse emerges.

That makes it useful for:

early warning

funding stress detection

hidden fragility monitoring

pre-crisis triage

liquidity risk review in systems that still appear contained

License

MIT

Structural Note

This dataset belongs to the Clarus family of stability benchmarks.

It is designed to measure whether a financial system that appears externally stable is already internally unstable due to hidden degradation and interacting variable pressure.

This places it in a pre-failure layer of the Clarus architecture, concerned with concealed activation pressure before overt instability becomes visible.

Production Deployment

This benchmark can support systems that monitor hidden financial risk before obvious liquidity breakdown appears.

Use cases include treasury monitoring, liquidity risk analysis, counterparty stress tracking, portfolio fragility review, and concealed instability detection in funding systems.

Enterprise and Research Collaboration

This dataset class is suitable for adaptation across banks, asset managers, hedge funds, insurers, regulators, market structure research, systemic risk monitoring, and financial stability analytics.

Label check

• —rows 1, 2, 5, 7, and 10 satisfy the default positive rule
• —row 9 stays negative because cross_coupling_intensity is below 0.60