ClarusC64/clinical-quad-infection-buffer-lag-coupling-sepsis-transition-v1.5

---
language: en
license: mit
task_categories:
- text-classification
tags:
- clarus
- clarus-v1.5
- quad-coupling
- clinical-trials
- respiratory-collapse
- rescue-path-sequencing
size_categories:
- 1K<n<10K
pretty_name: Clinical Quad Oxygen Demand Buffer Lag Coupling Respiratory Collapse v1.5
---
Clinical Quad Oxygen Demand Buffer Lag Coupling Respiratory Collapse v1.5
What this repo does

This repository contains a Clarus v1.5 benchmark dataset.

The v1.5 layer introduces Counterfactual Rescue Path Sequencing Geometry.

Earlier Clarus layers evaluate:

• system state
• trajectory
• intervention selection
• control sequence correctness
• temporal policy stability
• failure reconstruction
• intervention timing

v1.5 evaluates a deeper reasoning task.

The benchmark asks whether a model can determine:

• whether the system remained recoverable
• which rescue path was correct
• which ordered intervention sequence was required

This turns the task from intervention timing reasoning into rescue path sequencing reasoning.

Core quad

The system state is represented by four interacting variables.

• infection_load
• buffer_capacity
• lag_burden
• coupling_stress

These define the instability position of the system.

In clinical respiratory collapse scenarios they can correspond to:

Quad variable Clinical interpretation
infection_load infection or inflammatory burden
buffer_capacity physiological reserve
lag_burden treatment delay
coupling_stress systemic organ interaction pressure
Prediction targets

The model must predict three outputs.

label_counterfactual_recoverability

Binary classification.

1 = the system remained recoverable through a valid rescue path
0 = rescue paths were no longer viable

label_correct_rescue_path

The stabilizing rescue branch.

Example values:

• antibiotics_fluids_source_control
• oxygen_escalation_niv_airway_support
• diuresis_oxygen_support

label_correct_intervention_sequence

The correct ordered stabilizing intervention chain.

Example:

early_antibiotics > fluids > source_control

Correct actions in the wrong order count as incorrect.

New v1.5 signals

The dataset introduces sequencing geometry variables.

Signal Meaning
primary_rescue_path best stabilizing rescue branch
secondary_rescue_path alternative rescue branch
intervention_sequence_alignment_score similarity to optimal sequence
sequence_dependency_score dependency of later success on earlier actions
pathway_viability_score viability of rescue path
pathway_lockout_risk risk that rescue path is locked out
rescue_order_irreversibility_score severity of ordering mistakes
branch_switch_penalty cost of switching rescue branches

These signals represent the geometry of rescue paths rather than simple interventions.

Row structure
Train rows

Train rows contain system state, rescue geometry signals, and labels.

Columns include:

• scenario_id
• infection_load
• buffer_capacity
• lag_burden
• coupling_stress
• failure_decision_sequence
• primary_rescue_path
• secondary_rescue_path
• intervention_sequence_alignment_score
• sequence_dependency_score
• pathway_viability_score
• pathway_lockout_risk
• rescue_order_irreversibility_score
• branch_switch_penalty

Labels:

• label_counterfactual_recoverability
• label_correct_rescue_path
• label_correct_intervention_sequence

Tester rows

Tester rows include the same structural signals but exclude labels.

The model must predict:

• label_counterfactual_recoverability
• label_correct_rescue_path
• label_correct_intervention_sequence

Files

This repository contains:

data/train.csv
data/tester.csv
scorer.py
dataset_schema.json
benchmark_spec.json
README.md

Evaluation

The scorer evaluates three reasoning tasks simultaneously.

Primary metric

composite_rescue_sequence_success

A prediction is correct only if:

• recoverability classification is correct
• rescue path is correct
• intervention sequence is correct

Secondary metric

false_recoverability_prediction_rate

Fraction of predicted recoverable cases that fail one or more rescue conditions.

Additional metrics

recoverability_label_accuracy
correct_rescue_path_accuracy
correct_intervention_sequence_accuracy

Binary classification metrics are also reported for:

label_counterfactual_recoverability

including:

accuracy
precision
recall
f1
confusion matrix

Diagnostics

The scorer also evaluates performance under difficult rescue conditions.

high_lockout_risk_recoverability_accuracy

Accuracy on cases where rescue path lockout risk is high.

high_sequence_dependency_sequence_accuracy

Sequence prediction accuracy when rescue success depends strongly on ordering.

high_order_irreversibility_misclassification_rate

Fraction of high-order-irreversibility cases incorrectly predicted as recoverable.

high_branch_switch_penalty_composite_accuracy

Composite rescue success rate on cases with high branch-switch penalty.

Example row
scenario_id RESP_042
infection_load 0.81
buffer_capacity 0.32
lag_burden 0.74
coupling_stress 0.66

failure_decision_sequence delayed_antibiotics > respiratory_decline > inflammatory_amplification

primary_rescue_path antibiotics_fluids_source_control
secondary_rescue_path fluids_vasopressors_delayed_source_control

intervention_sequence_alignment_score 0.78
sequence_dependency_score 0.82
pathway_viability_score 0.64
pathway_lockout_risk 0.31
rescue_order_irreversibility_score 0.73
branch_switch_penalty 0.59

label_counterfactual_recoverability 1
label_correct_rescue_path antibiotics_fluids_source_control
label_correct_intervention_sequence early_antibiotics > fluids > source_control
Construction note

This dataset simulates respiratory collapse dynamics during infection escalation.

Signals approximate structural relationships between:

• infection burden
• physiological reserve
• treatment delay
• systemic coupling

The goal is to represent rescue path sequencing geometry rather than clinical guidelines.

Structural Note

Clarus datasets model instability geometry across complex systems.

The quad variables define the system state.
Derived signals represent proximity to collapse boundaries, recovery pathways, and control viability.

This benchmark focuses specifically on rescue path sequencing.

Production Deployment

In operational environments these signals could be generated from real-time system telemetry.

Potential use cases include:

• clinical decision monitoring
• safety-critical system stabilization
• infrastructure failure prevention
• autonomous control oversight

Enterprise & Research Collaboration

Organizations interested in large-scale stability modeling or cross-domain instability detection can contact:

[e-mail]

License

MIT License