ClarusC64/clinical-quad-infection-buffer-lag-coupling-sepsis-transition-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:

team@clarusinvariant.com

License

MIT License